WO2010022655A1 - 一种搜索方法和*** - Google Patents
一种搜索方法和*** Download PDFInfo
- Publication number
- WO2010022655A1 WO2010022655A1 PCT/CN2009/073534 CN2009073534W WO2010022655A1 WO 2010022655 A1 WO2010022655 A1 WO 2010022655A1 CN 2009073534 W CN2009073534 W CN 2009073534W WO 2010022655 A1 WO2010022655 A1 WO 2010022655A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- search
- user
- interest model
- personalized
- results
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 151
- 238000004422 calculation algorithm Methods 0.000 claims description 115
- 239000013598 vector Substances 0.000 claims description 59
- 238000012545 processing Methods 0.000 claims description 49
- 230000003068 static effect Effects 0.000 claims description 47
- 238000004891 communication Methods 0.000 claims description 19
- 238000004364 calculation method Methods 0.000 claims description 12
- 238000001914 filtration Methods 0.000 claims description 10
- 238000011156 evaluation Methods 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 4
- 239000000284 extract Substances 0.000 description 34
- 238000013500 data storage Methods 0.000 description 33
- 238000010586 diagram Methods 0.000 description 28
- 238000000605 extraction Methods 0.000 description 19
- 238000005516 engineering process Methods 0.000 description 6
- 230000003287 optical effect Effects 0.000 description 6
- 230000000717 retained effect Effects 0.000 description 3
- 238000007635 classification algorithm Methods 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000013075 data extraction Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Definitions
- the present invention relates to the field of communications, and in particular, to a search method and system. Background technique
- An embodiment of the present invention provides a search method for providing a personalized search for a search client user, where the personalized search can provide a related search according to the interest model of the search client user.
- the search method includes the following steps: receiving a search request sent by a search client, where the search request includes one or more search keywords; extracting a user's interest model according to the user's personalized data; Transmitting a plurality of search keywords and the interest model to one or more member search devices; receiving a search result returned by the one or more member search devices and a score value corresponding to the search result calculated according to the interest model; Sorting the search results according to the score value to obtain a final search result, and returning to the search client.
- the embodiment of the present invention further provides a search method, where the method includes: receiving a search request sent by a search client, where the search request includes one or more search keywords and user identification information; The keyword and the user identification information are distributed to one or more member search devices; receiving search results searched by the one or more member search devices according to the interest model extracted by the user's personalized data; to the search client Send search results.
- An embodiment of the present invention further provides a search method, where the method includes: receiving a search request, where the search request includes one or more keywords; performing search according to the one or more keywords; and authenticating according to the user
- the embodiment of the present invention further provides a search method, the method includes: receiving a search request sent by a search client, where the search request includes one or more search keywords; and distributing the one or more search keywords to the search Receiving a search result returned by the search device; extracting a user's interest model according to the user's personalized data, and calculating a score value of the search result according to the interest model; sorting the search result according to the score value; The client sends the sorted search results.
- the embodiment of the invention further provides a computer readable storage medium, comprising a computer program, the computer program running, the one or more processors may be configured to perform the following steps: receiving a search request sent by a search client, where the search request includes One or more search keywords; extracting a user's interest model according to the user's personalized data; searching according to the one or more search keywords, And sorting the search results according to the interest model; and sending the relevance sorted search results to the search client.
- An embodiment of the present invention further provides a search apparatus, the apparatus comprising: a tool for receiving a search request; a tool for extracting a user's interest model according to the user's personalized data; and the request and the interest model a tool for distributing to a member search device; a tool for receiving a search result returned by the member search device and a search result score value; a tool for sorting the search result according to the score value; for returning the search result to the search client The tool of the end.
- An embodiment of the present invention further provides a search apparatus, the apparatus comprising: a tool for receiving a search client search request; a tool for extracting a user's interest model according to the user's personalized data; The search request search tool obtains a search result for relevance ranking; a tool for sending the search result to the search client.
- An embodiment of the present invention further provides a search apparatus, where the apparatus includes: a tool for receiving a search request; a tool for searching for an search result according to the user's personalized data extracted by the user and the search request; A tool that returns the search results.
- the embodiment of the present invention further provides a search system, where the system includes: a search server, which can perform search communication with one or more member search devices; the search server, configured to receive a search request of the search client, to the One or more member search devices distributing the search request and receiving search results returned by the one or more member search devices; the one or more member search devices for obtaining the search according to the search request
- the search server or the one or more member search devices may be further configured to calculate an interest model according to user data, and calculate a score value of the search result according to the interest model and perform relevance ranking.
- the embodiment of the present invention further provides a search system, the system includes: a search server, which can perform search communication with one or more member search devices; the search server is configured to receive a search request of the search client, according to the personalizedization of the user Extracting a user's interest model, distributing the search request and the interest model to the one or more member search devices, and receiving search results returned by the one or more member search devices and rating values of the search results, According to the score value pair Sorting the search results, and returning the sorted search results to the search client; the one or more member search means for searching for the search results according to the search request; and calculating the search according to the interest model The rating value of the search results.
- a search server which can perform search communication with one or more member search devices
- the search server is configured to receive a search request of the search client, according to the personalizedization of the user Extracting a user's interest model, distributing the search request and the interest model to the one or more member search devices, and receiving search results returned by the one or more member search devices
- An embodiment of the present invention further provides a search method, where the method includes: receiving a search request, where the search request includes one or more search keywords; distributing the one or more search keywords to a search device; After the search result, the search result is sorted according to the score value of the search result calculated by the user interest model extracted by the user's personalized data; and the sorted search result is returned.
- An embodiment of the present invention further provides a search method, where the method includes: receiving a search request; carrying a user interest model, distributing the search request to a search device; and receiving a personality according to the interest model returned by the search device Search results and return the personalized search results.
- the embodiment of the present invention further provides a search method, the method includes: receiving a search request of a search client; extracting a user's interest model according to user personalized data or extracting a pre-stored interest model of the user; carrying the user The interest model, sending the search request to the search server; receiving the search result personalized by the search server according to the user's interest model; and returning the personalized search result to the search client.
- the embodiment of the present invention further provides a search method, the method includes: receiving a search request and a user's interest model; searching for a search result according to the search request; and personalizing the search result according to the user's interest model ; Returns the personalized search results.
- An embodiment of the present invention further provides a search apparatus, where the apparatus includes: a tool for receiving a search request; a tool for distributing the search request to a search device; and a tool for receiving a search result returned by the search device; a tool for calculating a score value of a search result according to a user interest model extracted by the user's personalized data; a tool for sorting the search results according to the score value; and a tool for returning the sorted search result.
- An embodiment of the present invention further provides a search apparatus, where the apparatus includes: a tool for receiving a search request; a method for carrying a user interest model, and distributing the search request to a search device a tool for receiving personalized search results according to the interest model returned by the search device; and means for returning the personalized search results.
- the embodiment of the present invention further provides a search device, where the device includes: a tool for receiving a search request of a search client; for extracting a user's interest model according to user personalized data or extracting a pre-stored interest of the user a tool of the model; a tool for carrying the user's interest model, sending the search request to the search server; for receiving a search result personalized by the search server according to the user's interest model;
- the search client returns a tool for the personalized search result.
- the embodiment of the present invention further provides a search device, where the device includes: a search request receiving module, configured to receive a search request and a user's interest model; and a search processing module, configured to search for a search result according to the search request; a result personalization processing module, configured to: perform personalized processing on the search result according to the user's interest model; and a search result returning module, configured to return the personalized search result.
- a search request receiving module configured to receive a search request and a user's interest model
- a search processing module configured to search for a search result according to the search request
- a result personalization processing module configured to: perform personalized processing on the search result according to the user's interest model
- a search result returning module configured to return the personalized search result.
- the embodiment of the present invention further provides a search system, where the system includes: a search server, which can perform search communication with one or more member search devices; the search server is configured to receive a search request and carry a user interest model, and The search request is distributed to the one or more member search devices; the one or more member search devices are configured to search for a search result according to the search request, and calculate the search result according to the user's interest model And a return value to the search server; the search server receives the personalized search result according to the interest model returned by the one or more member search devices, and returns the personalized search result.
- a search server which can perform search communication with one or more member search devices
- the search server is configured to receive a search request and carry a user interest model, and The search request is distributed to the one or more member search devices; the one or more member search devices are configured to search for a search result according to the search request, and calculate the search result according to the user's interest model And a return value to the search server; the search server receives the personalized search result
- the search result is obtained by searching according to the interest model extracted by the user's personalized data and the search request, so that the user gets the search result closer to the user's needs, and different users can obtain different search results, so that the search result is obtained. More personalized, which is conducive to the promotion of search business.
- DRAWINGS 1 is a block diagram of an embodiment of a search system;
- FIG. 2 is an internal structure diagram of an embodiment of a search system;
- FIG. 3 is a block diagram of an embodiment of a search system; and FIG.
- FIG. 4 is an internal structure diagram of an embodiment of a search system
- Figure 5 is a block diagram of an embodiment of a search system
- Figure 6 is an internal structure diagram of an embodiment of a search system
- Figure 7 is a flow chart of an embodiment of a search method
- Figure 8 is a flow chart of an embodiment of a search method
- 9 is a flow chart of an embodiment of a search method
- FIG. 10 is a block diagram of an embodiment of a search system
- FIG. 11 is an internal structure diagram of an embodiment of a search system
- FIG. 12 is a block diagram of an embodiment of a search system
- Figure 14 is a block diagram of an embodiment of a search system
- Figure 15 is a block diagram of an embodiment of a search system
- Figure 16 is a block diagram of an embodiment of a search system
- FIG. 18 is a framework diagram of an embodiment of a search system
- FIG. 19 is an internal structure diagram of an embodiment of a search system
- 20 is a framework diagram of an embodiment of a search system
- FIG. 21 is an internal structure diagram of an embodiment of a search system
- FIG. 22 is a flowchart of an embodiment of a search method
- FIG. 23 is a message flow diagram of an embodiment of a search method
- 24 is a flowchart of an embodiment of a search method
- FIG. 25 is a flowchart of an embodiment of a search method
- FIG. 26 is a flow chart of an embodiment of a search method
- Figure 27 is a flow diagram of an embodiment of a search method.
- the search system 100 includes: a search client 102, a search server 104, a user data storage device 106, and one or more member search devices 108. .
- the search client 102 is configured to send a search request to the search server 104 according to the keyword input by the user, and receive the search result returned by the search server 104.
- the search client 102 can be a personal computer (PC, Personal Computer), a notebook computer (NB, Notebook Computer), a personal digital assistant (PDA, Personal Digital Assistant), a mobile phone (HS, Handset), and a smart CD player.
- PC Personal Computer
- NB notebook computer
- PDA Personal Digital Assistant
- a mobile phone HS, Handset
- smart CD player smart CD player.
- IODD Intelligent Optical Disk Drive Terminal devices with communication functions.
- a mobile phone is taken as an example for description.
- the search server 104 can perform search communication with one or more member search devices 108, receive search requests from the search client 102, extract user's interest models based on user data, and distribute the search requests and interests to the member search device 108.
- the model receives the search results returned by the one or more member search devices 108 and the score values of the search results, and sorts the search results according to the score values, and sends them to the search client 102.
- the one or more member search means 108 is configured to search for the search result according to the search request, and calculate the score value of the search result according to the interest model, and return it to the search server 104.
- the user data storage device 106 is configured to store user data, such as a user static.
- the user data storage device 106 can be placed in an internal system on the operator side.
- the member search device 108 is responsible for receiving the search request distributed by the search server 104, completing the search, and scoring and sorting the search results according to the same personalized scoring algorithm and the transmitted user interest model, and personalizing the search results and the score values. Return to the search server 104. In other embodiments, the search server 104 or one or more member search devices 108 are also used to filter the search results according to the search results and the score values calculated by the interest model.
- the search server 104 can indicate that the member search device 108 personalizes the search results using some uniform personalized scoring algorithm and, at each member search device 108, a specified personalized scoring algorithm. After scoring the search results and returning the personalized search results and the score values, the search server 104 summarizes the search results and comprehensively sorts the search results according to the score values calculated by the same personalized scoring algorithm, and then the final The personalized search results are returned to the search client 102.
- the search server 104 may further include: a search request receiving module.
- the interest model extraction module 204 the personalized search request distribution module 206, the personalized search result ranking module 216, and the final search result return module 218.
- the search request receiving module 202 is configured to receive a search request of the search client 102, where the search request may include one or more search keywords input by the user.
- the interest model extraction module 204 is configured to extract the user's interest model according to the user's personalized data.
- the user data may include: a user's static profile user's search history, the user's location information, the user's presence information, and the like.
- the personalized search request distribution module 206 is configured to distribute the personalized search request carrying the interest model to one or more member search devices 108, and at the same time, the personalized search request distribution module 206 can simultaneously search for the one or more members.
- the personalized search result sorting module 216 is configured to summarize the search results of the member search device 108, and calculate a score value of the search result according to a unified personalized scoring algorithm, and comprehensively sort the search results of each member search device. , for example, ranking highly relevant search results in the top position, or ranking highly relevant search results after the search results in the auction ranking. This will allow users to quickly navigate to the highly relevant results they need to search.
- the final search result return module 218 is configured to send the final search result to the search client 102. If the search result is filtered, only the search result with high correlation is included, and the partial search result is provided to the user, which can reduce the traffic of the network and reduce the pressure of the search client 102.
- the member search device 108 may further include: a personalized search request receiving module 208, a search processing module 210, a search result personalization processing module 212, and a personalized search result return module personalized search request receiving module 208, configured to receive a search request,
- the search request may be from the search server 104
- the personalized search request receiving module 208 may also receive the user interest model and the scoring algorithm ID sent by the search server 104.
- the search processing module 210 is configured to perform a search according to the search keyword to obtain a search result.
- the search result personalization processing module 212 is configured to perform personalized processing on the search result according to the user's interest model, and may also adopt a unified unified personalized scoring algorithm.
- the personalized search result returning module 214 is configured to return the search result, and may also return the score value of the search result. In this embodiment, the search result and the score value are returned to the search server 104.
- the search system includes: a search client 302, a search server 304, a user data storage device 306, and one or more member search devices 308.
- the member search device 308 can access the user data storage device 306, and does not need to distribute the user's interest model through the search server 304, thereby saving network resources.
- the search client 302 is configured to send a search request to the search server 304 according to the keyword input by the user, and receive the search result returned by the search server 304.
- the search client 302 can be a personal computer (PC, Personal Computer), a notebook computer (NB, Notebook Computer), a personal digital assistant (PDA, Personal Digital Assistant), a mobile phone (HS, Handset), and a smart CD player.
- PC Personal Computer
- NB notebook computer
- PDA Personal Digital Assistant
- a mobile phone HS, Handset
- smart CD player smart CD player.
- IODD Intelligent Optical Disk Drive Terminal devices with communication functions.
- a mobile phone is taken as an example for description.
- the search server 304 can perform search communication with one or more member search devices for receiving a search request from the search client 302, and distributing the search request to the member search device 308.
- the search request includes the ID of the user
- receives the search result returned by the one or more member search devices 308 and the score value of the search result sorts the search results according to the score value, and sends the search result to the search client 302.
- the one or more member search devices 308 may also be configured to calculate a score value of the search result according to the interest model and the unified personalized score algorithm, and return to the search server 304.
- the user data storage device 306 is configured to store user data, such as a user's static profile, hobbies, search history, location information, and presence information.
- user data storage device 306 can be disposed in the internal system on the operator side and connected to the member search device 308.
- the search server 304 is responsible for receiving a search request from the search client 302 (the search request contains the user's ID), distributing the search request to one or more member search devices 308, and searching for the device based on the extracted interests at the one or more members After the model scores the search results and returns to the personalized search results and ratings, the search server 304 summarizes the search results returned by the one or more member search devices 308, and based on the search result score values returned by the member search device 308. The search results are subjected to a comprehensive ranking, and then the final personalized search results are returned to the search client 302.
- the member search device 308 is responsible for receiving the search request distributed by the search server 304 (the search request includes the ID of the user), accessing the user data storage device 306 according to the user ID, extracting the interest model from the user data, completing the search, and according to the extracted interest.
- the model scores and ranks the search results using the same personalized scoring algorithm and returns the personalized search results and their rating values to the search server 304.
- the search server 304 or one or more member search devices 308 are optionally used to filter the search results based on the search results and the score values calculated by the interest model.
- the search server 304 can indicate that the member search device 308 personalizes the search results using some uniform personalized scoring algorithm and is After the search device 308 scores the search results according to the specified personalized scoring algorithm and returns the personalized search results and the score values, the search server 304 summarizes the search results and searches according to the score values calculated by the same personalized scoring algorithm. The result is a comprehensive ranking and then the final personalized search results are returned to the search client 302.
- the search server 304 may further include: a search request receiving module.
- the personalized search request distribution module 404 the personalized search result sorting module 416, and the final search result returning module 418.
- the search request receiving module 402 is configured to receive a search request of the search client 302, where the search request may include one or more search keywords input by the user.
- a personalized search request distribution module 404 for distributing the search request to one or more member search devices 308, while the personalized search request distribution module 404 can simultaneously specify uniform personalization for the one or more member search devices 308
- the scoring algorithm personalizes the search results. In this embodiment, an algorithmic ID is used to indicate that a unified scoring algorithm is employed.
- the personalized search result sorting module 416 is configured to sort the search results according to the score values obtained by calculating the search results according to the interest model, for example, ranking the highly relevant search results in the front position, or correlating the correlation results. High search results are ranked after search results in the auction ranking.
- the final search result returning module 418 is configured to send the final search result to the search client 302, and the search result may be a filtered search result containing only relatively high correlation, and provide partial search results to the user, which may reduce the network. Traffic, reducing the pressure on the search client 302.
- the member search device 308 may further include: a personalized search request receiving module 406, a search processing module 408, an interest model extraction module 410, a search result personalization processing module 412, and a personalized search result returning module 414.
- the personalized search request receiving module 406 is configured to receive a search request.
- the search request may be from the search server 304, and the search request may include a search keyword and a user ID, and does not include the user's personality for searching. Data.
- the interest model extraction module 410 is configured to extract the user's interest model according to the user's personalized data.
- the user data may include: a user's static profile user's search history, the user's location information, the user's presence information, and the like.
- the search result personalization processing module 412 is configured to perform personalized processing on the search result according to the user's interest model, and may also adopt a specified unified personalized scoring algorithm.
- the personalized search result returning module 414 is configured to return the search result, and may also return the score value of the search result. In this embodiment, the search result and the score value are returned to the search server 304.
- the search result is obtained by searching according to the interest model extracted by the user's personalized data and the search request, so that the user gets the search result closer to the user's needs, and different users can obtain different search results, so that the search result is obtained. More personalized, which is conducive to the promotion of search business.
- the search system 500 includes: a search client 502, a search server 504, a user data storage device 506, and one or more member search devices 508. .
- the search server 504 scores and sorts the search results using the user's interest model, and does not distribute the interest model to the member search device 508, thereby saving network resources.
- the search client 502 is configured to send a search request to the search server 504 according to the keyword input by the user, and receive the search result returned by the search server 504.
- the search client 502 can be a personal computer (PC, Personal Computer), a notebook computer (NB, Notebook Computer), a personal digital assistant (PDA, Personal Digital Assistant), a mobile phone (HS, Handset), and a smart CD player.
- PC Personal Computer
- NB notebook computer
- PDA Personal Digital Assistant
- a mobile phone HS, Handset
- smart CD player smart CD player.
- IODD Intelligent Optical Disk Drive Terminal devices with communication functions.
- a mobile phone is taken as an example for description.
- the search server 504 can perform search communication with one or more member search devices 508, receive a search request of the search client 502, extract a user's interest model based on the user data, and distribute the search request to the member search device 508, and Receiving search results returned by one or more member search devices 508, calculating a score value of the search result according to the interest model, and according to the score value
- the search results are sorted in relevance and sent to the search client 502.
- One or more member search means 508 for returning the search result according to the search request to the search server 504.
- the user data storage device 506 is configured to store user data, such as a user's static profile, hobbies, search history, location information, and presence information.
- user data storage device 506 can be placed in an internal system on the operator side.
- the member search device 508 is responsible for receiving the search request distributed by the search server 504, completing the search, and returning the search result to the search server 504.
- the search server 504 may be further configured to filter the search result according to the preset threshold according to the search result and the score value calculated by the interest model.
- the search server 504 may further include: a search request receiving module.
- Search request distribution module 604 interest model extraction module 612, personalized search result ranking module 614, and final search result return module 616.
- the search request receiving module 602 is configured to receive a search request of the search client 502, where the search request may include one or more search keywords input by the user.
- a personalized search request distribution module 604 is operative to distribute the search request to one or more member search devices 508.
- the interest model extraction module 612 is configured to extract the user's interest model based on the user data.
- the personalized search result sorting module 614 is configured to sort the search results by using the score values obtained by calculating the search results according to the interest model, for example, ranking the highly relevant search results in the front position, or correlating the correlation results. High search results are ranked after search results in the auction ranking. This will allow users to quickly navigate to the highly relevant results they need to search.
- the final search result returning module 616 is configured to send the final search result to the search client 502, and the search result may be a filtered search result containing only relatively high correlation, and provide partial search results to the user, which may reduce the network. Traffic reduces the pressure on the search client 502.
- the member search device 508 may further include: a search request receiving module 606, a search processing module 608, and a search result returning module 610.
- the search request receiving module 606 is configured to receive a search request.
- the search request may be from the search server 504, and the search request may include a search keyword without including the user's personalized data for searching.
- the search processing module 608 is configured to perform a search according to the search keyword to obtain a search result.
- the search result returns to the module 610, which is used to return the search result to the search server 504.
- the search result is obtained by searching according to the interest model extracted by the user's personalized data and the search request, so that the user gets the search result closer to the user's needs, and different users can obtain different search results, so that the search result is obtained. More personalized, which is conducive to the promotion of search business.
- an embodiment of a search method may be used to provide a personalized search for a search client user based on the above search system, and the personalized search may provide a related search according to a search client user's interest model.
- the method includes the following steps:
- Step 702 Receive a search request sent by a search client, where the search request includes one or more search keywords.
- the search request may be a signal sent by the mobile terminal to the network side.
- Step 704 Extract a user's interest model according to the user's personalized data.
- the personalized data of the user includes one or more of the following: Static User Profile User Profile Search history, location information, or presence information.
- Step 706 The searching for one or more search keywords and the interest model are distributed to one or more member search devices.
- the user interest model is carried to distribute the search request to the member search device, and the unified algorithm is used to personalize the search result.
- Step 708 The one or more member search devices complete the search, and apply the interest model of the user sent by the search server and the specified same algorithm to calculate the score value of the search result, and sort the search results according to the score value.
- the search results can also be filtered according to preset thresholds.
- Step 710 Receive personalized search results returned by the member search device and corresponding score values.
- Step 712 Perform comprehensive personalized relevance ranking on the search results returned by each member search device according to the score value of the search result.
- the method may further include: filtering the search result according to the score value, including: according to the set threshold, retaining a search result whose score value is greater than or equal to the set threshold, where the domain value is greater than, for example, greater than Search results equal to 0.8 are retained.
- Step 714 Send the final personalized search result to the search client user.
- a search method may be used to provide a personalized search for a search client user based on the above search system, and the personalized search may provide related search results according to a search client user's interest model.
- the method includes the following steps:
- Step 802 The search client sends a search request to the search server, where the search request includes one or more search keywords.
- the mobile terminal sends the search signal to the search server.
- Step 804 The search server distributes the search request and the user ID to the member search device.
- Step 806 The member search device accesses the user data storage device according to the user ID, and extracts the user interest model from the personalized data of the user.
- Step 808 Each member search device completes the search, and uses the same personalized scoring algorithm to perform personalized correlation scoring and sorting according to the extracted user interest model.
- Step 810 The member search device returns the personalized search result and the corresponding score value to the search server.
- Step 812 Perform comprehensive personalized relevance ranking on the search results returned by each member search device according to the score value of the search result.
- the method may further include: filtering the search result according to the score value, including: according to the set threshold, retaining a search result whose score value is greater than or equal to the set threshold, where the domain value is greater than, for example, greater than Search results equal to 0.8 are retained.
- Step 814 The search server returns a final personalized search result to the search client user.
- a search method may be used to provide a personalized search for a search client user based on the above search system, and the personalized search may provide related search results according to a search client user's interest model. The method includes the following steps:
- Step 902 The search client sends a search request to the search server, where the search request includes one or more search keywords.
- the mobile terminal sends the search signal to the search server.
- Step 904 The search server distributes the search request to the member search device.
- Step 906 The member search device searches for the search result according to the one or more keywords.
- Step 908 The member search device returns a search result to the search server.
- Step 910 Extract a user's interest model according to user data, calculate a score value of the search result according to the interest model, and perform comprehensive personalized relevance ranking on the search results returned by each member search device according to the score value of the search result.
- the method may further include: filtering the search result according to the score value, including: according to the set threshold, retaining a search result whose score value is greater than or equal to the set threshold, where the domain value is greater than, for example, greater than Search results equal to 0.8 are retained.
- Step 912 The search server returns a final personalized search result to the search client user.
- the following is a specific application example
- User interest is represented by n dimensions such as: news, sports, entertainment, finance, technology, real estate, games, women, forums, weather, merchandise, home appliances, music, reading, blogs, mobile phones, military, education, travel, MMS, ring tones, catering, civil aviation, industry, agriculture, computers, geography, etc.
- a vector W "l, r2, r 3, ..., m) composed of the user's scores of interest for each dimension is the user's interest model.
- the search server extracts the interest model from the user data
- Static interest model W1 corresponding to the user's static profile
- pi is the sum of the word frequencies of all words of the type belonging to the i-th interest dimension in the static profile.
- W2 d 1 +d2+d3 + ⁇ ⁇ ⁇ +di+ ⁇ ⁇ ⁇ dm , where di is the interest model vector corresponding to the user's click document.
- the value of tj is automatically reduced by a certain percentage, indicating that its importance decreases over time, until the value of tj is reduced to zero after a long time, then di is removed from the history. . (For example, every month, the value of tj is reduced by 10%)
- the meta search engine carries the extracted interest model data to initiate a personalized search request to one or more member search devices, and indicates that the plurality of member search devices personalize the search results using some specified personalization algorithm.
- a member search device performs a personalized search with a specified personalization algorithm
- the member search means retrieves the candidate result document based on the inverted index.
- the member search device performs personalized correlation scoring and sorting on the candidate result documents according to the interest model data and the specified personalized algorithm.
- the member search device returns the n most relevant documents (the highest score) and the personalized relevance score of the document.
- the meta search engine sorts the personalized search results returned by each member search device according to the size of the relevance score calculated by the same algorithm, and returns the most relevant result to the search client.
- the interest model vector in the above embodiment may also be applied to other embodiments of the present invention and will not be described again.
- the search result is obtained by searching according to the interest model extracted by the user's personalized data and the search request, so that the user gets the search result closer to the user's needs, and different users can obtain different search results, so that the search result is obtained. More personalized, which is conducive to the promotion of search business.
- the process of personalizing the scoring in one embodiment is accomplished by the member search device, enabling the member search device to return the most relevant personalized search results, making the personalized search results integrated by the final meta search engine more accurate.
- each member search device is specified to be personalized by the same algorithm, so that the scores returned by each member search device are comparable, and the search server only needs to perform a comprehensive evaluation on the scores returned by each member search device. Sorting, you can achieve the purpose of comprehensive personalized sorting of search results, without having to retrieve snapshots of all documents for real-time word-cutting and scoring, greatly reducing network traffic, greatly reducing the burden of meta-search engines, and improving personalized search. s efficiency.
- the search system 1000 includes: a search client 1002, a search server 1004, a user data storage device 1006, and one or more member search devices 1008.
- the search client 1002 is configured to send a search request to the search server 1004 based on the keyword input by the user in a textual or voice manner, and receive the search result returned by the search server 1004.
- the search client may be a personal computer (PC, Personal Computer), a notebook computer (NB, Notebook Computer), a personal digital assistant (PDA, Personal Digital Assistant), a mobile phone (HS, Handset), and a smart disk drive ( Terminal devices with communication functions such as IODD, Intelligent Optical Disk Drive).
- a mobile phone is taken as an example for description, and details are not described herein again.
- the search server 1004 can perform search communication with one or more member search devices, wherein the member search device further includes a member search server 1010.
- the search server 1004 is configured to receive a search request, carry a user interest model, distribute the search request to one or more member search devices 1008 and search server 1010, and receive the basis returned by one or more member search devices 1008 or member search servers 1010. Personalized search results for the interest model, returning the search results.
- the search server 1004 extracts the user's interest model from the user data (user static profile, search history, etc.) or directly extracts the interest model extracted according to the user data in advance, and carries The interest model distributes the personalized search request to one or more member search devices 1008 and member search servers 1010, and uses a unified algorithm ID to indicate that the personalized scoring algorithm scores the search results, and then returns personalized search results and ratings. Thereafter, the search server 1004 summarizes the search results, and comprehensively sorts the search results according to the score values of the search results calculated by the same personalized scoring algorithm, and then returns the final personalized search results to the search client 1002.
- the user data storage device 1006 is configured to store user data, including: a user's interest model, such as a user's static profile, hobbies, search history, location information, and presence information.
- a user's interest model such as a user's static profile, hobbies, search history, location information, and presence information.
- the user data storage device 1006 can be installed in the internal system on the carrier side.
- the member search device 1008 can be a separate vertical engine, and the member search server 1010 can be connected to the next-level member engine 1012.
- the member search device 1008 searches for a search result according to the search request, and calculates a score value of the search result according to the user's interest model, returns the search result and the score value to the search server 1004, and the member search device 1008 can also sort and send the search result to the search server. 1004.
- the member search server 1010 can distribute the search request to the next-level member engine 1012, and the member search server 1010 or the next-level member engine 1012 can also personalize the search result, and details are not described herein.
- the search server 1004 may further include: a search request receiving module 1102, an interest model extraction module 1104, a search request distribution module 1106, a personalized search result ranking module 1116, and a final search result returning module 1118.
- the search request receiving module 1102 is configured to receive a search request of the search client 1002, where the search request may include one or more search keywords input by the user, and the keyword may be input by the user in a text manner or a voice manner.
- the interest model extraction module 1104 is configured to extract the user's interest model according to the user's personalized data.
- the user data may include: a user's static profile user's search history, the user's location information, the user's presence information, and the like.
- a search request distribution module 1106, configured to distribute the personalized search request carrying the interest model to one or more The member search device 1008 and the member search server 1010.
- the personalized search request distribution module 1106 can simultaneously specify a unified personalized scoring algorithm for the one or more member search devices 1008 to personalize the search results, and the unified personalized scoring algorithm can be identified by an algorithm identification ID. Said.
- the personalized search result sorting module 1116 is configured to summarize the search results of the member search device 1008 and the member search server 1010, and perform comprehensive sorting according to a personalized scoring algorithm. For example, ranking highly relevant search results in the top position, or ranking highly relevant search results after the search results in the auction ranking. This is advantageous for the user to quickly browse to the highly relevant results that they need to search. In other embodiments, the ranking also includes priority information of the member search device and related factor information (such as price factor information, etc.).
- the final search result returning module 1118 is configured to send a final search result to the search client 1002, where the search result may be a filtered search result containing only relatively high correlation, and provide partial search results to the user, which may reduce the network. Traffic reduces the pressure on the search client 1002.
- the member search device 1008 may further include: a search request receiving module 1108, a search processing module 1110, a search result personalization processing module 1112, and a search result returning module 1114.
- the search request receiving module 1108 is configured to receive a search request.
- the search request may be from the search server 1004, and the personalized search request receiving module 1108 may further receive the user interest model and the scoring algorithm ID sent by the search server 1004.
- the search processing module 1110 is configured to perform search according to the search keyword to obtain search results.
- the search result personalization processing module 1112 is configured to personalize the search result according to the user's interest model, and may also adopt a unified unified personalized scoring algorithm.
- the personalized search result returning module 1114 is configured to return the search result, and may also return the score value of the search result. In this embodiment, the search result and the score value are returned to the search server 1004.
- the search system 1200 includes: a search client 1202.
- Search server 1204 user data storage device 1206, one or more member search devices 1208, member search server 1210, and one or more lower level member engines 1212.
- the search client 1202 is configured to send a search request to the search server 1204 according to a keyword input by the user in a text or voice manner, and receive the search result returned by the search server 1204.
- the search server 1204 can perform search communication with one or more member search devices 1208, wherein the member search device further includes a member search server 1210.
- the search server 1204 is configured to receive a search request, distribute the search request to one or more member search devices 1208 or search server 1210, and receive search results returned by one or more member search devices 1208 or member search servers 1210, and search results After personalized processing, the personalized search result is returned.
- the search server 1204 extracts the user's interest model from the user data (user static profile, search history, etc.) or directly extracts the interest model extracted in advance according to the user data, and personalizes the search result according to the interest model.
- the user data storage device 1206 is configured to store user data, including: a user's interest model, such as a user's static profile, hobbies, search history, location information, and presence information.
- a user's interest model such as a user's static profile, hobbies, search history, location information, and presence information.
- the user data storage device 1206 can be disposed in an internal system on the carrier side.
- the member search device 1208 can be a separate vertical engine, and the member search server 1210 can be connected to the next level member engine 1212.
- the member search means 1208 searches for the search result based on the search request and returns it to the search server 1204, wherein the member search server 1210 can distribute the search request to the next-level member engine 1212, and the search is completed by the next-level member engine 1212.
- the search server 1204 may further include: a search request receiving module 1302, a search request distribution module 1304, an interest model extraction module 1312, a personalized search result sorting module 1314, and a final search result returning module 1316.
- the search request receiving module 1302 is configured to receive a search request of the search client 1202, where the search request may include one or more search keywords input by the user, and the keyword may be used by the user. Enter text or voice.
- the search request distribution module 1304 is configured to distribute the search request to the member search device 1208 and the member search server 1210.
- the interest model extraction module 1312 is configured to extract the user's interest model according to the user's personalized data.
- the user data may include: a user's static profile user's search history, user's location information, user's presence information, and the like.
- the personalized search ranking module 1314 is configured to summarize the search results of the member search device 1208 and the member search server 1210, and calculate a score value of the search result according to the interest model extracted by the interest model extraction module 1312, and search for the search result according to the score value.
- the results are sorted. For example, ranking highly relevant search results in the top position, or ranking highly relevant search results after the search results in the auction ranking. This is advantageous for the user to quickly browse to the highly relevant results that they need to search.
- the ranking also includes level rating information and related factor information (such as price factor information, etc.) of the member search device.
- the final search result returning module 1316 is configured to send a final search result to the search client 1202, where the search result may be a filtered search result containing only relatively high correlation, and provide partial search results to the user, which may reduce the network. Traffic reduces the pressure on the search client 1202.
- the member search device 1208 may further include: a search request receiving module 1306, a search processing module 1308, and a search result returning module 1310.
- the search request receiving module 1306 is configured to receive a search request.
- the search request comes from the search server 1204 and does not include the user's interest model data.
- the search processing module 1308 is configured to perform search according to the search keyword to obtain search results.
- the search result returns to module 1310, which is used to return search results.
- a search server distributes a search request and an interest model to a member search device, and the member search device calculates a score value of the search result according to the interest model, and returns To the search server, the search server is personalized re-ranking and sorting, and the personalized search results are returned to the search client.
- the search system 1400 includes: a search client 1402, a search server 1404, and user data storage.
- the storage device 1406, the member search device 1408, the member search server 1410, and the next-level member engine search client 1402 are configured to send a search request to the search server 1404 according to a keyword input by the user in a textual manner or a voice manner, and receive the search.
- the search server 1404 can perform search communication with one or more member search devices 1408, wherein the member search device further includes a member search server 1410.
- the search server 1404 is configured to receive a search request, carry a user interest model, distribute the search request to one or more member search devices 1408 and search server 1410, and receive the personality according to the interest model returned by the member search device 1408 and the member search server 1410.
- the search results are re-ranked, the search results are sorted according to the results of the re-rank, and the sorted search results are returned to the search client 1402.
- the scoring algorithm may or may not be specified.
- the user data storage device 1406 is configured to store user data, including: a user's interest model, such as a user's static profile, hobbies, search history, location information, and presence information.
- a user's interest model such as a user's static profile, hobbies, search history, location information, and presence information.
- the user data storage device 1006 can be installed in the internal system on the carrier side.
- user data device 1406 is coupled to search server 1404.
- the member search device 1408 can be a separate vertical engine, and the member search server 1410 can be reconnected to the next level member engine 1412, the function of which can be similar to or different from the search server 1404.
- the member search means 1408 searches for a search result based on the search request, and calculates a score value of the search result based on the user's interest model.
- the calculated score value may calculate a score value of the search result by using a specified scoring algorithm when the search server 1404 specifies the scoring algorithm. If no scoring algorithm is specified, the member search device 1408 may calculate the score value of the search result by using the private algorithm.
- the member search means 1408 returns the search result and the score value to the search server 1404.
- the member search server 1410 can distribute the search request to the next-level member engine 1412, member search The server 1410 or the next-level member engine 1412 can also perform personalized processing for the search results, and details are not described herein.
- the search server 1404 includes: a search request receiving module 1502, an interest model extraction module 1504, a search request distribution module 1506, a re-scoring module 1516, a personalized search request sorting module 1518, and a final search result returning module 1520.
- the search request receiving module 1502 is configured to receive a search request of the search client 1402, and the search request may include one or more search keywords input by the user, and the keyword may be input by the user in a text mode or a voice manner.
- the interest model extraction module 1504 is configured to extract the user's interest model according to the user's personalized data or to extract the pre-stored extracted user interest model.
- the user data may include: a user's static profile user search history, User's location information, user's presence information, and so on.
- the search request distribution module 1506 is configured to distribute the personalized search request carrying the interest model to one or more member search devices 1408 and member search servers 1410.
- the personalized search request distribution module 1506 can simultaneously specify a unified personalized scoring algorithm for the one or more member search devices 1408 and the member search server 1410 to personalize the search results, and the unified personalized scoring algorithm can pass An algorithm identifies the ID to represent.
- the re-scoring module 1516 is configured to re-ranking the search results returned by the member search device 1408 and the member search server 1410.
- the re-ranking includes: calculating a score value of the search result according to the extracted interest model, and sorting the search result according to the score value. For example, ranking highly relevant search results in the top position, or ranking highly relevant search results after the search results in the auction ranking. This facilitates the user's quick access to the highly relevant results that they need to search.
- the ranking also includes the level rating information of the member search device and related factor information (e.g., price factor information, etc.). For example, to calculate level rating information, the formula is as follows:
- P rl* returned score +r2* level factor
- P represents a level score
- rl represents a weight of the returned score value
- r2 represents a weight of the level factor
- the returned score value is a score value returned by the member search device
- the level factor represents a member search device level
- the personalized search result sorting module 1518 is configured to comprehensively sort the search results according to the re-scoring score values, for example, ranking the highly relevant search results in the top position, or ranking the highly relevant search results. After the search results of the auction ranking. This facilitates the user's ability to quickly navigate to the highly relevant results that they need to search. In other embodiments, the ranking also includes priority information and related factor information (e.g., price factor information, etc.) of the member search device.
- the final search result returning module 1520 is configured to send the final search result to the search client 1402, where the search result may be a filtered search result containing only relatively high correlation, and provide partial search results to the user, which may reduce the network. Traffic reduces the pressure on the search client 1402.
- the member search device 1408 further includes: a search request receiving module 1508, a search processing module
- the search result personalization processing module 1512 and the search result return module 1514 are the search result personalization processing module 1512 and the search result return module 1514.
- the search request receiving module 1508 is configured to receive a search request.
- the search request may be from the search server 1404, and the search request receiving module 1508 may further receive the user interest model and the scoring algorithm ID sent by the search server 1404.
- the search processing module 1510 is configured to perform a search according to the search keyword to obtain a search result.
- the search result personalization processing module 1512 is configured to perform personalized processing on the search result according to the user's interest model, and may also adopt a specified unified personalized scoring algorithm. If no unified algorithm is specified, the private algorithm is used to search. The results are personalized.
- the search result returning module 1514 is configured to return the search result, and may also return the score value of the search result. In this embodiment, the search result and the score value are returned to the search server 1404.
- FIG. 16 is a framework diagram of an embodiment of a search system.
- the search system 1600 extracts the user's interest model from the application server or extracts the pre-stored extracted interest model, and performs personalized processing by the search server 1608.
- the search system 1600 includes: a search client 1602, a user data storage device 1604, an application server 1606, a search server 1608, a member search device 1610, a member search server 1612, and a next-level member engine 1614.
- the search client 1602 is configured to send a search request to the application server 1606 according to the keyword input by the user in a text mode or a voice mode, and then the search result returned by the application server 1606.
- the user data storage device 1604 is configured to store user data, including: a user's interest model, such as a user's static profile, hobbies, search history, location information, and presence information.
- a user's interest model such as a user's static profile, hobbies, search history, location information, and presence information.
- the user data storage device 1604 can be disposed in an internal system on the operator side.
- the application server 1606 is connected to the user data storage device 1604, and is configured to extract a user's interest model or extract a pre-stored extracted user's interest model, send the received search request and interest model to the search server 1608, and receive the search server 1608.
- the returned personalized search results are returned to the search client 1602.
- the application server 1606 extracts the user's interest model from the user data (user static profile search history, etc.) or directly extracts the interest model extracted according to the user data in advance, and carries the interest model to distribute the personalized search request to the search server. 1608.
- the search server 1608 can communicate with one or more member search devices, wherein the member search device can also include a member search server.
- the search server 1608 is configured to receive the search request of the application server 1606 and the interest model of the user, distribute the search request and the interest model to the member search device 1610 and the member search server 1612, receive the returned personalized search result and the score value, and search results. After summarization, a comprehensive re-ranking re-ranking is performed and returned to the application server 1606.
- the member search device 1610 can be a separate vertical engine, and the member search server 1612 can be connected to the next level member engine 1614.
- the member search device 1610 searches for a search result according to the search request, and calculates a score value of the search result according to the user's interest model, returns the search result and the score value to the search server 1608, and the member search device 1610 can also sort and send the search result to the search server. 1608.
- the member search server 1612 may distribute the search request to the next-level member engine 1614, and the member search server 1612 or the next-level member engine 1614 may also perform personalized processing for the search result, and details are not described herein.
- the application server 1606 may further include: an interest model extraction module 1702, a search request sending module 1704, and a search result receiving module 1724.
- the interest model extraction module 1702 is configured to extract the user's interest model according to the user's personalized data or to extract the pre-stored extracted user interest model.
- the user data may include: a user's static profile user's search history, User's location information, user's presence information, and so on.
- the search request sending module 1704 is configured to send the search request and the interest model to the search server 1608.
- the search result receiving module 1724 is configured to receive the personalized search result returned by the search server 1608, and return the search result to the search client 1602.
- the search server 1608 also includes a search request receiving module 1706, a search request distribution module 1708, a re-scoring module 1718, a personalized search result ranking module 1720, and a search result return module 1722.
- the search request receiving module 1706 is configured to receive the search request sent by the application server 1606 and the interest model of the user.
- the search request distribution module 1708 is configured to distribute the search request and the user's interest model to the member search device, and specify or not specify a unified scoring algorithm.
- the re-scoring module 1718 is configured to receive the search result and the score value returned by each member search device 1610, and summarize the search results to perform re-ranking.
- the re-ranking includes: calculating a score value of the search result according to the extracted interest model, and sorting the search result according to the score value. For example, ranking highly relevant search results in the top position, or ranking highly relevant search results after the search results in the auction ranking. This will help users to quickly browse to the relevant searches they need to search.
- the ranking also includes level rating information and related factor information (such as price factor information, etc.) of the member search device. For example, to calculate level rating information, the formula is as follows:
- P represents a level score
- rl represents a weight of the returned score value
- r2 represents a weight of the level factor
- the returned score value is a score value returned by the member search means
- the level factor represents a member search device level
- the personalized search result sorting module 1720 is configured to sort the search results according to the result of re-mnking, for example, ranking the highly relevant search results in the top position, or ranking the highly relevant search results in the auction ranking. After the search results. This facilitates the user's quick access to the highly relevant results that they need to search. In other embodiments, the ranking also includes priority information and related factor information (e.g., price factor information, etc.) of the member search device.
- the search result returning module 1722 is configured to return a search result to the application server 1606, where the search result may be a filtered search result containing only relatively high correlation, and provide partial search results to the user, which can reduce network traffic and reduce search.
- the member search device 1610 may further include: a search request receiving module 1710, a search processing module 1712, a search result personalization processing module 1714, and a search result returning module 1716.
- the search request receiving module 1710 is configured to receive a search request.
- the search request may be from the search server 1608, and the search request receiving module 1710 may further receive the user interest model and the scoring algorithm ID sent by the search server 1608.
- the search processing module 1712 is configured to perform a search according to the search keyword to obtain a search result.
- the search result personalization processing module 1714 is configured to perform personalized processing on the search result according to the user's interest model, and may also adopt a specified unified personalized scoring algorithm. If no unified algorithm is specified, the private algorithm is used to search. The results are personalized.
- the search result returning module 1716 is configured to return the search result, and may also return the score value of the search result. In this embodiment, the search result and the score value are returned to the search server.
- the search system 1800 extracts the user's interest model from the application server or retrieves the pre-stored extracted interest model, which is personalized by the search server 1808, but does not require re-rating.
- the search system 1800 includes: a search client 1802, a user data storage device 1804, an application server 1806, a search server 1808, a member search device 1810, a member search server 1812, and a next-level member engine 1814.
- the search client 1802 is configured to send a search request to the application server 1806 according to the keyword input by the user in a text mode or a voice manner, and then the search result returned by the application server 1806.
- the user data storage device 1804 is configured to store user data, including: a user's interest model, such as a user's static profile, hobbies, search history, location information, and presence information.
- a user's interest model such as a user's static profile, hobbies, search history, location information, and presence information.
- the user data storage device 1804 can be disposed in an internal system on the operator side.
- the application server 1806 is connected to the user data storage device 1804 for extracting the user's interest model or extracting the pre-stored extracted user's interest model, transmitting the received search request and interest model to the search server 1808, and receiving the search server 1808.
- the returned personalized search results are returned to the search client 1802.
- the application server 1806 extracts the user's interest model from the user data (user static profile search history, etc.) or directly extracts the interest model extracted according to the user data in advance, and carries the interest model to distribute the personalized search request to the search server. 1808.
- the search server 1808 can communicate with one or more member search devices, wherein the member search device can also include a member search server.
- the search server 1808 is configured to receive a search request of the application server 1806 and a user's interest model, and distribute the search request and the interest model to The member search device 1810 and the member search server 1812, and use the unified algorithm ID to indicate that the personalized scoring algorithm scores the search result, receives the returned personalized search result and the score value, summarizes the search results, and returns according to each member search device.
- the rating values are scored comprehensively and returned to the application server 1806.
- the member search device 1810 can be a separate vertical engine, and the member search server 1812 can be connected to the next level member engine 1814.
- the member search device 1810 searches for a search result according to the search request, and calculates a score value of the search result according to the user's interest model, returns the search result and the score value to the search server 1808, and the member search device 1810 can also sort and send the search result to the search server. 1808.
- the member search server 1812 can distribute the search request to the next-level member engine 1814, and the member search server 1812 or the next-level member engine 1814 can also personalize the search result, and details are not described herein.
- the application server 1806 may further include: an interest model extraction module 1902, a search request sending module 1904, and a search result receiving module 1922.
- the interest model extraction module 1902 is configured to extract the user's interest model according to the user's personalized data or to extract the pre-stored extracted user interest model.
- the user data may include: a user's static profile user search history, User's location information, user's presence information, and so on.
- the search request sending module 1904 is configured to send the search request and the interest model to the search server 1808.
- the search result receiving module 1922 is configured to receive the personalized search result returned by the search server 1808, and return it to the search client 1802.
- the search server 1808 also includes: a search request receiving module 1906, a search request distribution module
- the search request receiving module 1906 is configured to receive the search request sent by the application server 1806 and the user's interest model.
- the search request distribution module 1908 is configured to distribute the search request and the user's interest model to the member search device, and specify a unified scoring algorithm, which can be represented by an algorithm ID.
- the personalized search result sorting module 1918 is configured to perform comprehensive sorting of the search results according to the score value of the returned search result, for example, the search result with high relevance Ranked in the top position, or ranked highly relevant search results after the search results in the auction ranking. This is advantageous for the user to quickly browse to the highly relevant results that they need to search.
- the ranking also includes priority information of the member search device and related factor information (such as price factor information, etc.).
- the search result returning module 1920 is configured to return a search result to the application server 1806, and the search result may be a filtered search result containing only relatively high correlation, and provide partial search results to the user, which can reduce network traffic and reduce search.
- the member search device 1810 may further include: a search request receiving module 1910, a search processing module 1912, a search result personalization processing module 1914, and a search result returning module 1916.
- the search request receiving module 1910 is configured to receive a search request.
- the search request may be from the search server 1808, and the search request receiving module 1910 may further receive the user interest model and the scoring algorithm ID sent by the search server 1808.
- the search processing module 1912 is configured to perform a search according to the search keyword to obtain a search result.
- the search result personalization processing module 1914 is configured to personalize the search result according to the user's interest model, and may also adopt a specified unified personalized scoring algorithm.
- the search result returning module 1916 is configured to return a search result, and may also return a score value of the search result. In this embodiment, the search result and the score value are returned to the search server 1808.
- the search system 2000 extracts the user's interest model from the application server or extracts the pre-stored extracted interest model, and performs personalized processing by the search server 2008, and does not need to pass the interest model to the member search device and the member search server.
- the search system 2000 includes: a search client 2002, a user data storage device 2004, an application server 2006, a search server 2008, a member search device 2010, a member search server 2012, and a next-level member engine 2014.
- the search client 2002 is configured to send a search request to the application server 2006 according to the keyword input by the user in a text mode or a voice manner, and then the search result returned by the application server 2006.
- the user data storage device 2004 is configured to store user data, including: a user's interest model, such as a user's static profile, hobbies, search history, location information, and presence information.
- the user data storage device 2004 may be disposed in an internal system on the operator side.
- the application server 2006 is connected to the user data storage device 2004 for extracting the user's interest model or extracting the pre-stored extracted user's interest model, transmitting the received search request and interest model to the search server 2008, and receiving the search server 2008.
- the returned personalized search results are returned to the search client 2002.
- the application server 2006 extracts the user's interest model from the user data (user static profile search history, etc.) or directly extracts the interest model extracted according to the user data in advance, and carries the interest model to distribute the personalized search request to the search server. 2008.
- the search server 2008 can communicate with one or more member search devices, wherein the member search device can also include a member search server.
- the search server 2008 is configured to receive the search request of the application server 2006 and the interest model of the user, distribute the search request to the member search device 2010 and the member search server 2012, receive the returned search result, and calculate the score value of the search result according to the user's interest model. And sending the application server 2006 after sorting the search results according to the score value.
- the member search device 2010 can be a separate vertical engine, and the member search server 2012 can be connected to the next-level member engine 2014.
- the member search device 2010 obtains the search result based on the search request, and returns to the search server 2008, and the member search server 2012 can distribute the search request to the next-level member engine 2014 for search processing.
- the application server 2006 may further include: an interest model extraction module 2102, a search request sending module 2104, and a search result receiving module 2120.
- the interest model extraction module 2102 is configured to extract a user's interest model according to the user's personalized data or to extract the pre-stored extracted user interest model.
- the user data may include: a user's static profile user search history, User's location information, user's presentation Information, etc.
- the search request sending module 2104 is configured to send the search request and the interest model to the search server 2008.
- the search result receiving module 2120 is configured to receive the personalized search result returned by the search server 2008, and return it to the search client 2002.
- the search server 2008 may further include: a search request receiving module 2106, a search request distribution module 2108, a personalized search result sorting module 2116, and a search result returning module 2118.
- the search request receiving module 2106 is configured to receive a search request sent by the application server 2006 and a user's interest model.
- the search request distribution module 2108 is configured to distribute the search request to the member search device.
- the personalized search result sorting module 2116 is configured to receive the returned search result, calculate a score value of the search result according to the user's interest model, and sort the search result according to the score value, for example, rank the search result with high relevance In the top position, or rank highly relevant search results after the search results in the auction ranking. This facilitates the user's quick access to the highly relevant results that they need to search.
- the ranking also includes priority information and related factor information (e.g., price factor information, etc.) of the member search device.
- the search result returning module 2118 is configured to return a search result to the application server 2006, where the search result may be a filtered search result containing only relatively high correlation, and provide partial search results to the user, which can reduce network traffic and reduce search.
- the member search device 2010 may further include: a search request receiving module 2110, a search processing module 2112, and a search result returning module 2114. Its function is the same as that of the aforementioned member search device 1208, and will not be described again.
- the search result is obtained by searching according to the interest model extracted by the user's personalized data and the search request, so that the user gets the search result closer to the user's demand, and different users can obtain different search results, so that The search results are more personalized and are conducive to the promotion of search business.
- the personalized search is performed by the member search device, so that the member search device can return the most relevant search result, and the search server makes the search result more accurate when performing the comprehensive search result.
- the search method includes the following steps:
- Step 2202 Receive a search request sent by a search client, where the search request includes one or more search keywords, and the search keyword may be input by a user in a text manner or a voice manner.
- the search request may be a signal sent by the mobile terminal to the network side.
- Step 2204 Extract a user's interest model according to the user's personalized data or take out the pre-stored extracted interest model.
- the personalized data of the user includes one or more of the following: Static User Profile User Profile Search history, location information, or presence information.
- Step 2206 distributing the search one or more search keywords and the interest model to one or more member search devices or member search servers.
- the user interest model is carried to distribute the search request to the member search device, and indicates or does not indicate that the search result is personalized by using a unified algorithm. If a unified algorithm is specified, it may be indicated by an algorithm ID.
- Step 2208 The member search device completes the search. If there is a specified algorithm, the score value of the search result is calculated by using the same personalized personalization scoring algorithm. If there is no specified algorithm, the member search device may calculate the score of the search result by using a proprietary algorithm. Value, and sort the search results based on the rating value. At the same time, the search results can be filtered according to preset thresholds.
- Step 2210 Receive personalized search results returned by the member search device and corresponding score values.
- Step 2212 Re-rating the search result according to the score value of the search result and related factors (member search device level, price factor, etc.).
- re-ranking includes: calculating a score value of the search result according to the extracted interest model. For example, calculate level rating information and calculate the formula as follows:
- P rl* returned score +r2* level factor
- P represents a level score
- rl represents a weight of the returned score value
- r2 represents a weight of the level factor
- the returned score value is a score value returned by the member search device
- the level factor represents a member search device level
- Step 2214 sorting the search results according to the result of re-ranking. For example, rank highly relevant search results in the top position, or rank highly relevant search results after the search results in the auction position. This facilitates the user's quick access to the highly relevant results that they need to search.
- the ranking also includes level rating information and related factor information (e.g., price factor information, etc.) of the member search device.
- Step 2216 returning the final search result to the search client.
- the returned search result can be filtered and only contains highly relevant search results, and provides partial search results to the user, which can reduce network traffic and reduce the pressure on the search client.
- the search server receives a search request from the application server, and returns the search result to the application server, wherein the application server provides the user's interest model.
- Step 2302 The search client sends a search request to the application server, where the search request includes one or more search keywords, and the search keyword can be input by the user in text mode or voice mode.
- the search request may be a signal sent by the mobile terminal to the network side.
- Step 2304 The application server extracts the user interest model from the user's personalized data (such as the user's static profile and the user's click history), or directly extracts the interest model extracted in advance according to the user's personalized data.
- Step 2306 The application server carries the user's interest model, and sends the search request to the search server.
- Step 2308 The searching one or more search keywords and the interest model are distributed to one or more member search devices.
- the user interest model is carried to distribute the search request to the member search device, and indicates or does not indicate that the search result is personalized by using a unified algorithm. If the unified algorithm is specified, it may be indicated by an algorithm ID.
- Step 2310 The member search device completes the search. If there is a specified algorithm, the score value of the search result is calculated by using the same personalized personalization scoring algorithm. If there is no specified algorithm, the member search device may calculate the score of the search result by using a proprietary algorithm. And sorting the search results according to the score value, wherein the private algorithm may be the same algorithm as the specified scoring algorithm, or may be a different algorithm. At the same time, the search results can be filtered according to preset thresholds.
- Step 2312 Receive personalized search results returned by the member search device and corresponding score values.
- Step 2314 re-rating the search result according to the score value of the search result and related factors (member search device level, price factor, etc.).
- re-ranking includes: calculating a score value of the search result according to the extracted interest model. For example, calculate level rating information and calculate the formula as follows:
- P represents a level score
- rl represents a weight of the returned score value
- r2 represents a weight of the level factor
- the returned score value is a score value returned by the member search means
- the level factor represents a member search device level
- Step 2316 sorting the search results according to the result of re-ranking. For example, ranking highly relevant search results in the top position, or ranking highly relevant search results after the search results in the auction ranking. This will help users to quickly navigate to the highly relevant knots they need to search.
- the ranking further includes level rating information and related factor information (such as price factor information, etc.) of the member search device.
- Step 2318 returning the sorted result to the application server.
- Step 2320 the application server returns the search result to the search client.
- the search results returned in step 2318 and step 2320 may be filtered and only contain relatively high correlation search results, and provide partial search results to the user, which can reduce network traffic and reduce the pressure on the search client.
- the search server indicates a unified personalized scoring algorithm to the member search device.
- the method includes the following steps:
- Step 2402 The search client sends a search request to the application server, where the search request includes one or more search keywords, and the search keyword can be input by the user in text or voice.
- the search request may be a signal sent by the mobile terminal to the network side.
- Step 2404 The application server extracts the user interest model from the user's personalized data (such as the user's static profile and the user's click history), or directly extracts the interest model extracted in advance according to the user's personalized data.
- Step 2406 The application server carries the user's interest model and sends the search request to the search server.
- Step 2408 distributing the search one or more search keywords and the interest model to one or more member search devices.
- the user interest model is carried to distribute the search request to the member search device, and the unified algorithm is used to personalize the search result, and the unified algorithm can be indicated by an algorithm ID.
- Step 2410 The member search device completes the search, calculates a score value of the search result by using the same personalized personalization scoring algorithm, and sorts the search result according to the score value. At the same time, the search results can be filtered according to preset thresholds, for example: The returned results are controlled at 100.
- Step 2412 The search server receives the personalized search result returned by the member search device and the corresponding score value.
- Step 2414 Sort the search results according to the returned score value. For example, place highly relevant search results in the top position, or rank highly relevant search results after the search results in the auction position. This facilitates the user's quick access to the highly relevant results that they need to search.
- the ranking also includes level rating information and related factor information (e.g., price factor information, etc.) of the member search device.
- step 2416 the sorted result is returned to the application server.
- step 2418 the application server returns the search result to the search client.
- the search results returned in step 2416 and step 2418 may be filtered and only contain relatively high correlation search results, and provide partial search results to the user, which can reduce network traffic and reduce the pressure on the search client.
- the search server only distributes the search request to the member search device or the member search server.
- the method includes the following steps:
- Step 2502 The search client sends a search request to the application server, where the search request includes one or more search keywords, and the search keyword can be input by the user in a text mode or a voice manner.
- the search request may be a signal sent by the mobile terminal to the network side.
- Step 2504 The application server extracts the user interest model from the user's personalized data (such as the user's static profile and the user's click history), or directly extracts the interest model extracted in advance according to the user's personalized data.
- Step 2506 The application server carries the user's interest model, and sends the search request to the search server.
- Step 2510 the member search device completes the search, or the member search server completes the search.
- Step 2512 The search server receives the search result returned by the member search device or the member search server.
- Step 2514 The search server summarizes the search results, calculates a score value of the search result according to the user's interest model, and sorts the search result according to the score value.
- Step 2516 returning the sorted result to the application server.
- Step 2518 the application server returns the search result to the search client.
- the search results returned in step 2516 and step 2518 may be filtered and only contain relatively high correlation search results, and provide partial search results to the user, which can reduce network traffic and reduce the pressure on the search client.
- the search server extracts the user's interest model or extracts the pre-stored extracted interest model, distributes the search request and interest model to the member search device, and indicates a unified personalized scoring algorithm.
- the method includes:
- Step 2602 Receive a search request sent by a search client, where the search request includes one or more search keywords, and the search keyword may be input by a user in a text manner or a voice manner.
- the search request may be a signal sent by the mobile terminal to the network side.
- Step 2604 Extract a user's interest model according to the user's personalized data or take out the pre-stored extracted interest model.
- the personalized data of the user includes one or more of the following: Static User Profile User Profile Search history, location information, or presence information.
- Step 2606 distributing the search one or more search keywords and the interest model to one or more member search devices or member search servers.
- carrying a user interest model The type distributes the search request to the member search device, and indicates that the search result is personalized by a unified algorithm, indicating that the unified algorithm can be indicated by an algorithm ID.
- Step 2608 the member search device completes the search, calculates the score value of the search result by using the same personalized personalized scoring algorithm, and sorts the search result according to the score value.
- the search results can be filtered according to the preset threshold. For example: The return result is controlled at 100.
- Step 2610 The search server receives the personalized search result returned by the member search device and the corresponding score value.
- Step 2612 Sort the search results according to the returned score value. For example, place highly relevant search results in the top position, or rank highly relevant search results after the search results in the auction position. This facilitates the user's quick access to the highly relevant results that they need to search.
- the ranking also includes level rating information and related factor information (e.g., price factor information, etc.) of the member search device.
- Step 2614 returning the final search result to the search client.
- the returned search result can be filtered and only contains highly relevant search results, and provides partial search results to the user, which can reduce network traffic and reduce the pressure on the search client.
- the search server only distributes the search request to the member search device or the member search server.
- the method includes the following steps:
- Step 2702 Receive a search request sent by a search client, where the search request includes one or more search keywords, and the search keyword may be input by a user in a text manner or a voice manner.
- the search request may be a signal sent by the mobile terminal to the network side.
- Step 2704 Extract a user's interest model according to the user's personalized data or take out the pre-stored extracted interest model.
- the personalized data of the user includes one or more of the following: a static user profile, a user profile search history, location information, or presence information.
- Step 2706 The one or more search keywords are distributed to one or more member search devices or member search servers, and the member search server may continue to be distributed to the next-level engine, and details are not described herein.
- Step 2708 the member search device completes the search, and/or the member search server completes the search.
- Step 2710 The search server receives the personalized search result returned by the member search device.
- Step 2712 The search server summarizes the search results, calculates a score value of the search result according to the user's interest model, and sorts the search results according to the score value. For example, rank highly relevant search results in the top position, or rank highly relevant search results after the search results in the auction position. This facilitates the user's quick access to the highly relevant results that they need to search.
- the ranking also includes level rating information and related factor information (e.g., price factor information, etc.) of the member search device.
- Step 2714 returning the final search result to the search client.
- the returned search result can be filtered and only contains highly relevant search results, and provides partial search results to the user, which can reduce network traffic and reduce the pressure on the search client.
- User interest is represented by n dimensions such as: news, sports, entertainment, finance, technology, real estate, games, women, forums, weather, merchandise, home appliances, music, reading, blogs, mobile phones, military, education, travel, MMS, ring tones, catering, civil aviation, industry, agriculture, computers, geography, etc.
- a vector W(rl, r 2, r3, ..., m) composed of the user's scores of interest for each dimension is the user's interest model.
- the search server extracts the interest model from the user data
- Wl (pl, p2, p3, ..., ⁇ ), where pi is the sum of the word frequencies of all words of the type belonging to the i-th interest dimension in the static profile.
- Di (tlAt3, ... ., tn)
- the value of tj is automatically reduced by a certain percentage, indicating that its importance decreases over time, until the value of tj is reduced to zero after a long time, then di is removed from the history. . (For example, every month, the value of tj is reduced by 10%)
- the search server carries the extracted interest model data to initiate a personalized search request to one or more member search devices, and indicates that the plurality of member search devices personalize the search results using a specified personalized algorithm.
- a member search device performs a personalized search with a specified personalization algorithm.
- the member search means retrieves the candidate result document based on the inverted index.
- the member search device performs personalized correlation scoring and sorting on the candidate result documents according to the interest model data and the specified personalization algorithm.
- Algorithm (a): W (rl,r2,r3,...,m) is the interest model vector passed by the meta search engine.
- D (tl, t2, t3, ..., t n ) is the interest model vector corresponding to the document.
- W (rl,r2,r3,...,m) for the interest model passed by the search server
- the documents are classified by a common document classification algorithm such as Knn, Cvm, etc.
- the classified result document belongs to the type of C, and the type C matches the type of each dimension of the interest model, and the interest model matches the document type.
- the score value ri corresponding to the dimension i is paid to the document.
- the member search device returns n documents that are most relevant (the highest score value) and the personalized relevance score value of the document.
- the search server sorts the personalized search results returned by each member search device according to the size of the relevance score calculated by the same algorithm, and returns the most relevant result to the search client.
- the search result is obtained by searching according to the interest model extracted by the user's personalized data and the search request, so that the user gets the search result closer to the user's demand, and different users can obtain different search results, so that The search results are more personalized and are conducive to the promotion of search business.
- the personalized search is performed by the member search device, so that the member search device can return the most relevant search result, and the search server makes the search result more accurate when performing the comprehensive search result.
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Description
一种搜索方法和***
本申请要求了 2008年 8月 26日提交的, 申请号为 200810142177.X, 发明名称为 "一种搜索方法和***" ; 2008年 12月 15 日提交的, 申请号 为 200810187004.X, 发明名称为 "一种搜索方法和***" ; 2008年 12月 15 日提交的, 申请号为 200810187020.9, 发明名称为 "一种搜索方法和系 统" ; 2008年 12月 15日提交的, 申请号为 200810187019.6, 发明名称为 "一种搜索方法和***" ; 2008 年 12 月 15 日提交的, 申请号为 200810187049.7, 发明名称为 "一种搜索方法和***" 的中国申请优先权, 以上全部内容通过引用结合在本申请中。 技术领域
本发明涉及通信领域, 尤其涉及一种搜索方法和***。 背景技术
随着科学的发展和进歩, 通信技术也随着飞速发展, 其中移动搜索的也 是一个在伴随着通信技术发展的新亮点。 移动搜索技术的研究也成为业界研 发的重点。 传统的移动搜索技术很大程度上依赖于用户输入的关键词, 根据 用户提供的关键词给用户提供相关的搜索结果。 发明人在实现本发明的过程 中发现: 在互联网领域的今天, 根据用户关键词搜索的结果可能有会成百上 千个, 提供给用户的搜索结果很大程度上不能很好满足用户的搜索需求。 发明内容
有鉴于此, 有必要提供一种搜索方法, 来解决现有技术中搜索结果不能 很好的满足用户需求的问题。
同时, 提供了计算机可读存储介质、 搜索装置和搜索***。
本发明实施例提供一种搜索方法, 用于为搜索客户端用户提供个性化搜 索, 该个性化搜索可以根据所述搜索客户端用户的兴趣模型提供相关的搜索
结果, 所述搜索方法包括如下歩骤: 接收搜索客户端发送的搜索请求, 该搜 索请求中包含一个或多个搜索关键字; 根据用户的个性化数据提取用户的兴 趣模型; 将所述一个或多个搜索关键字和所述兴趣模型分发给一个或多个成 员搜索装置; 接收所述一个或多个成员搜索装置返回的搜索结果及根据所述 兴趣模型计算得到的搜索结果相应的评分值; 根据所述评分值对所述搜索结 果进行排序得到最终搜索结果, 返回给所述搜索客户端。
本发明实施例还提供一种搜索方法, 所述的方法包括: 接收搜索客户端 发送的搜索请求, 该搜索请求中包含一个或多个搜索关键字及用户标识信 息; 将所述一个或多个关键字及所述用户标识信息分发给一个或多个成员搜 索装置; 接收所述一个或多个成员搜索装置根据用户的个性化数据提取的兴 趣模型搜索到的搜索结果; 向所述搜索客户端发送搜索结果。
本发明实施例还提供一种搜索方法, 所述的方法包括: 接收搜索请求, 该搜索请求中包含一个或多个关键字; 根据所述一个或多个关键字进行搜 索; 根据用户的个性化数据提取的兴趣模型对搜索结果进行评分和排序; 返 回所述排序后的搜索结果。
本发明实施例还提供一种搜索方法, 该方法包括: 接收搜索客户端发送 的搜索请求, 该搜索请求中包含一个或多个搜索关键字; 将所述一个或多个 搜索关键字分发给搜索装置; 接收搜索装置返回的搜索结果; 根据用户的个 性化数据提取用户的兴趣模型, 并根据所述兴趣模型计算搜索结果的评分 值; 根据所述评分值对搜索结果进行排序; 向所述搜索客户端发送排序后的 搜索结果。
本发明实施例还提供一种计算机可读存储介质, 包括计算机程序, 该计 算机程序运行时可以使得一个或多个处理器执行如下歩骤: 接收搜索客户端 发送的搜索请求, 该搜索请求中包含一个或多个搜索关键字; 根据用户的个 性化数据提取用户的兴趣模型; 根据所述一个或多个搜索关键字进行搜索,
并根据所述兴趣模型对搜索结果进行相关性排序; 向所述搜索客户端发送相 关性排序后的搜索结果。
本发明实施例还提供一种搜索装置, 该装置包括: 用于接收搜索请求的 工具; 用于根据用户的个性化数据提取用户的兴趣模型的工具; 用于将所述 请求和所述兴趣模型分发给成员搜索装置的工具; 用于接收成员搜索装置返 回的搜索结果及搜索结果评分值的工具; 用于根据所述评分值对搜索结果进 行排序的工具; 用于将搜索结果返回给搜索客户端的工具。
本发明实施例还提供一种搜索装置, 该装置包括: 用于接收搜索客户端 搜索请求的工具; 用于根据用户的个性化数据提取用户的兴趣模型的工具; 用于根据所述兴趣模型对所述搜索请求搜索得到的搜索结果进行相关性排序 的工具; 用于向所述搜索客户端发送所述搜索结果的工具。
本发明实施例还提供一种搜索装置, 该装置包括: 用于接收搜索请求的 工具; 用于根据用户的个性化数据提取的兴趣模型及所述搜索请求进行搜索 得到搜索结果的工具; 用于返回所述搜索结果的工具。
本发明实施例还提供一种搜索***, 所述的***包括: 搜索服务器, 可 以与一个或多个成员搜索装置进行搜索通信; 所述搜索服务器, 用于接收搜 索客户端的搜索请求, 向所述一个或多个成员搜索装置分发所述搜索请求, 并接收所述一个或多个成员搜索装置返回的搜索结果; 所述一个或多个成员 搜索装置, 用于根据所述搜索请求搜索得到所述搜索结果; 所述搜索服务器 或所述一个或多个成员搜索装置, 还可以被配置为根据用户数据提取的兴趣 模型, 并根据所述兴趣模型计算所述搜索结果的评分值并进行相关性排序。
本发明实施例还提供一种搜索***, 该***包括: 搜索服务器, 可以与 一个或多个成员搜索装置进行搜索通信; 所述搜索服务器, 用于接收搜索客 户端的搜索请求, 根据用户的个性化数据提取用户的兴趣模型, 向所述一个 或多个成员搜索装置分发所述搜索请求和所述兴趣模型, 并接收所述一个或 多个成员搜索装置返回的搜索结果及搜索结果的评分值, 根据所述评分值对
搜索结果进行排序, 并向搜索客户端返回排序后的搜索结果; 所述一个或多 个成员搜索装置, 用于根据所述搜索请求搜索得到所述搜索结果; 并根据所 述兴趣模型计算所述搜索结果的所述评分值。
本发明实施例还提供一种搜索方法, 所述的方法包括: 接收搜索请求, 该搜索请求中包含一个或多个搜索关键字; 将所述一个或多个搜索关键字分 发给搜索装置; 返回搜索结果后, 根据用户的个性化数据提取的用户兴趣模 型计算的搜索结果的评分值对搜索结果进行排序; 返回排序后的搜索结果。
本发明实施例还提供一种搜索方法, 所述的方法包括: 接收搜索请求; 携带用户兴趣模型, 将所述搜索请求分发给搜索装置; 接收所述搜索装置返 回的根据所述兴趣模型的个性化搜索结果, 并返回所述个性化搜索结果。
本发明实施例还提供一种搜索方法, 所述的方法包括: 接收搜索客户端 的搜索请求; 根据用户个性化数据提取用户的兴趣模型或取出预先存储的所 述用户的兴趣模型; 携带所述用户的兴趣模型, 向搜索服务器发送所述搜索 请求; 接收所述搜索服务器返回的根据所述用户的兴趣模型个性化的搜索结 果; 向所述搜索客户端返回所述个性化处理后的搜索结果。
本发明实施例还提供一种搜索方法, 该方法包括: 接收搜索请求和用户 的兴趣模型; 根据所述搜索请求搜索得到搜索结果; 根据所述用户的兴趣模 型对所述搜索结果进行个性化处理; 返回所述个性化处理后的搜索结果。
本发明实施例还提供一种搜索装置, 所述的装置包括: 用于接收搜索请 求工具; 用于将所述搜索请求分发给搜索装置的工具; 用于接收搜索装置返 回的搜索结果的工具; 用于根据用户的个性化数据提取的用户兴趣模型计算 搜索结果的评分值的工具; 用于根据所述评分值对搜索结果进行排序的工 具; 和用于返回排序后的搜索结果的工具。
本发明实施例还提供一种搜索装置, 所述的装置包括; 用于接收搜索请 求的工具; 用于携带用户兴趣模型, 将所述搜索请求分发给搜索装置的工
具; 用于接收所述搜索装置返回的根据所述兴趣模型的个性化搜索结果的工 具; 和用于返回所述个性化搜索结果的工具。
本发明实施例还提供一种搜索装置, 所述的装置包括: 用于接收搜索客 户端的搜索请求的工具; 用于根据用户个性化数据提取用户的兴趣模型或取 出预先存储的所述用户的兴趣模型的工具; 用于携带所述用户的兴趣模型, 向搜索服务器发送所述搜索请求的工具; 用于接收所述搜索服务器返回的根 据所述用户的兴趣模型个性化的搜索结果; 用于向所述搜索客户端返回所述 个性化处理后的搜索结果的工具。
本发明实施例还提供一种搜索装置, 所述的装置包括: 搜索请求接收模 块, 用于接收搜索请求和用户的兴趣模型; 搜索处理模块, 用于根据所述搜 索请求搜索得到搜索结果; 搜索结果个性化处理模块, 用于根据所述用户的 兴趣模型对所述搜索结果进行个性化处理; 和搜索结果返回模块, 用于返回 所述个性化处理后的搜索结果。
本发明实施例还提供一种搜索***, 所述的***包括: 搜索服务器, 可 以与一个或多个成员搜索装置进行搜索通信; 所述搜索服务器, 用于接收搜 索请求, 携带用户兴趣模型, 将所述搜索请求分发给所述一个或多个成员搜 索装置; 所述一个或多个成员搜索装置, 用于根据所述搜索请求搜索得到搜 索结果, 并根据所述用户的兴趣模型计算该搜索结果的评分值, 并返回给所 述搜索服务器; 所述搜索服务器接收所述一个或多个成员搜索装置返回的根 据所述兴趣模型的个性化搜索结果, 并返回所述个性化搜索结果。
在一个实施例中通过根据用户的个性化数据提取的兴趣模型及所述搜索 请求进行搜索得到搜索结果, 使得用户得到搜索结果更加贴近用户需求, 不 同的用户可以得到不同的搜索结果, 使得搜索结果更加个性化, 有利于搜索 业务的推广使用。 附图说明
图 1是一个搜索***实施例的框架图; 图 2是一个搜索***实施例的内部结构图; 图 3是一个搜索***实施例的框架图; 图 4是一个搜索***实施例的内部结构图; 图 5是一个搜索***实施例的框架图; 图 6是一个搜索***实施例的内部结构图; 图 7是一个搜索方法实施例的流程图; 图 8是一个搜索方法实施例的流程图; 图 9是一个搜索方法实施例的流程图; 图 10是一个搜索***实施例的框架图; 图 11是一个搜索***实施例的内部结构图; 图 12是一个搜索***实施例的框架图; 图 13是一个搜索***实施例的内部结构图; 图 14是一个搜索***实施例的框架图; 图 15是一个搜索***实施例的内部结构图; 图 16是一个搜索***实施例的框架图; 图 17是一个搜索***实施例的内部结构图; 图 18是一个搜索***实施例的框架图; 图 19是一个搜索***实施例的内部结构图; 图 20是一个搜索***实施例的框架图; 图 21是一个搜索***实施例的内部结构图; 图 22是一个搜索方法实施例的流程图; 图 23是一个搜索方法实施例的消息流程图; 图 24是一个搜索方法实施例的流程图; 图 25是一个搜索方法实施例的流程图; 图 26是一个搜索方法实施例的流程图;
图 27是一个搜索方法实施例的流程图。 具体实施方式
请参看图 1, 是一个搜索***实施例 100的框架图, 在本实施例中, 搜 索*** 100包括: 搜索客户端 102、 搜索服务器 104、 用户数据存储设备 106及一个或多个成员搜索装置 108。
搜索客户端 102, 用于根据用户输入的关键字向搜索服务器 104发送搜 索请求, 并接收搜索服务器 104返回的搜索结果。 本实施例中, 搜索客户端 102 可以是个人电脑 (PC , Personal Computer ) 、 笔记本电脑 (NB, Notebook Computer) 、 个人数字助理 (PDA, Personal Digital Assistant) 、 手机 (HS, Handset ) 和智能光碟机 ( IODD, Intelligent Optical Disk Drive) 等具有通信功能的终端设备。 本实施例中, 以手机为例进行说明。
搜索服务器 104, 可以与一个或多个成员搜索装置 108进行搜索通信, 用于接收搜索客户端 102的搜索请求, 根据用户数据提取用户的兴趣模型, 向成员搜索装置 108分发所述搜索请求和兴趣模型, 并接收一个或多个成员 搜索装置 108返回的搜索结果及搜索结果的评分值, 并根据评分值对搜索结 果进行相关性排序, 发送给搜索客户端 102。 一个或多个成员搜索装置 108, 用于根据搜索请求搜索得到搜索结果, 并根据兴趣模型计算搜索结果 的评分值, 返回给搜索服务器 104。
其中, 用户数据存储设备 106, 用于存储用户数据, 例如用户静态的
Profile (描述) 、 兴趣爱好、 搜索历史、 位置信息和呈现信息等。 在本实施 例中, 用户数据存储设备 106可以设置于运营商侧的内部***中。
成员搜索装置 108负责接收搜索服务器 104分发的搜索请求, 完成搜 索, 并根据相同的个性化评分算法及传送过来的用户兴趣模型对搜索结果进 行评分与排序, 并将个性化的搜索结果及评分值返回给搜索服务器 104。
在其他实施例中, 可选地, 搜索服务器 104或一个或多个成员搜索装置 108, 还用于根据搜索结果及兴趣模型计算出的评分值对搜索结果进行过 滤。
在其他实施例中, 可选地, 搜索服务器 104可以指明成员搜索装置 108 采用某种统一的个性化评分算法对搜索结果进行个性化处理, 并在各个成员 搜索装置 108按指定的个性化评分算法对搜索结果进行评分再返回个性化搜 索结果及评分值后, 搜索服务器 104对搜索结果进行汇总并根据用同一种个 性化评分算法算出来的评分值对搜索结果进行综合的排序, 然后将最终的个 性化搜索结果返回给搜索客户端 102。
请结合参看图 2, 其中搜索服务器 104还可以包括: 搜索请求接收模块
202、 兴趣模型提取模块 204、 个性化搜索请求分发模块 206、 个性化搜索结 果排序模块 216和最终搜索结果返回模块 218。
搜索请求接收模块 202, 用于接收搜索客户端 102的搜索请求, 该搜索 请求中可以包含用户输入的一个或多个搜索关键字。 兴趣模型提取模块 204, 用于根据用户的个性化数据提取用户的兴趣模型, 本实施例中, 用户 数据可以包括: 用户的静态 Profile 用户的搜索历史、 用户的位置信息、 用 户的呈现信息等。 个性化搜索请求分发模块 206, 用于将携带兴趣模型的个 性化搜索请求分发给一个或多个成员搜索装置 108, 同时, 个性化搜索请求 分发模块 206可以同时为这一个或多个成员搜索装置 108指定统一的个性化 评分算法对搜索结果进行个性化处理, 该统一的个性化评分算法可以通过一 个算法标识 ID来表示。 个性化搜索结果排序模块 216, 用于对成员搜索装 置 108的搜索结果进行汇总, 并根据统一的个性化评分算法来计算出搜索结 果的评分值, 对各个成员搜索装置的搜索结果进行综合性排序, 例如, 将相 关性高的搜索结果排在靠前的位置, 或将相关性高的搜索结果排在竞价排名 的搜索结果之后。 这样有利于用户能快速浏览到其需要搜索的相关性高的结 果。 最终搜索结果返回模块 218, 用于向搜索客户端 102发送最终的搜索结
果, 该搜索结果可以是经过过滤后的只包含相关性比较高的搜索结果, 向用 户提供部分搜索结果, 可以减少网络的流量, 减轻搜索客户端 102的压力。
成员搜索装置 108, 还可以包括: 个性化搜索请求接收模块 208、 搜索 处理模块 210、 搜索结果个性化处理模块 212和个性化搜索结果返回模块 个性化搜索请求接收模块 208, 用于接收搜索请求, 本实施例中, 搜索 请求可以来自于搜索服务器 104, 个性化搜索请求接收模块 208还可以接收 搜索服务器 104发送的用户兴趣模型和评分算法 ID。 搜索处理模块 210, 用于根据搜索关键字进行搜索得到搜索结果。 搜索结果个性化处理模块 212, 用于根据用户的兴趣模型对搜索结果进行个性化处理, 同时, 也可以 采用被指定的统一的个性化评分算法。 个性化搜索结果返回模块 214, 用于 返回搜索结果, 也可以返回搜索结果的评分值, 本实施例中, 返回搜索结果 及评分值给搜索服务器 104。
请参看图 3, 是一个搜索***实施例 300的框架图, 在本实施例中, 搜 索***包括: 搜索客户端 302、 搜索服务器 304、 用户数据存储设备 306、 及一个或多个成员搜索装置 308。 本实施例中, 成员搜索装置 308可以访问 用户数据存储设备 306, 不需要通过搜索服务器 304分发用户的兴趣模型, 可以节省网络资源。
搜索客户端 302, 用于根据用户输入的关键字向搜索服务器 304发送搜 索请求, 并接收搜索服务器 304返回的搜索结果。 本实施例中, 搜索客户端 302 可以是个人电脑 (PC , Personal Computer ) 、 笔记本电脑 (NB, Notebook Computer) 、 个人数字助理 (PDA, Personal Digital Assistant) 、 手机 (HS, Handset ) 和智能光碟机 ( IODD, Intelligent Optical Disk Drive) 等具有通信功能的终端设备。 本实施例中, 以手机为例进行说明。
搜索服务器 304, 可以与一个或多个成员搜索装置进行搜索通信, 用于 接收搜索客户端 302的搜索请求, 向成员搜索装置 308分发所述搜索请求
(搜索请求中包含用户的 ID) , 并接收一个或多个成员搜索装置 308返回 的搜索结果及搜索结果的评分值, 根据评分值对搜索结果进行相关性排序, 发送给搜索客户端 302。 一个或多个成员搜索装置 308, 也可以被配置成根 据用户数据提取的兴趣模型, 并根据兴趣模型及统一的个性化评分算法计算 搜索结果的评分值, 返回给搜索服务器 304。
其中, 用户数据存储设备 306, 用于存储用户数据, 例如用户静态的 Profile (描述) 、 兴趣爱好、 搜索历史、 位置信息和呈现信息等。 在本实施 例中, 用户数据存储设备 306可以设置于运营商侧的内部***中, 并与成员 搜索装置 308连接。
搜索服务器 304负责接收搜索客户端 302的搜索请求 (搜索请求中包含 用户的 ID) , 将搜索请求分发给一个或多个成员搜索装置 308, 并在这一 个或多个成员搜索装置根据提取的兴趣模型对搜索结果进行评分, 再返回个 性化搜索结果及评分后, 搜索服务器 304对这一个或多个成员搜索装置 308 返回的搜索结果进行汇总, 并根据成员搜索装置 308返回的搜索结果评分值 对搜索结果进行一个综合的排序, 然后将最后的个性化搜索结果返回给搜索 客户端 302。
成员搜索装置 308负责接收搜索服务器 304分发过来的搜索请求 (搜索 请求中包含用户的 ID) , 根据用户 ID访问用户数据存储设备 306, 从用户 数据中提取兴趣模型, 完成搜索, 并根据提取的兴趣模型用相同的个性化评 分算法对搜索结果进行评分与排序, 并将个性化的搜索结果及其评分值返回 给搜索服务器 304。
在其他实施例中, 可选地, 搜索服务器 304或一个或多个成员搜索装置 308, 还用于根据搜索结果及兴趣模型计算出的评分值对搜索结果进行过 滤。
在其他实施例中, 可选地, 搜索服务器 304可以指明成员搜索装置 308 采用某种统一的个性化评分算法对搜索结果进行个性化处理, 并在各个成员
搜索装置 308按指定的个性化评分算法对搜索结果进行评分再返回个性化搜 索结果及评分值后, 搜索服务器 304对搜索结果进行汇总并根据用同一种个 性化评分算法算出来的评分值对搜索结果进行综合的排序, 然后将最终的个 性化搜索结果返回给搜索客户端 302。
请结合参看图 4, 其中搜索服务器 304还可以包括: 搜索请求接收模块
402、 个性化搜索请求分发模块 404、 个性化搜索结果排序模块 416和最终 搜索结果返回模块 418。
搜索请求接收模块 402, 用于接收搜索客户端 302的搜索请求, 该搜索 请求中可以包含用户输入的一个或多个搜索关键字。 个性化搜索请求分发模 块 404, 用于将搜索请求分发给一个或多个成员搜索装置 308, 同时, 个性 化搜索请求分发模块 404可以同时为这一个或多个成员搜索装置 308指定统 一的个性化评分算法对搜索结果进行个性化处理, 本实施例中, 通过一个算 法 ID来指明采用统一的评分算法。 个性化搜索结果排序模块 416, 用于将 根据兴趣模型计算搜索结果得到的评分值, 对搜索结果进行相关性排序, 例 如, 将相关性高的搜索结果排在靠前的位置, 或将相关性高的搜索结果排在 竞价排名的搜索结果之后。 这样有利于用户能快速浏览到其需要搜索的相关 性高的结果。 最终搜索结果返回模块 418, 用于向搜索客户端 302发送最终 的搜索结果, 该搜索结果可以是经过过滤后的只包含相关性比较高的搜索结 果, 向用户提供部分搜索结果, 可以减少网络的流量, 减轻搜索客户端 302 的压力。
本实施例中, 成员搜索装置 308还可以包括: 个性化搜索请求接收模块 406、 搜索处理模块 408、 兴趣模型提取模块 410、 搜索结果个性化处理模块 412和个性化搜索结果返回模块 414。
个性化搜索请求接收模块 406, 用于接收搜索请求, 本实施例中, 搜索 请求可以来自于搜索服务器 304, 搜索请求中可以包含搜索关键字及用户 ID, 而不包含用户的用于搜索的个性化数据。 搜索处理模块 408, 用于根据
搜索关键字进行搜索得到搜索结果。 兴趣模型提取模块 410, 用于根据用户 的个性化数据提取用户的兴趣模型, 本实施例中, 用户数据可以包括: 用户 的静态 Profile 用户的搜索历史、 用户的位置信息、 用户的呈现信息等。 搜 索结果个性化处理模块 412, 用于根据用户的兴趣模型对搜索结果进行个性 化处理, 同时, 也可以采用被指定的统一的个性化评分算法。 个性化搜索结 果返回模块 414, 用于返回搜索结果, 也可以返回搜索结果的评分值, 本实 施例中, 返回搜索结果及评分值给搜索服务器 304。
在一个实施例中通过根据用户的个性化数据提取的兴趣模型及所述搜索 请求进行搜索得到搜索结果, 使得用户得到搜索结果更加贴近用户需求, 不 同的用户可以得到不同的搜索结果, 使得搜索结果更加个性化, 有利于搜索 业务的推广使用。
请参看图 5, 是一个搜索***实施例 500的框架图, 在本实施例中, 搜 索*** 500包括: 搜索客户端 502、 搜索服务器 504、 用户数据存储设备 506及一个或多个成员搜索装置 508。 本实施例中, 搜索服务器 504, 使用 用户的兴趣模型对搜索结果进行评分和排序, 不用分发兴趣模型给成员搜索 装置 508, 节省网络资源。
搜索客户端 502, 用于根据用户输入的关键字向搜索服务器 504发送搜 索请求, 并接收搜索服务器 504返回的搜索结果。 本实施例中, 搜索客户端 502 可以是个人电脑 (PC , Personal Computer ) 、 笔记本电脑 (NB, Notebook Computer) 、 个人数字助理 (PDA, Personal Digital Assistant) 、 手机 (HS, Handset ) 和智能光碟机 ( IODD, Intelligent Optical Disk Drive) 等具有通信功能的终端设备。 本实施例中, 以手机为例进行说明。
搜索服务器 504, 可以与一个或多个成员搜索装置 508进行搜索通信, 用于接收搜索客户端 502的搜索请求, 根据用户数据提取用户的兴趣模型, 向成员搜索装置 508分发所述搜索请求, 并接收一个或多个成员搜索装置 508返回的搜索结果, 根据兴趣模型计算搜索结果的评分值, 并根据评分值
对搜索结果进行相关性排序, 发送给搜索客户端 502。 一个或多个成员搜索 装置 508, 用于根据搜索请求搜索得到搜索结果返回给搜索服务器 504。
其中, 用户数据存储设备 506, 用于存储用户数据, 例如用户静态的 Profile (描述) 、 兴趣爱好、 搜索历史、 位置信息和呈现信息等。 在本实施 例中, 用户数据存储设备 506可以设置于运营商侧的内部***中。
成员搜索装置 508负责接收搜索服务器 504分发的搜索请求, 完成搜 索, 并将搜索结果返回给搜索服务器 504。
在其他实施例中, 可选地, 搜索服务器 504, 还可以被配置为根据搜索 结果及兴趣模型计算出的评分值, 根据预设的阈值对搜索结果进行过滤。
请结合参看图 6, 其中搜索服务器 504还可以包括: 搜索请求接收模块
602、 搜索请求分发模块 604、 兴趣模型提取模块 612、 个性化搜索结果排序 模块 614和最终搜索结果返回模块 616。
搜索请求接收模块 602, 用于接收搜索客户端 502的搜索请求, 该搜索 请求中可以包含用户输入的一个或多个搜索关键字。 个性化搜索请求分发模 块 604, 用于将搜索请求分发给一个或多个成员搜索装置 508。 兴趣模型提 取模块 612, 用于根据用户数据提取用户的兴趣模型。 个性化搜索结果排序 模块 614, 用于将根据兴趣模型计算搜索结果得到的评分值, 对搜索结果进 行相关性排序, 例如, 将相关性高的搜索结果排在靠前的位置, 或将相关性 高的搜索结果排在竞价排名的搜索结果之后。 这样有利于用户能快速浏览到 其需要搜索的相关性高的结果。 最终搜索结果返回模块 616, 用于向搜索客 户端 502发送最终的搜索结果, 该搜索结果可以是经过过滤后的只包含相关 性比较高的搜索结果, 向用户提供部分搜索结果, 可以减少网络的流量, 减 轻搜索客户端 502的压力。
本实施例中, 成员搜索装置 508还可以包括: 搜索请求接收模块 606、 搜索处理模块 608和搜索结果返回模块 610。
搜索请求接收模块 606, 用于接收搜索请求, 本实施例中, 搜索请求可 以来自于搜索服务器 504, 搜索请求中可以包含搜索关键字, 而不包含用户 的用于搜索的个性化数据。 搜索处理模块 608, 用于根据搜索关键字进行搜 索得到搜索结果。 搜索结果返回模块 610, 用于返回搜索结果给搜索服务器 504
在一个实施例中通过根据用户的个性化数据提取的兴趣模型及所述搜索 请求进行搜索得到搜索结果, 使得用户得到搜索结果更加贴近用户需求, 不 同的用户可以得到不同的搜索结果, 使得搜索结果更加个性化, 有利于搜索 业务的推广使用。
请结合参看图 7, 一种搜索方法实施例, 可用于基于上述的搜索***, 用于为搜索客户端用户提供个性化搜索, 该个性化搜索可以根据搜索客户端 用户的兴趣模型提供相关的搜索结果, 该方法包括如下歩骤:
歩骤 702, 接收搜索客户端发送的搜索请求, 该搜索请求中包含一个或 多个搜索关键字。 该搜索请求可以是移动终端向网络侧发出的一个信号。
歩骤 704, 根据用户的个性化数据提取用户的兴趣模型。 本实施例中, 用户的个性化数据包括以下一种或多种: 静态用户档案 User Profile 搜索 历史、 位置信息或呈现信息。 兴趣模型为: 用户对 N个维度的评分值所组 成的兴趣模型向量, N>=2。
歩骤 706, 将所述搜索一个或多个搜索关键字和所述兴趣模型分发给一 个或多个成员搜索装置。 本实施例中, 携带用户兴趣模型将搜索请求分发给 成员搜索装置, 并指明采用统一的算法对搜索结果做个性化处理。
歩骤 708, 这一个或多个成员搜索装置完成搜索, 并应用搜索服务器发 送过来的用户的兴趣模型及指定的相同算法计算搜索结果的评分值, 根据评 分值对搜索结果进行排序。 同时也可以根据预设的阈值, 对搜索结果进行过 滤。
歩骤 710, 接收成员搜索装置返回的个性化搜索结果及相应的评分值。
歩骤 712, 根据搜索结果的评分值对各个成员搜索装置返回的搜索结果 进行综合的个性化相关性排序。 进行相关性排序后, 该方法还可以包括: 根 据评分值对搜索结果进行过滤包括: 根据设定的阈值, 保留评分值大于或等 于所述设定阈值的搜索结果, 该域值例如相关性大于等于 0.8的搜索结果予 以保留。
歩骤 714, 向搜索客户端用户发送最终的个性化搜索结果。
请结合参看图 8, 一种搜索方法, 可用于基于上述的搜索***, 用于为 搜索客户端用户提供个性化搜索, 该个性化搜索可以根据搜索客户端用户的 兴趣模型提供相关的搜索结果, 该方法包括如下歩骤:
歩骤 802, 搜索客户端将搜索请求发送给搜索服务器, 该搜索请求中包 含一个或多个搜索关键字, 本实施例中, 以移动终端为例, 移动终端向搜索 服务器发送搜索信号。
歩骤 804, 搜索服务器将搜索请求及用户 ID分发给成员搜索装置。 歩骤 806, 成员搜索装置根据用户 ID访问用户数据存储设备, 从用户 的个性化数据中提取用户兴趣模型。
歩骤 808, 各成员搜索装置完成搜索, 并根据提取的用户兴趣模型, 用 相同的个性化评分算法对搜索结果进行个性化相关性评分与排序。
歩骤 810, 成员搜索装置向搜索服务器返回个性化搜索结果及相应的评 分值。
歩骤 812, 根据搜索结果的评分值对各个成员搜索装置返回的搜索结果 进行综合的个性化相关性排序。 进行相关性排序后, 该方法还可以包括: 根 据评分值对搜索结果进行过滤包括: 根据设定的阈值, 保留评分值大于或等 于所述设定阈值的搜索结果, 该域值例如相关性大于等于 0.8的搜索结果予 以保留。
歩骤 814, 搜索服务器向搜索客户端用户返回最终的个性化搜索结果。
请结合参看图 9, 一种搜索方法, 可用于基于上述的搜索***, 用于为 搜索客户端用户提供个性化搜索, 该个性化搜索可以根据搜索客户端用户的 兴趣模型提供相关的搜索结果, 该方法包括如下歩骤:
歩骤 902, 搜索客户端将搜索请求发送给搜索服务器, 该搜索请求中包 含一个或多个搜索关键字, 本实施例中, 以移动终端为例, 移动终端向搜索 服务器发送搜索信号。
歩骤 904, 搜索服务器将搜索请求分发给成员搜索装置。
歩骤 906, 成员搜索装置根据这一个或多个关键字搜索得到搜索结果。 歩骤 908, 成员搜索装置向搜索服务器返回搜索结果。
歩骤 910, 根据用户数据提取用户的兴趣模型, 根据兴趣模型计算搜索 结果的评分值, 根据搜索结果的评分值对各个成员搜索装置返回的搜索结果 进行综合的个性化相关性排序。 进行相关性排序后, 该方法还可以包括: 根 据评分值对搜索结果进行过滤包括: 根据设定的阈值, 保留评分值大于或等 于所述设定阈值的搜索结果, 该域值例如相关性大于等于 0.8的搜索结果予 以保留。
歩骤 912, 搜索服务器向搜索客户端用户返回最终的个性化搜索结果。 以下举一个具体的应用实例说明
1.兴趣模型的定义
将用户的兴趣用 n个维度来表示如: 新闻、 体育、 娱乐、 财经、 科技、 房产、 游戏、 女性、 论坛、 天气、 商品、 家电、 音乐、 读书、 博客、 手机、 军事、 教育、 旅游、 彩信、 彩铃、 餐饮、 民航、 工业、 农业、 电脑、 地理 等。 用户对每个维度的兴趣的评分值所组成的一个向量 W "l,r2,r3,…… ,m) 则为用户的兴趣模型。
2.搜索服务器从用户数据中提取兴趣模型
(2) 用户的搜索点击历史所对应的动态兴趣模型 W2
W2=d 1 +d2+d3 + · · · +di+ · · · dm , 其中 di为用户某个点击文档所对应 的兴趣模型向量。
di =(tlAt3,…… .,tn), 当用户最新点击了这个文档, tj等于文档中类 型属于第 j个兴趣维度的所有词的词频之和。 如果用户对某个点击过的文档 进行评价, 如果评价好, di 向量乘以一个正的常数 c ( Ol ) 表示文档的重要 性增加 di = c*di = (c*ti,c*t2,c*t3,…… ,c*tn), 如果评价不好, di向量乘以一 个正的常数 c 的倒数表示文档的重要性减小 di = l/c*di =
( 1 /c*ti, 1 /c*t2, 1 /c*t3 , ,l/c*tn);
过了一段时间, tj 的值又自动减少一定的百分比, 表示随着时间的推移 其重要性减弱, 直到过了较长的时间 tj 的值减为零为止, 这时将 di从历史 记录中删除。 (例如, 每过一个月, tj的值减少 10% )
(3 ) 综合的兴趣模型, 将 W1和 W2分别归一化后相加, SP : 兴趣模 型 向 量 W=W1+W2 , 或 者 加 权 相 加 , 如 兴 趣 模 型 向 量 W=W1*30%+W2*70%, 然后再对 W进行归一化处理。 本领域普通技术人 员可知, 上述特征也可以应用与本发明的其他实施例中, 不再赘述。
3.元搜索引擎携带提取的兴趣模型数据向一个或多个成员搜索装置发起 个性化搜索请求, 并指明多个成员搜索装置采用某种指定的个性化算法对搜 索结果进行个性化的处理。
4.某个成员搜索装置用指定个性化算法进行个性化的搜索
( 1 ) 成员搜索装置根据倒排索引检索出候选的结果文档。
(2 ) 成员搜索装置根据兴趣模型数据和指定的个性化算法对候选的结 果文档进行个性化相关性评分与排序
算法 (a) : 兴趣模型向量 W=(rl,r2,r3,…… ,m)为搜索服务器传过来的 兴趣模型向量, D=(tl,t2,t3,…… .,ΐη)为文档所对应的兴趣模型向量
Ranking score = W *D =rl*tl+r2*t2+r3*t3+ ,+ rn*tn
或者
用算法 (b ) : W=(rl,r2,r3,…… ,m)为元搜索引擎传过来的兴趣模型向 用通用的文档分类算法如 Knn, Cvm等对文档进行分类, 分类的结果文 档属于 C的类型, 用类型 C与兴趣模型各个维度所属的类型相匹配, 将兴 趣模型中与文档类型相匹配的某个维度 i对应的评分值 ri付给该文档
Ranking score = ri.
(3 ) 成员搜索装置返回最相关 (评分值最大) 的 n个文档以及文档的个 性化相关性评分值
5.元搜索引擎将各个成员搜索装置返回的个性化搜索结果根据用同一种 算法算出来的相关性评分值的大小做一个综合的相关性排序, 并将最相关的 结果返回给搜索客户端。 上述实施例中的兴趣模型向量也可以应用与本发明 的其他实施例中, 不再赘述。
本领域普通技术人员可以理解, 上述方法中的全部或部分歩骤可以通过 程序指令相关的硬件完成, 该程序可以使得一个或多个计算机处理器执行上 述方法, 并且, 该程序可以存储于计算机可读存储介质中, 该计算机可读存 储介质, 例如 ROM、 RAM或光盘等。
在一个实施例中通过根据用户的个性化数据提取的兴趣模型及所述搜索 请求进行搜索得到搜索结果, 使得用户得到搜索结果更加贴近用户需求, 不 同的用户可以得到不同的搜索结果, 使得搜索结果更加个性化, 有利于搜索 业务的推广使用。
在一个实施例中个性化评分的过程由成员搜索装置来完成, 使得成员搜 索装置能够返回最相关的个性化搜索结果, 使得最后元搜索引擎综合的个性 化搜索结果更加精确。
在本发明实施例中, 指明各个成员搜索装置用同一种算法进行个性化的 处理, 使得各个成员搜索装置返回的评分具有可比性, 搜索服务器只需要对 各个成员搜索装置返回的评分做一个综合的排序, 就能达到搜索结果综合个 性化排序的目的, 而不用取回所有文档的快照进行实时切词与评分, 大大较 少来网络的流量, 大大减少了元搜索引擎的负担, 提高个性化搜索的效率。
请结合参看图 10, 是一个搜索***实施例的框架图, 在本实施例中, 搜索*** 1000包括: 搜索客户端 1002、 搜索服务器 1004、 用户数据存储设 备 1006、 一个或多个成员搜索装置 1008、 成员搜索服务器 1010及一个或多 个下一级成员引擎 1012。
搜索客户端 1002, 用于根据用户以文本方式或语音方式输入的关键 字, 向搜索服务器 1004发送搜索请求, 并接收搜索服务器 1004返回的搜索 结果。 本实施例中, 搜索客户端可以是个人电脑 (PC , Personal Computer ) 、 笔记本电脑 (NB, Notebook Computer ) 、 个人数字助理 ( PDA, Personal Digital Assistant) 、 手机 (HS, Handset ) 和智能光碟机 (IODD, Intelligent Optical Disk Drive) 等具有通信功能的终端设备。 本发 明实施例中, 以手机为例进行说明, 不再赘述。
搜索服务器 1004, 可以与一个或多个成员搜索装置进行搜索通信, 其 中成员搜索装置还包括成员搜索服务器 1010。 搜索服务器 1004用于接收搜 索请求, 携带用户兴趣模型, 将搜索请求分发给一个或多个成员搜索装置 1008和搜索服务器 1010, 并接收一个或多个成员搜索装置 1008或成员搜索 服务器 1010返回的根据兴趣模型的个性化搜索结果, 返回该搜索结果。 本 实施例中, 搜索服务器 1004从用户数据 (用户静态 profile、 搜索历史等) 中提取用户的兴趣模型或直接取出预先根据用户数据提取的兴趣模型, 携带
兴趣模型将个性化的搜索请求分发给一个或多个成员搜索装置 1008及成员 搜索服务器 1010, 并利用统一的算法 ID指明个性化评分算法对搜索结果进 行评分, 再返回个性化的搜索结果及评分后, 搜索服务器 1004对搜索结果 进行汇总, 并根据同一种个性化评分算法计算出来的搜索结果的评分值对搜 索结果进行综合性排序, 然后将最后的个性化搜索结果返回给搜索客户端 1002
其中, 用户数据存储设备 1006, 用于存储用户数据, 包括: 用户的兴 趣模型, 例如用户静态的 Profile (描述) 、 兴趣爱好、 搜索历史、 位置信息 和呈现信息等。 在本实施例中, 用户数据存储设备 1006可以设置于运营商 侧的内部***中。
成员搜索装置 1008可以为单独的垂直引擎, 成员搜索服务器 1010可以 再连接下一级成员引擎 1012。 成员搜索装置 1008根据搜索请求搜索得到搜 索结果, 并根据用户的兴趣模型计算该搜索结果的评分值, 将搜索结果和评 分值返回给搜索服务器 1004, 成员搜索装置 1008也可以排序后发送给搜索 服务器 1004。 成员搜索服务器 1010可以将搜索请求分发给下一级成员引擎 1012, 成员搜索服务器 1010或下一级成员引擎 1012也可以针对搜索结果进 行个性化处理, 不再赘述。
请结合参看图 11, 其中搜索服务器 1004还可以包括: 搜索请求接收模 块 1102、 兴趣模型提取模块 1104、 搜索请求分发模块 1106、 个性化搜索结 果排序模块 1116和最终搜索结果返回模块 1118。
搜索请求接收模块 1102, 用于接收搜索客户端 1002的搜索请求, 该搜 索请求中可以包含用户输入的一个或多个搜索关键字, 该关键字可以由用户 以文本方式或语音方式输入。 兴趣模型提取模块 1104, 用于根据用户的个 性化数据提取用户的兴趣模型, 本实施例中, 用户数据可以包括: 用户的静 态 Profile 用户的搜索历史、 用户的位置信息、 用户的呈现信息等。 搜索请 求分发模块 1106, 用于将携带兴趣模型的个性化搜索请求分发给一个或多
个成员搜索装置 1008和成员搜索服务器 1010。 同时, 个性化搜索请求分发 模块 1106可以同时为这一个或多个成员搜索装置 1008指定统一的个性化评 分算法对搜索结果进行个性化处理, 该统一的个性化评分算法可以通过一个 算法标识 ID来表示。 个性化搜索结果排序模块 1116, 用于对成员搜索装置 1008及成员搜索服务器 1010的搜索结果进行汇总, 并根据一个性化评分算 法进行综合排序。 例如, 将相关性高的搜索结果排在靠前的位置, 或将相关 性高的搜索结果排在竞价排名的搜索结果之后。 这样有利于用户能快速浏览 到其需要搜索的相关性高的结果, 在其他实施例中, 该排序还包括成员搜索 装置的优先级信息和相关因素信息 (如价格因素信息等) 。 最终搜索结果返 回模块 1118, 用于向搜索客户端 1002发送最终的搜索结果, 该搜索结果可 以是经过过滤后的只包含相关性比较高的搜索结果, 向用户提供部分搜索结 果, 可以减少网络的流量, 减轻搜索客户端 1002的压力。
成员搜索装置 1008还可以包括: 搜索请求接收模块 1108、 搜索处理模 块 1110、 搜索结果个性化处理模块 1112及搜索结果返回模块 1114。
搜索请求接收模块 1108, 用于接收搜索请求, 本实施例中, 搜索请求 可以来自于搜索服务器 1004, 个性化搜索请求接收模块 1108还可以接收搜 索服务器 1004发送的用户兴趣模型和评分算法 ID。 搜索处理模块 1110, 用于根据搜索关键字进行搜索得到搜索结果。 搜索结果个性化处理模块 1112, 用于根据用户的兴趣模型对搜索结果进行个性化处理, 同时, 也可以 采用被指定的统一的个性化评分算法。 个性化搜索结果返回模块 1114, 用 于返回搜索结果, 也可以返回搜索结果的评分值, 本实施例中, 返回搜索结 果及评分值给搜索服务器 1004。
请结合参看图 12, 是一个搜索***实施例的框架图, 该实施例中搜索 *** 1200通过向搜索服务器仅发送搜索请求, 而不发送用户的兴趣模型数 据, 由搜索服务器完成个性化处理。 搜索*** 1200包括: 搜索客户端
1202、 搜索服务器 1204、 用户数据存储设备 1206、 一个或多个成员搜索装 置 1208、 成员搜索服务器 1210及一个或多个下一级成员引擎 1212。
搜索客户端 1202, 用于根据用户以文本方式或语音方式输入的关键 字, 向搜索服务器 1204发送搜索请求, 并接收搜索服务器 1204返回的搜索 结果。
搜索服务器 1204, 可以与一个或多个成员搜索装置 1208进行搜索通 信, 其中成员搜索装置还包括成员搜索服务器 1210。 搜索服务器 1204用于 接收搜索请求, 将搜索请求分发给一个或多个成员搜索装置 1208或搜索服 务器 1210, 并接收一个或多个成员搜索装置 1208或成员搜索服务器 1210 返回的搜索结果, 将搜索结果进行个性化处理后返回该个性化处理后的搜索 结果。 本实施例中, 搜索服务器 1204从用户数据 (用户静态 profile、 搜索 历史等) 中提取用户的兴趣模型或直接取出预先根据用户数据提取的兴趣模 型, 根据该兴趣模型对搜索结果进行个性化处理。
其中用户数据存储设备 1206, 用于存储用户数据, 包括: 用户的兴趣 模型, 例如用户静态的 Profile (描述) 、 兴趣爱好、 搜索历史、 位置信息和 呈现信息等。 在本实施例中, 用户数据存储设备 1206可以设置于运营商侧 的内部***中。
成员搜索装置 1208可以为单独的垂直引擎, 成员搜索服务器 1210可以 再连接下一级成员引擎 1212。 成员搜索装置 1208根据搜索请求搜索得到搜 索结果, 并返回给搜索服务器 1204, 其中成员搜索服务器 1210可以将搜索 请求分发给下一级成员引擎 1212, 由下一级成员引擎 1212完成搜索。
请结合参看图 13, 其中搜索服务器 1204还可以包括: 搜索请求接收模 块 1302、 搜索请求分发模块 1304、 兴趣模型提取模块 1312、 个性化搜索结 果排序模块 1314和最终搜索结果返回模块 1316。
搜索请求接收模块 1302, 用于接收搜索客户端 1202的搜索请求, 该搜 索请求中可以包含用户输入的一个或多个搜索关键字, 该关键字可以由用户
以文本方式或语音方式输入。 搜索请求分发模块 1304, 用于将搜索请求分 发给成员搜索装置 1208和成员搜索服务器 1210。 兴趣模型提取模块 1312, 用于根据用户的个性化数据提取用户的兴趣模型, 本实施例中, 用户数据可 以包括: 用户的静态 Profile 用户的搜索历史、 用户的位置信息、 用户的呈 现信息等。 个性化搜索排序模块 1314, 用于对成员搜索装置 1208和成员搜 索服务器 1210的搜索结果进行汇总, 并根据兴趣模型提取模块 1312提取的 兴趣模型计算搜索结果的评分值, 并根据该评分值对搜索结果进行排序。 例 如, 将相关性高的搜索结果排在靠前的位置, 或将相关性高的搜索结果排在 竞价排名的搜索结果之后。 这样有利于用户能快速浏览到其需要搜索的相关 性高的结果, 在其他实施例中, 该排序还包括成员搜索装置的级别评分信息 和相关因素信息 (如价格因素信息等) 。 最终搜索结果返回模块 1316, 用 于向搜索客户端 1202发送最终的搜索结果, 该搜索结果可以是经过过滤后 的只包含相关性比较高的搜索结果, 向用户提供部分搜索结果, 可以减少网 络的流量, 减轻搜索客户端 1202的压力。
成员搜索装置 1208还可以包括: 搜索请求接收模块 1306、 搜索处理模 块 1308和搜索结果返回模块 1310。
搜索请求接收模块 1306, 用于接收搜索请求, 本实施例中, 搜索请求 来自与搜索服务器 1204, 不包含用户的兴趣模型数据。 搜索处理模块 1308, 用于根据搜索关键字进行搜索得到搜索结果。 搜索结果返回模块 1310, 用于返回搜索结果。
请结合参看图 14, 是一个搜索***实施例的框架图, 在本实施例中, 搜索服务器将搜索请求及兴趣模型分发给成员搜索装置, 成员搜索装置根据 兴趣模型计算搜索结果的评分值, 返回给搜索服务器, 搜索服务器进行个性 化重评分 (re-ranking ) 和排序, 得到个性化搜索结果后返回给搜索客户 端。 搜索*** 1400包括: 搜索客户端 1402、 搜索服务器 1404、 用户数据存
储设备 1406、 成员搜索装置 1408、 成员搜索服务器 1410及下一级成员引擎 搜索客户端 1402, 用于根据用户以文本方式或语音方式输入的关键 字, 向搜索服务器 1404发送搜索请求, 并接收搜索服务器 1404返回的搜索 结果。
搜索服务器 1404, 可以与一个或多个成员搜索装置 1408进行搜索通 信, 其中成员搜索装置还包括成员搜索服务器 1410。 搜索服务器 1404用于 接收搜索请求, 携带用户兴趣模型, 将搜索请求分发给一个或多个成员搜索 装置 1408和搜索服务器 1410, 并接收成员搜索装置 1408和成员搜索服务 器 1410返回的根据兴趣模型的个性化搜索结果, 进行 re-ranking, 根据 re- ranking的结果对搜索结果进行排序, 将排序后的搜索结果返回给搜索客户 端 1402。 其中搜索服务器 1404携带用户兴趣模型分发搜索请求时, 可以指 定或不指定评分算法。
其中, 用户数据存储设备 1406, 用于存储用户数据, 包括: 用户的兴 趣模型, 例如用户静态的 Profile (描述) 、 兴趣爱好、 搜索历史、 位置信息 和呈现信息等。 在本实施例中, 用户数据存储设备 1006可以设置于运营商 侧的内部***中。 本实施例中, 用户数据设备 1406与搜索服务器 1404相 连。
成员搜索装置 1408可以为单独的垂直引擎, 成员搜索服务器 1410可以 再连接下一级成员引擎 1412, 其功能可以与搜索服务器 1404类似, 也可以 不同。 成员搜索装置 1408根据搜索请求搜索得到搜索结果, 并根据用户的 兴趣模型计算该搜索结果的评分值。 其中, 该计算评分值在搜索服务器 1404指定评分算法时可以用指定评分算法计算搜索结果的评分值, 若没有 指定评分算法, 成员搜索装置 1408可以利用其私有算法计算搜索结果的评 分值。 成员搜索装置 1408将搜索结果和评分值返回给搜索服务器 1404。 成 员搜索服务器 1410可以将搜索请求分发给下一级成员引擎 1412, 成员搜索
服务器 1410或下一级成员引擎 1412也可以针对搜索结果进行个性化处理, 不再赘述。
请结合参看图 15, 其中搜索服务器 1404包括: 搜索请求接收模块 1502、 兴趣模型提取模块 1504、 搜索请求分发模块 1506、 重评分模块 1516、 个性化搜索请求排序模块 1518和最终搜索结果返回模块 1520。
搜索请求接收模块 1502, 用于接收搜索客户端 1402的搜索请求, 该搜 索请求可以包含用户输入的一个或多个搜索关键字, 该关键字可以由用户以 文本方式或语音方式输入。 兴趣模型提取模块 1504, 用于根据用户的个性 化数据提取用户的兴趣模型或取出预先存储的提取好的用户兴趣模型, 本实 施例中, 用户数据可以包括: 用户的静态 Profile 用户的搜索历史、 用户的 位置信息、 用户的呈现信息等。 搜索请求分发模块 1506, 用于将携带兴趣 模型的个性化搜索请求分发给一个或多个成员搜索装置 1408和成员搜索服 务器 1410。 同时, 个性化搜索请求分发模块 1506可以同时为这一个或多个 成员搜索装置 1408和成员搜索服务器 1410指定统一的个性化评分算法对搜 索结果进行个性化处理, 该统一的个性化评分算法可以通过一个算法标识 ID来表示。 重评分模块 1516, 用于对个成员搜索装置 1408和成员搜索服务 器 1410返回的搜索结果进行重评分 (re-ranking) 。 其中, re-ranking包括: 根据提取的兴趣模型计算搜索结果的评分值, 并根据该评分值对搜索结果进 行排序。 例如, 将相关性高的搜索结果排在靠前的位置, 或将相关性高的搜 索结果排在竞价排名的搜索结果之后。 这样有利于用户能快速浏览到其需要 搜索的相关性高的结果, 在其他实施例中, 该排序还包括成员搜索装置的级 别评分信息和相关因素信息 (如价格因素信息等) 。 例如, 计算级别评分信 息, 计算公式如下:
P=rl*返回的评分值 +r2*级别因子
其中, P表示级别评分, rl表示返回的评分值占的权重, r2表示级别因 子占的权重, 所述返回的评分值为成员搜索装置返回的评分值, 所述级别因 子表示成员搜索装置级别。
计算综合评分信息, 计算公式如下:
R=P+r3 *价格因素评分值
其中, R表示综合评分, r3表示价格因素评分权重, rl+r2+r3=l。
个性化搜索结果排序模块 1518, 用于根据重新评分的评分值对搜索结 果进行综合性排序, 例如, 将相关性高的搜索结果排在靠前的位置, 或将相 关性高的搜索结果排在竞价排名的搜索结果之后。 这样有利于用户能快速浏 览到其需要搜索的相关性高的结果, 在其他实施例中, 该排序还包括成员搜 索装置的优先级信息和相关因素信息 (如价格因素信息等) 。 最终搜索结果 返回模块 1520, 用于向搜索客户端 1402发送最终的搜索结果, 该搜索结果 可以是经过过滤后的只包含相关性比较高的搜索结果, 向用户提供部分搜索 结果, 可以减少网络的流量, 减轻搜索客户端 1402的压力。
成员搜索装置 1408还包括: 搜索请求接收模块 1508、 搜索处理模块
1510、 搜索结果个性化处理模块 1512和搜索结果返回模块 1514。
搜索请求接收模块 1508, 用于接收搜索请求, 本实施例中, 搜索请求 可以来自于搜索服务器 1404, 搜索请求接收模块 1508还可以接收搜索服务 器 1404发送的用户兴趣模型和评分算法 ID。 搜索处理模块 1510, 用于根 据搜索关键字进行搜索得到搜索结果。 搜索结果个性化处理模块 1512, 用 于根据用户的兴趣模型对搜索结果进行个性化处理, 同时, 也可以采用被指 定的统一的个性化评分算法, 若没有指定统一算法时, 采用私有算法对搜索 结果进行个性化处理。 搜索结果返回模块 1514, 用于返回搜索结果, 也可 以返回搜索结果的评分值, 本实施例中, 返回搜索结果及评分值给搜索服务 器 1404。
请结合参看图 16, 是一个搜索***实施例的框架图。 本实施例中, 搜 索*** 1600 由应用服务器提取用户的兴趣模型或取出预先存储的提取好的 兴趣模型, 并由搜索服务器 1608进行个性化处理。 搜索*** 1600包括: 搜 索客户端 1602、 用户数据存储设备 1604、 应用服务器 1606、 搜索服务器 1608、 成员搜索装置 1610、 成员搜索服务器 1612及下一级成员引擎 1614。
搜索客户端 1602, 用于根据用户以文本方式或语音方式输入的关键 字, 向应用服务器 1606发送搜索请求, 并接应用服务器 1606返回的搜索结 果。
用户数据存储设备 1604, 用于存储用户数据, 包括: 用户的兴趣模 型, 例如用户静态的 Profile (描述) 、 兴趣爱好、 搜索历史、 位置信息和呈 现信息等。 在本实施例中, 用户数据存储设备 1604可以设置于运营商侧的 内部***中。
应用服务器 1606与用户数据存储设备 1604相连, 用于提取用户的兴趣 模型或取出预先存储的提取的用户的兴趣模型, 将接收到的搜索请求和兴趣 模型发送给搜索服务器 1608, 并接收搜索服务器 1608返回的个性化搜索结 果, 返回给搜索客户端 1602。 本实施例中, 应用服务器 1606从用户数据 (用户静态 profile 搜索历史等) 中提取用户的兴趣模型或直接取出预先根 据用户数据提取的兴趣模型, 携带兴趣模型将个性化的搜索请求分发给搜索 服务器 1608。
搜索服务器 1608, 可以与一个或多个成员搜索装置进行通信, 其中, 成员搜索装置还可以包括成员搜索服务器。 搜索服务器 1608用于接收应用 服务器 1606的搜索请求和用户的兴趣模型, 将搜索请求和兴趣模型分发给 成员搜索装置 1610和成员搜索服务器 1612, 接收返回的个性化搜索结果和 评分值, 将搜索结果汇总后, 进行综合重评分 re-ranking, 并返回给应用服 务器 1606。
成员搜索装置 1610可以为单独的垂直引擎, 成员搜索服务器 1612可以 再连接下一级成员引擎 1614。 成员搜索装置 1610根据搜索请求搜索得到搜 索结果, 并根据用户的兴趣模型计算该搜索结果的评分值, 将搜索结果和评 分值返回给搜索服务器 1608, 成员搜索装置 1610也可以排序后发送给搜索 服务器 1608。 成员搜索服务器 1612可以将搜索请求分发给下一级成员引擎 1614, 成员搜索服务器 1612或下一级成员引擎 1614也可以针对搜索结果进 行个性化处理, 不再赘述。
请结合参看图 17, 其中应用服务器 1606还可以包括: 兴趣模型提取模 块 1702、 搜索请求发送模块 1704、 搜索结果接收模块 1724。
兴趣模型提取模块 1702, 用于根据用户的个性化数据提取用户的兴趣 模型或取出预先存储的提取好的用户兴趣模型, 本实施例中, 用户数据可以 包括: 用户的静态 Profile 用户的搜索历史、 用户的位置信息、 用户的呈现 信息等。 搜索请求发送模块 1704, 用于将搜索请求和兴趣模型发送给搜索 服务器 1608。 搜索结果接收模块 1724, 用于接收搜索服务器 1608返回的个 性化搜索结果, 返回给搜索客户端 1602。
搜索服务器 1608还包括: 搜索请求接收模块 1706、 搜索请求分发模块 1708、 重评分模块 1718、 个性化搜索结果排序模块 1720和搜索结果返回模 块 1722ο
搜索请求接收模块 1706, 用于接收应用服务器 1606发送的搜索请求和 用户的兴趣模型。 搜索请求分发模块 1708, 用于将搜索请求和用户的兴趣 模型分发给成员搜索装置, 并指定或不指定统一的评分算法。 重评分模块 1718, 用于接收各成员搜索装置 1610返回的搜索结果和评分值, 将搜索结 果汇总后, 进行重评分 (re-ranking) 。 其中, re-ranking包括: 根据提取的 兴趣模型计算搜索结果的评分值, 并根据该评分值对搜索结果进行排序。 例 如, 将相关性高的搜索结果排在靠前的位置, 或将相关性高的搜索结果排在 竞价排名的搜索结果之后。 这样有利于用户能快速浏览到其需要搜索的相关
性高的结果, 在其他实施例中, 该排序还包括成员搜索装置的级别评分信息 和相关因素信息 (如价格因素信息等) 。 例如, 计算级别评分信息, 计算公 式如下:
P=rl*返回的评分值 +r2*级别因子
其中, P表示级别评分, rl表示返回的评分值占的权重, r2表示级别因 子占的权重, 所述返回的评分值为成员搜索装置返回的评分值, 所述级别因 子表示成员搜索装置级别。
计算综合评分信息, 计算公式如下:
R=P+r3*价格因素评分值
其中, R表示综合评分, r3表示价格因素评分权重, rl+r2+r3=l。
个性化搜索结果排序模块 1720, 用于根据 re-mnking的结果对搜索结果 进行排序, 例如, 将相关性高的搜索结果排在靠前的位置, 或将相关性高的 搜索结果排在竞价排名的搜索结果之后。 这样有利于用户能快速浏览到其需 要搜索的相关性高的结果, 在其他实施例中, 该排序还包括成员搜索装置的 优先级信息和相关因素信息 (如价格因素信息等) 。 搜索结果返回模块 1722, 用于向应用服务器 1606返回搜索结果, 该搜索结果可以是经过过滤 后的只包含相关性比较高的搜索结果, 向用户提供部分搜索结果, 可以减少 网络的流量, 减轻搜索客户端 1602的压力。
成员搜索装置 1610还可以包括: 搜索请求接收模块 1710、 搜索处理模 块 1712、 搜索结果个性化处理模块 1714和搜索结果返回模块 1716。
搜索请求接收模块 1710, 用于接收搜索请求, 本实施例中, 搜索请求 可以来自于搜索服务器 1608, 搜索请求接收模块 1710还可以接收搜索服务 器 1608发送的用户兴趣模型和评分算法 ID。 搜索处理模块 1712, 用于根 据搜索关键字进行搜索得到搜索结果。 搜索结果个性化处理模块 1714, 用 于根据用户的兴趣模型对搜索结果进行个性化处理, 同时, 也可以采用被指 定的统一的个性化评分算法, 若没有指定统一算法时, 采用私有算法对搜索
结果进行个性化处理。 搜索结果返回模块 1716, 用于返回搜索结果, 也可 以返回搜索结果的评分值, 本实施例中, 返回搜索结果及评分值给搜索服务 器画。
请结合参看图 18, 是一个搜索***实施例的框架图。 本实施例中, 搜 索*** 1800 由应用服务器提取用户的兴趣模型或取出预先存储的提取好的 兴趣模型, 由搜索服务器 1808 进行个性化处理, 但不需要重评分 re- ranking。 搜索*** 1800包括: 搜索客户端 1802、 用户数据存储设备 1804、 应用服务器 1806、 搜索服务器 1808、 成员搜索装置 1810、 成员搜索 服务器 1812和下一级成员引擎 1814。
搜索客户端 1802, 用于根据用户以文本方式或语音方式输入的关键 字, 向应用服务器 1806发送搜索请求, 并接应用服务器 1806返回的搜索结 果。
用户数据存储设备 1804, 用于存储用户数据, 包括: 用户的兴趣模 型, 例如用户静态的 Profile (描述) 、 兴趣爱好、 搜索历史、 位置信息和呈 现信息等。 在本实施例中, 用户数据存储设备 1804可以设置于运营商侧的 内部***中。
应用服务器 1806与用户数据存储设备 1804相连, 用于提取用户的兴趣 模型或取出预先存储的提取好的用户的兴趣模型, 将接收到搜索请求和兴趣 模型发送给搜索服务器 1808, 并接收搜索服务器 1808返回的个性化搜索结 果, 返回给搜索客户端 1802。 本实施例中, 应用服务器 1806从用户数据 (用户静态 profile 搜索历史等) 中提取用户的兴趣模型或直接取出预先根 据用户数据提取的兴趣模型, 携带兴趣模型将个性化的搜索请求分发给搜索 服务器 1808。
搜索服务器 1808, 可以与一个或多个成员搜索装置进行通信, 其中, 成员搜索装置还可以包括成员搜索服务器。 搜索服务器 1808用于接收应用 服务器 1806的搜索请求和用户的兴趣模型, 将搜索请求和兴趣模型分发给
成员搜索装置 1810和成员搜索服务器 1812, 并利用统一的算法 ID指明个 性化评分算法对搜索结果进行评分, 接收返回的个性化搜索结果和评分值, 将搜索结果汇总后, 根据各成员搜索装置返回的评分值进行综合评分, 并返 回给应用服务器 1806。
成员搜索装置 1810可以为单独的垂直引擎, 成员搜索服务器 1812可以 再连接下一级成员引擎 1814。 成员搜索装置 1810根据搜索请求搜索得到搜 索结果, 并根据用户的兴趣模型计算该搜索结果的评分值, 将搜索结果和评 分值返回给搜索服务器 1808, 成员搜索装置 1810也可以排序后发送给搜索 服务器 1808。 成员搜索服务器 1812可以将搜索请求分发给下一级成员引擎 1814, 成员搜索服务器 1812或下一级成员引擎 1814也可以针对搜索结果进 行个性化处理, 不再赘述。
请结合参看图 19, 其中应用服务器 1806还可以包括: 兴趣模型提取模 块 1902、 搜索请求发送模块 1904、 搜索结果接收模块 1922。
兴趣模型提取模块 1902, 用于根据用户的个性化数据提取用户的兴趣 模型或取出预先存储的提取好的用户兴趣模型, 本实施例中, 用户数据可以 包括: 用户的静态 Profile 用户的搜索历史、 用户的位置信息、 用户的呈现 信息等。 搜索请求发送模块 1904, 用于将搜索请求和兴趣模型发送给搜索 服务器 1808。 搜索结果接收模块 1922, 用于接收搜索服务器 1808返回的个 性化搜索结果, 返回给搜索客户端 1802。
搜索服务器 1808还包括: 搜索请求接收模块 1906、 搜索请求分发模块
1908、 个性化搜索结果排序模块 1918和搜索结果返回模块 1920。
搜索请求接收模块 1906, 用于接收应用服务器 1806发送的搜索请求和 用户的兴趣模型。 搜索请求分发模块 1908, 用于将搜索请求和用户的兴趣 模型分发给成员搜索装置, 并指定统一的评分算法, 该指定统一评分算法可 以通过算法 ID来表示。 个性化搜索结果排序模块 1918, 用于根据返回的搜 索结果的评分值对搜索结果进行综合的排序, 例如, 将相关性高的搜索结果
排在靠前的位置, 或将相关性高的搜索结果排在竞价排名的搜索结果之后。 这样有利于用户能快速浏览到其需要搜索的相关性高的结果, 在其他实施例 中, 该排序还包括成员搜索装置的优先级信息和相关因素信息 (如价格因素 信息等) 。 搜索结果返回模块 1920, 用于向应用服务器 1806返回搜索结 果, 该搜索结果可以是经过过滤后的只包含相关性比较高的搜索结果, 向用 户提供部分搜索结果, 可以减少网络的流量, 减轻搜索客户端 1802 的压 力。
成员搜索装置 1810还可以包括: 搜索请求接收模块 1910、 搜索处理模 块 1912、 搜索结果个性化处理模块 1914和搜索结果返回模块 1916。
搜索请求接收模块 1910, 用于接收搜索请求, 本实施例中, 搜索请求 可以来自于搜索服务器 1808, 搜索请求接收模块 1910还可以接收搜索服务 器 1808发送的用户兴趣模型和评分算法 ID。 搜索处理模块 1912, 用于根 据搜索关键字进行搜索得到搜索结果。 搜索结果个性化处理模块 1914, 用 于根据用户的兴趣模型对搜索结果进行个性化处理, 同时, 也可以采用被指 定的统一的个性化评分算法。 搜索结果返回模块 1916, 用于返回搜索结 果, 也可以返回搜索结果的评分值, 本实施例中, 返回搜索结果及评分值给 搜索服务器 1808。
请结合参看图 20, 是个搜索***实施例的框架图。 本实施例中, 搜索 *** 2000 由应用服务器提取用户的兴趣模型或取出预先存储的提取好的兴 趣模型, 由搜索服务器 2008进行个性化处理, 不需要将兴趣模型传递给成 员搜索装置和成员搜索服务器。 搜索*** 2000包括: 搜索客户端 2002、 用 户数据存储设备 2004、 应用服务器 2006、 搜索服务器 2008、 成员搜索装置 2010、 成员搜索服务器 2012和下一级成员引擎 2014。
搜索客户端 2002, 用于根据用户以文本方式或语音方式输入的关键 字, 向应用服务器 2006发送搜索请求, 并接应用服务器 2006返回的搜索结 果。
用户数据存储设备 2004, 用于存储用户数据, 包括: 用户的兴趣模 型, 例如用户静态的 Profile (描述) 、 兴趣爱好、 搜索历史、 位置信息和呈 现信息等。 在本实施例中, 用户数据存储设备 2004可以设置于运营商侧的 内部***中。
应用服务器 2006与用户数据存储设备 2004相连, 用于提取用户的兴趣 模型或取出预先存储的提取好的用户的兴趣模型, 将接收到搜索请求和兴趣 模型发送给搜索服务器 2008, 并接收搜索服务器 2008返回的个性化搜索结 果, 返回给搜索客户端 2002。 本实施例中, 应用服务器 2006从用户数据 (用户静态 profile 搜索历史等) 中提取用户的兴趣模型或直接取出预先根 据用户数据提取的兴趣模型, 携带兴趣模型将个性化的搜索请求分发给搜索 服务器 2008。
搜索服务器 2008, 可以与一个或多个成员搜索装置进行通信, 其中, 成员搜索装置还可以包括成员搜索服务器。 搜索服务器 2008用于接收应用 服务器 2006的搜索请求和用户的兴趣模型, 将搜索请求分发给成员搜索装 置 2010和成员搜索服务器 2012, 接收返回的搜索结果, 根据用户的兴趣模 型计算搜索结果的评分值, 并对搜索结果根据该评分值进行排序后发送应用 服务器 2006。
成员搜索装置 2010可以为单独的垂直引擎, 成员搜索服务器 2012可以 再连接下一级成员引擎 2014。 成员搜索装置 2010根据搜索请求得到搜索结 果, 返回给搜索服务器 2008, 成员搜索服务器 2012可以将搜索请求分发给 下一级成员引擎 2014进行搜索处理。
请结合参看图 21, 其中应用服务器 2006还可以包括: 兴趣模型提取模 块 2102、 搜索请求发送模块 2104、 搜索结果接收模块 2120。
兴趣模型提取模块 2102, 用于根据用户的个性化数据提取用户的兴趣 模型或取出预先存储的提取好的用户兴趣模型, 本实施例中, 用户数据可以 包括: 用户的静态 Profile 用户的搜索历史、 用户的位置信息、 用户的呈现
信息等。 搜索请求发送模块 2104, 用于将搜索请求和兴趣模型发送给搜索 服务器 2008。 搜索结果接收模块 2120, 用于接收搜索服务器 2008返回的个 性化搜索结果, 返回给搜索客户端 2002。
搜索服务器 2008还可以包括: 搜索请求接收模块 2106、 搜索请求分发 模块 2108、 个性化搜索结果排序模块 2116和搜索结果返回模块 2118。
搜索请求接收模块 2106, 用于接收应用服务器 2006发送的搜索请求和 用户的兴趣模型。 搜索请求分发模块 2108, 用于将搜索请求分发给成员搜 索装置。 个性化搜索结果排序模块 2116, 用于接收返回的搜索结果, 并根 据用户的兴趣模型计算搜索结果的评分值, 并根据该评分值对搜索结果进行 排序, 例如, 将相关性高的搜索结果排在靠前的位置, 或将相关性高的搜索 结果排在竞价排名的搜索结果之后。 这样有利于用户能快速浏览到其需要搜 索的相关性高的结果, 在其他实施例中, 该排序还包括成员搜索装置的优先 级信息和相关因素信息 (如价格因素信息等) 。 搜索结果返回模块 2118, 用于向应用服务器 2006返回搜索结果, 该搜索结果可以是经过过滤后的只 包含相关性比较高的搜索结果, 向用户提供部分搜索结果, 可以减少网络的 流量, 减轻搜索客户端 2002的压力。
成员搜索装置 2010还可以包括: 搜索请求接收模块 2110、 搜索处理模 块 2112和搜索结果返回模块 2114。 其功能与前述成员搜索装置 1208功能 相同, 不再赘述。
在本发明的实施例中通过根据用户的个性化数据提取的兴趣模型及所述 搜索请求进行搜索得到搜索结果, 使得用户得到搜索结果更加贴近用户需 求, 不同的用户可以得到不同的搜索结果, 使得搜索结果更加个性化, 有利 于搜索业务的推广使用。 此外, 通过成员搜索装置来完成个性化的评分, 使 得成员搜索装置能够返回最相关的搜索结果, 搜索服务器在进行综合搜索结 果时, 使搜索结果更加精确。 通过指明各成员搜索装置利用统一的算法来进
行个性化处理, 使得各成员搜索装置的返回的评分具有可比性, 大大减少了 网络的流量, 提高了个性化的效率。
请结合参看图 22, 是一个搜索方法实施例的流程图。 该搜索方法包括 以下歩骤:
歩骤 2202, 接收搜索客户端发送的搜索请求, 该搜索请求中包含一个 或多个搜索关键字, 该搜索关键字可以由用户以文本方式或语音方式输入。 该搜索请求可以是移动终端向网络侧发出的一个信号。
歩骤 2204, 根据用户的个性化数据提取用户的兴趣模型或取出预先存 储的提取好的兴趣模型。 本实施例中, 用户的个性化数据包括以下一种或多 种: 静态用户档案 User Profile 搜索历史、 位置信息或呈现信息。 兴趣模 型为: 用户对 N个维度的评分值所组成的兴趣模型向量, N>=2。
歩骤 2206, 将所述搜索一个或多个搜索关键字和所述兴趣模型分发给 一个或多个成员搜索装置或成员搜索服务器。 本实施例中, 携带用户兴趣模 型将搜索请求分发给成员搜索装置, 并指明或不指明采用统一的算法对搜索 结果做个性化处理, 若指明统一的算法可以通过一个算法 ID指示。
歩骤 2208, 成员搜索装置完成搜索, 若有指定的算法, 用指定的相同 的个性化评分算法计算搜索结果的评分值, 若没有指定的算法, 成员搜索装 置可以用私有算法计算搜索结果的评分值, 并根据该评分值对搜索结果进行 排序。 同时也可以根据预设的阈值, 对搜索结果进行过滤。
歩骤 2210, 接收成员搜索装置返回的个性化搜索结果及相应的评分 值。
歩骤 2212, 根据搜索结果的评分值和相关因素 (成员搜索装置级别和 价格因素等) 对搜索结果进行重评分。 其中, re-ranking包括: 根据提取的 兴趣模型计算搜索结果的评分值。 例如, 计算级别评分信息, 计算公式如 下:
P=rl*返回的评分值 +r2*级别因子
其中, P表示级别评分, rl表示返回的评分值占的权重, r2表示级别因 子占的权重, 所述返回的评分值为成员搜索装置返回的评分值, 所述级别因 子表示成员搜索装置级别。
计算综合评分信息, 计算公式如下:
R=P+r3 *价格因素评分值
其中, R表示综合评分, r3表示价格因素评分权重, rl+r2+r3=l。
歩骤 2214, 根据 re-ranking的结果对搜索结果进行排序。 例如, 将相关 性高的搜索结果排在靠前的位置, 或将相关性高的搜索结果排在竞价排名的 搜索结果之后。 这样有利于用户能快速浏览到其需要搜索的相关性高的结 果, 在其他实施例中, 该排序还包括成员搜索装置的级别评分信息和相关因 素信息 (如价格因素信息等) 。
歩骤 2216, 将最终搜索结果返回给搜索客户端。 返回的搜索结果可以 是经过过滤后的只包含相关性比较高的搜索结果, 向用户提供部分搜索结 果, 可以减少网络的流量, 减轻搜索客户端的压力。
请结合参看图 23, 是一个搜索方法实施例的流程图。 该实施例中, 搜 索服务器接收来来自于应用服务器的搜索请求, 将搜索结果返回给应用服务 器, 其中, 应用服务器提供用户的兴趣模型。
歩骤 2302, 搜索客户端将搜索请求发送给应用服务器, 该搜索请求中 包含一个或多个搜索关键字, 该搜索关键字可以由用户以文本方式或语音方 式输入。 该搜索请求可以是移动终端向网络侧发出的一个信号。
歩骤 2304, 应用服务器从用户的个性化数据 (如用户的静态 profile和 用户的点击历史) 中提取用户兴趣模型, 或直接取出预先根据用户个性化数 据提取的兴趣模型。 兴趣模型可以为: 用户对 N个维度的评分值所组成的 兴趣模型向量, N>=2。
歩骤 2306, 应用服务器携带用户的兴趣模型, 将搜索请求发送给搜索 服务器。
歩骤 2308, 将所述搜索一个或多个搜索关键字和所述兴趣模型分发给 一个或多个成员搜索装置。 本实施例中, 携带用户兴趣模型将搜索请求分发 给成员搜索装置, 并指明或不指明采用统一的算法对搜索结果做个性化处 理, 若指明统一的算法可以通过一个算法 ID指示。
歩骤 2310, 成员搜索装置完成搜索, 若有指定的算法, 用指定的相同 的个性化评分算法计算搜索结果的评分值, 若没有指定的算法, 成员搜索装 置可以用私有算法计算搜索结果的评分值, 并根据该评分值对搜索结果进行 排序, 其中, 私有算法可以是与指定的评分算法相同的算法, 也可以是不相 同的算法。 同时也可以根据预设的阈值, 对搜索结果进行过滤。
歩骤 2312, 接收成员搜索装置返回的个性化搜索结果及相应的评分 值。
歩骤 2314, 根据搜索结果的评分值和相关因素 (成员搜索装置级别和 价格因素等) 对搜索结果进行重评分。 其中, re-ranking包括: 根据提取的 兴趣模型计算搜索结果的评分值。 例如, 计算级别评分信息, 计算公式如 下:
P=rl*返回的评分值 +r2*级别因子
其中, P表示级别评分, rl表示返回的评分值占的权重, r2表示级别因 子占的权重, 所述返回的评分值为成员搜索装置返回的评分值, 所述级别因 子表示成员搜索装置级别。
计算综合评分信息, 计算公式如下:
R=P+r3*价格因素评分值
其中, R表示综合评分, r3表示价格因素评分权重, rl+r2+r3=l。
歩骤 2316, 根据 re-ranking的结果对搜索结果进行排序。 例如, 将相关 性高的搜索结果排在靠前的位置, 或将相关性高的搜索结果排在竞价排名的 搜索结果之后。 这样有利于用户能快速浏览到其需要搜索的相关性高的结
果, 在其他实施例中, 该排序还包括成员搜索装置的级别评分信息和相关因 素信息 (如价格因素信息等) 。
歩骤 2318, 将排序后的结果返回给应用服务器。
歩骤 2320, 应用服务器将搜索结果返回给搜索客户端。 歩骤 2318和歩 骤 2320中返回的搜索结果可以是经过过滤后的只包含相关性比较高的搜索 结果, 向用户提供部分搜索结果, 可以减少网络的流量, 减轻搜索客户端的 压力。
请结合参看图 24, 是一个搜索方法实施例的流程图。 该实施例中, 搜 索服务器向成员搜索装置指明统一的个性化评分算法。 该方法包括如下歩 骤:
歩骤 2402, 搜索客户端将搜索请求发送给应用服务器, 该搜索请求中 包含一个或多个搜索关键字, 该搜索关键字可以由用户以文本方式或语音方 式输入。 该搜索请求可以是移动终端向网络侧发出的一个信号。
歩骤 2404, 应用服务器从用户的个性化数据 (如用户的静态 profile和 用户的点击历史) 中提取用户兴趣模型, 或直接取出预先根据用户个性化数 据提取的兴趣模型。 兴趣模型可以为: 用户对 N个维度的评分值所组成的 兴趣模型向量, N>=2。
歩骤 2406, 应用服务器携带用户的兴趣模型, 将搜索请求发送给搜索 服务器。
歩骤 2408, 将所述搜索一个或多个搜索关键字和所述兴趣模型分发给 一个或多个成员搜索装置。 本实施例中, 携带用户兴趣模型将搜索请求分发 给成员搜索装置, 并指明采用统一的算法对搜索结果做个性化处理, 指明统 一的算法可以通过一个算法 ID指示。
歩骤 2410, 成员搜索装置完成搜索, 用指定的相同的个性化评分算法 计算搜索结果的评分值, 并根据该评分值对搜索结果进行排序。 同时也可以 根据预设的阈值, 对搜索结果进行过滤, 例如: 返回结果控制在 100个。
歩骤 2412, 搜索服务器接收成员搜索装置返回的个性化搜索结果及相 应的评分值。
歩骤 2414, 根据返回的评分值对搜索结果进行排序。 例如, 将相关性 高的搜索结果排在靠前的位置, 或将相关性高的搜索结果排在竞价排名的搜 索结果之后。 这样有利于用户能快速浏览到其需要搜索的相关性高的结果, 在其他实施例中, 该排序还包括成员搜索装置的级别评分信息和相关因素信 息 (如价格因素信息等) 。
歩骤 2416, 将排序后的结果返回给应用服务器。
歩骤 2418, 应用服务器将搜索结果返回给搜索客户端。 歩骤 2416和歩 骤 2418 中返回的搜索结果可以是经过过滤后的只包含相关性比较高的搜索 结果, 向用户提供部分搜索结果, 可以减少网络的流量, 减轻搜索客户端的 压力。
请结合参看图 25, 是一个搜索方法实施例的流程图。 该实施例中, 搜 索服务器只将搜索请求分发给成员搜索装置或成员搜索服务器。 该方法包括 如下歩骤:
歩骤 2502, 搜索客户端将搜索请求发送给应用服务器, 该搜索请求中 包含一个或多个搜索关键字, 该搜索关键字可以由用户以文本方式或语音方 式输入。 该搜索请求可以是移动终端向网络侧发出的一个信号。
歩骤 2504, 应用服务器从用户的个性化数据 (如用户的静态 profile和 用户的点击历史) 中提取用户兴趣模型, 或直接取出预先根据用户个性化数 据提取的兴趣模型。 兴趣模型可以为: 用户对 N个维度的评分值所组成的 兴趣模型向量, N>=2。
歩骤 2506, 应用服务器携带用户的兴趣模型, 将搜索请求发送给搜索 服务器。
歩骤 2508, 将所述搜索一个或多个搜索关键字分发给一个或多个成员 搜索装置或成员搜索服务器, 成员搜索服务器可以继续分发给下一级引擎, 不再赘述。
歩骤 2510, 成员搜索装置完成搜索, 或成员搜索服务器完成搜索。 歩骤 2512, 搜索服务器接收成员搜索装置或成员搜索服务器返回的搜 索结果。
歩骤 2514, 搜索服务器对搜索结果进行汇总, 根据用户的兴趣模型计 算搜索结果的评分值, 并根据该评分值对搜索结果进行排序。
歩骤 2516, 将排序后的结果返回给应用服务器。
歩骤 2518, 应用服务器将搜索结果返回给搜索客户端。 歩骤 2516和歩 骤 2518 中返回的搜索结果可以是经过过滤后的只包含相关性比较高的搜索 结果, 向用户提供部分搜索结果, 可以减少网络的流量, 减轻搜索客户端的 压力。
请结合参看图 26, 是一个搜索方法实施例的流程图。 该实施例中, 搜 索服务器提取用户的兴趣模型或取出预先存储的提取好的兴趣模型, 将搜索 请求和兴趣模型分发给成员搜索装置, 并指明统一的个性化评分算法。 该方 法包括:
歩骤 2602, 接收搜索客户端发送的搜索请求, 该搜索请求中包含一个 或多个搜索关键字, 该搜索关键字可以由用户以文本方式或语音方式输入。 该搜索请求可以是移动终端向网络侧发出的一个信号。
歩骤 2604, 根据用户的个性化数据提取用户的兴趣模型或取出预先存 储的提取好的兴趣模型。 本实施例中, 用户的个性化数据包括以下一种或多 种: 静态用户档案 User Profile 搜索历史、 位置信息或呈现信息。 兴趣模 型为: 用户对 N个维度的评分值所组成的兴趣模型向量, N>=2。
歩骤 2606, 将所述搜索一个或多个搜索关键字和所述兴趣模型分发给 一个或多个成员搜索装置或成员搜索服务器。 本实施例中, 携带用户兴趣模
型将搜索请求分发给成员搜索装置, 并指明采用统一的算法对搜索结果做个 性化处理, 指明统一的算法可以通过一个算法 ID指示。
歩骤 2608, 成员搜索装置完成搜索, 用指定的相同的个性化评分算法 计算搜索结果的评分值, 并根据该评分值对搜索结果进行排序。 同时也可以 根据预设的阈值, 对搜索结果进行过滤, 例如: 返回结果控制在 100个。
歩骤 2610, 搜索服务器接收成员搜索装置返回的个性化搜索结果及相 应的评分值。
歩骤 2612, 根据返回的评分值对搜索结果进行排序。 例如, 将相关性 高的搜索结果排在靠前的位置, 或将相关性高的搜索结果排在竞价排名的搜 索结果之后。 这样有利于用户能快速浏览到其需要搜索的相关性高的结果, 在其他实施例中, 该排序还包括成员搜索装置的级别评分信息和相关因素信 息 (如价格因素信息等) 。
歩骤 2614, 将最终搜索结果返回给搜索客户端。 返回的搜索结果可以 是经过过滤后的只包含相关性比较高的搜索结果, 向用户提供部分搜索结 果, 可以减少网络的流量, 减轻搜索客户端的压力。
请结合参看图 27, 是一个搜索方法实施例的流程图。 该实施例中, 搜 索服务器只将搜索请求分发给成员搜索装置或成员搜索服务器。 该方法包括 如下歩骤:
歩骤 2702, 接收搜索客户端发送的搜索请求, 该搜索请求中包含一个 或多个搜索关键字, 该搜索关键字可以由用户以文本方式或语音方式输入。 该搜索请求可以是移动终端向网络侧发出的一个信号。
歩骤 2704, 根据用户的个性化数据提取用户的兴趣模型或取出预先存 储的提取好的兴趣模型。 本实施例中, 用户的个性化数据包括以下一种或多 种: 静态用户档案 User Profile 搜索历史、 位置信息或呈现信息。 兴趣模 型为: 用户对 N个维度的评分值所组成的兴趣模型向量, N>=2。
歩骤 2706, 将所述搜索一个或多个搜索关键字分发给一个或多个成员 搜索装置或成员搜索服务器, 成员搜索服务器可以继续分发给下一级引擎, 不再赘述。
歩骤 2708, 成员搜索装置完成搜索, 和 /或成员搜索服务器完成搜索。 歩骤 2710, 搜索服务器接收成员搜索装置返回的个性化搜索结果。
歩骤 2712, 搜索服务器对搜索结果进行汇总, 根据用户的兴趣模型计 算搜索结果的评分值, 并根据该评分值对搜索结果进行排序。 例如, 将相关 性高的搜索结果排在靠前的位置, 或将相关性高的搜索结果排在竞价排名的 搜索结果之后。 这样有利于用户能快速浏览到其需要搜索的相关性高的结 果, 在其他实施例中, 该排序还包括成员搜索装置的级别评分信息和相关因 素信息 (如价格因素信息等) 。
歩骤 2714, 将最终搜索结果返回给搜索客户端。 返回的搜索结果可以 是经过过滤后的只包含相关性比较高的搜索结果, 向用户提供部分搜索结 果, 可以减少网络的流量, 减轻搜索客户端的压力。
为便于理解, 以下结合具体的应用实例来进行说明。
1.兴趣模型的定义
将用户的兴趣用 n个维度来表示如: 新闻、 体育、 娱乐、 财经、 科技、 房产、 游戏、 女性、 论坛、 天气、 商品、 家电、 音乐、 读书、 博客、 手机、 军事、 教育、 旅游、 彩信、 彩铃、 餐饮、 民航、 工业、 农业、 电脑、 地理 等。 用户对每个维度的兴趣的评分值所组成的一个向量 W(rl,r2,r3,…… ,m) 则为用户的兴趣模型。
2.搜索服务器从用户数据中提取兴趣模型
( 1 ) 用户的静态 profile所对应的兴趣模型 W1
Wl=(pl,p2,p3,…… ,ρη),其中 pi为静态 profile中类型属于第 i个兴趣 维度的所有词的词频之和。
(2) 用户的搜索点击历史所对应的兴趣模型 W2
W2=dl+d2+d3 + ··· ··· dm, 其中 di为用户某个点击文档所对应的兴 趣模型向量;
di =(tlAt3,…… .,tn), 当用户最新点击了这个文档, tj等于文档中类 型属于第 j个兴趣维度的所有词的词频之和。 如果用户对某个点击过的文档 进行评价, 如果评价好, di 向量乘以一个正的常数 C表示文档的重要性增加 di = c*di = (c*ti,c*t2,c*t3,…… ,c*tn), 如果评价不好, di向量乘以一个正的常 数 c 的 倒 数 表 示 文 档 的 重 要 性 减 小 di = l/c*di =
(l/c*ti,l/c*t2,l/c*t3, ,l/c*tn);
过了一段时间, tj 的值又自动减少一定的百分比, 表示随着时间的推移 其重要性减弱, 直到过了较长的时间 tj 的值减为零为止, 这时将 di从历史 记录中删除。 (例如, 每过一个月, tj的值减少 10% )
(3 ) 综合的兴趣模型 W=W1+W2
3.搜索服务器携带提取的兴趣模型数据向一个或多个成员搜索装置发起 个性化搜索请求, 并指明多个成员搜索装置采用某种指定的个性化算法对搜 索结果进行个性化的处理。
4.某个成员搜索装置用指定个性化算法进行个性化的搜索。
( 1 ) 成员搜索装置根据倒排索引检索出候选的结果文档。
(2 ) 成员搜索装置根据兴趣模型数据和指定的个性化算法对候选的结 果文档进行个性化相关性评分与排序。
算法 (a) : W=(rl,r2,r3,…… ,m)为元搜索引擎传过来的兴趣模型向量,
D=(tl,t2,t3,…… .,tn)为文档所对应的兴趣模型向量。
Ranking score = W *D =rl*tl+r2*t2+r3*t3+ ,+ rn*tn
或者
用算法 (b ) : W=(rl,r2,r3,…… ,m)为搜索服务器传过来的兴趣模型向
用通用的文档分类算法如 Knn, Cvm等对文档进行分类, 分类的结果文 档属于 C的类型, 用类型 C与兴趣模型各个维度所属的类型相匹配, 将兴 趣模型中与文档类型相匹配的某个维度 i对应的评分值 ri付给该文档。
Ranking score = ri.
(3 ) 成员搜索装置返回最相关 (评分值最大) 的 n个文档以及文档的个 性化相关性评分值。
5.搜索服务器将各个成员搜索装置返回的个性化搜索结果根据用同一种 算法算出来的相关性评分值的大小做一个综合的相关性排序, 并将最相关的 结果返回给搜索客户端。
本领域普通技术人员可以理解, 上述方法中的全部或部分歩骤可以通过 程序指令相关的硬件完成, 该程序可以使得一个或多个计算机处理器执行上 述方法, 并且, 该程序可以存储于计算机可读存储介质中, 该计算机可读存 储介质, 例如 ROM、 RAM或光盘等。
在本发明的实施例中通过根据用户的个性化数据提取的兴趣模型及所述 搜索请求进行搜索得到搜索结果, 使得用户得到搜索结果更加贴近用户需 求, 不同的用户可以得到不同的搜索结果, 使得搜索结果更加个性化, 有利 于搜索业务的推广使用。 此外, 通过成员搜索装置来完成个性化的评分, 使 得成员搜索装置能够返回最相关的搜索结果, 搜索服务器在进行综合搜索结 果时, 使搜索结果更加精确。 通过指明各成员搜索装置利用统一的算法来进 行个性化处理, 使得各成员搜索装置的返回的评分具有可比性, 大大减少了 网络的流量, 提高了个性化的效率。
综上所述, 以上仅为本发明的较佳实施例而已, 并非用于限定本发明的 保护范围。 凡在本发明的精神和原则之内, 所作的任何修改、 等同替换、 改 进等, 均应包含在本发明的保护范围之内。
Claims
1.一种搜索方法, 用于为搜索客户端用户提供个性化搜索, 该个性化搜 索可以根据所述搜索客户端用户的兴趣模型提供相关的搜索结果, 其特征在 于, 所述搜索方法包括如下歩骤:
接收搜索客户端发送的搜索请求, 该搜索请求中包含一个或多个搜索关 键字;
根据用户的个性化数据提取用户的兴趣模型;
将所述一个或多个搜索关键字和所述兴趣模型分发给一个或多个成员搜 索装置;
接收所述一个或多个成员搜索装置返回的搜索结果及根据所述兴趣模型 计算得到的搜索结果相应的评分值;
根据所述评分值对所述搜索结果进行排序得到最终搜索结果, 返回给所 述搜索客户端。
2.如权利要求 1所述的搜索方法, 其特征在于, 所述用户的个性化数据 包括以下一种或多种: 静态用户档案 User Profile 搜索历史、 位置信息或 呈现信息。
3.如权利要求 1所述的搜索方法, 其特征在于, 所述兴趣模型为: 用户 对 N个维度的评分值所组成的兴趣模型向量, N>=2。
4.如权利要求 3所述的搜索方法, 其特征在于, 所述兴趣模型向量为: 一个或多个静态兴趣模型与一个或多个动态兴趣模型的向量之和或向量加权 之和。
5.如权利要求 4所述的搜索方法, 其特征在于, 该方法还包括: 将所述 兴趣模型向量进行归一化处理。
7.如权利要求 6所述的搜索方法, 其特征在于, 所述 pi取值可根据历史 搜索时间的推移而发生变化, 或根据用户的评价发生变化。
8.如权利要求 1所述的搜索方法, 其特征在于, 该方法还包括: 将所述 搜索结果根据设定的阈值对搜索结果进行过滤, 向所述搜索客户端发送的最 终搜索结果为过滤后的最终搜索结果。
9.如权利要求 1所述的搜索方法, 其特征在于, 该方法还包括: 向所述 一个或多个成员搜索装置指定统一的评分算法, 并将算法标识 ID传递给所 述一个或多个成员搜索装置。
10.—种搜索方法, 其特征在于, 包括:
接收搜索客户端发送的搜索请求, 该搜索请求中包含一个或多个搜索关 键字及用户标识信息;
将所述一个或多个关键字及所述用户标识信息分发给一个或多个成员搜 索装置;
接收所述一个或多个成员搜索装置根据用户的个性化数据提取的兴趣模 型搜索到的搜索结果;
向所述搜索客户端发送搜索结果。
11.如权利要求 10所述的搜索方法, 其特征在于, 所述用户的个性化数 据包括以下一种或多种: 静态用户档案 User Profile 搜索历史、 位置信息 或呈现信息。
12.如权利要求 10所述的搜索方法, 其特征在于, 所述兴趣模型为: 用 户对 N个维度的评分值所组成的兴趣模型向量, N>=2。
13.如权利要求 12所述的搜索方法, 其特征在于, 所述兴趣模型向量 为: 一个或多个静态兴趣模型与一个或多个动态兴趣模型的向量之和或加权 向量之和。
15.如权利要求 14所述的搜索方法, 其特征在于, 所述 pi取值可根据历 史搜索时间的推移而发生变化, 或根据用户的评价发生变化。
16.如权利要求 10所述的搜索方法, 其特征在于, 该方法还包括根据所 述评分值对搜索结果进行过滤; 所述向发送给所述搜索客户端为发送过滤后 的搜索结果。
17.—种搜索方法, 其特征在于, 包括:
接收搜索请求, 该搜索请求中包含一个或多个关键字;
根据所述一个或多个关键字进行搜索;
根据用户的个性化数据提取的兴趣模型对搜索结果进行评分和排序; 返回所述排序后的搜索结果。
18.如权利要求 17所述的搜索方法, 其特征在于, 该方法还包括: 根据 所述兴趣模型计算搜索结果的评分值, 根据所述评分值对所述搜索结果进行 过滤, 所述返回所述排序后的搜索结果为向所述搜索客户端返回过滤后的搜 索结果。
19.如权利要求 18所述的搜索方法, 其特征在于, 所述根据所述兴趣模 型计算搜索结果的评分值包括: 根据兴趣模型向量与搜索结果中的文档提取 的向量进行点乘后得到所述评分值。
20.如权利要求 19所述搜索方法, 其特征在于, 该方法还包括将所述兴 趣模型向量或所述搜索结果中的文档的向量进行归一化处理。
21.—种搜索方法, 其特征在于, 该方法包括:
接收搜索客户端发送的搜索请求, 该搜索请求中包含一个或多个搜索关
将所述一个或多个搜索关键字分发给搜索装置;
接收搜索装置返回的搜索结果;
根据用户的个性化数据提取用户的兴趣模型, 并根据所述兴趣模型计算 搜索结果的评分值;
根据所述评分值对搜索结果进行排序;
向所述搜索客户端发送排序后的搜索结果。
22.如权利要求 21所述的搜索方法, 其特征在于, 所述用户的个性化数 据包括以下一种或多种: 静态用户档案 User Profile 搜索历史、 位置信息 或呈现信息。
23.如权利要求 21所述的搜索方法, 其特征在于, 该方法还包括: 根据 所述评分值对搜索结果进行过滤; 所述向所述搜索客户端发送相关性排序后 的搜索结果为发送过滤后的搜索结果。
24.—种计算机可读存储介质, 其特征在于, 包括计算机程序, 该计算 机程序运行时使得一个或多个处理器执行如下歩骤:
接收搜索客户端发送的搜索请求, 该搜索请求中包含一个或多个搜索关 键字;
根据用户的个性化数据提取用户的兴趣模型;
根据所述一个或多个搜索关键字进行搜索, 并根据所述兴趣模型对搜索 结果进行相关性排序;
向所述搜索客户端发送相关性排序后的搜索结果。
25.—种搜索装置, 其特征在于, 该装置包括:
用于接收搜索请求的工具;
用于根据用户的个性化数据提取用户的兴趣模型的工具;
用于将所述请求和所述兴趣模型分发给成员搜索装置的工具; 用于接收成员搜索装置返回的搜索结果及搜索结果评分值的工具; 用于根据所述评分值对搜索结果进行排序的工具;
用于将搜索结果返回给搜索客户端的工具。
26.—种搜索装置, 其特征在于, 该装置包括:
用于接收搜索客户端搜索请求的工具;
用于根据用户的个性化数据提取用户的兴趣模型的工具;
用于根据所述兴趣模型对所述搜索请求搜索得到的搜索结果进行评分排 序的工具;
用于向所述搜索客户端发送所述搜索结果的工具。
27.如权利要求 26所述的搜索装置, 其特征在于, 该装置还包括: 用于根据搜索结果及所述兴趣模型对所述搜索结果进行过滤的工具。
28.如权利要求 26所述的搜索装置, 其特征在于, 该装置还包括: 用于将所述搜索请求分发给其他搜索装置的工具。
29.如权利要求 26所述的搜索装置, 其特征在于, 该装置基于搜索*** 实施, 该***还包括: 一个或多个用于根据所述搜索请求向所述搜索装置返 回搜索结果的成员搜索装置。
30.—种搜索装置, 其特征在于, 该装置包括:
用于接收搜索请求的工具;
用于根据用户的个性化数据提取的兴趣模型及所述搜索请求进行搜索得 到搜索结果的工具;
用于返回所述搜索结果的工具。
31.如权利要求 30所述的搜索装置, 其特征在于, 该装置还包括: 用于 根据搜索结果及所述兴趣模型对所述搜索结果进行过滤的工具。
32.如权利要求 30所述的搜索装置, 其特征在于, 该装置基于搜索*** 实施, 该搜索***还包括: 用于将所述搜索请求分发给所述搜索装置的装 置。
33.—种搜索***, 其特征在于, 包括: 搜索服务器, 可以与一个或多 个成员搜索装置进行搜索通信; 所述搜索服务器, 用于接收搜索客户端的搜
索请求, 向所述一个或多个成员搜索装置分发所述搜索请求, 并接收所述一 个或多个成员搜索装置返回的搜索结果; 所述一个或多个成员搜索装置, 用 于根据所述搜索请求搜索得到所述搜索结果; 所述搜索服务器或所述一个或 多个成员搜索装置, 还可以被配置为根据用户数据提取的兴趣模型, 并根据 所述兴趣模型计算所述搜索结果的评分值并进行相关性排序。
34.如权利要求 33所述的搜索***, 其特征在于, 所述搜索服务器或所 述一个或多个成员搜索装置, 还可被配置为根据所述评分值对所述搜索结果 进行过滤。
35.如权利要求 33所述的搜索***, 其特征在于, 所述搜索服务器, 还 可以被配置为向所述一个或多个成员搜索装置指定统一的评分算法, 并分发 统一的评分算法标识 ID。
36.—种搜索***, 其特征在于, 该***包括: 搜索服务器, 可以与一 个或多个成员搜索装置进行搜索通信; 所述搜索服务器, 用于接收搜索客户 端的搜索请求, 根据用户的个性化数据提取用户的兴趣模型, 向所述一个或 多个成员搜索装置分发所述搜索请求和所述兴趣模型, 并接收所述一个或多 个成员搜索装置返回的搜索结果及搜索结果的评分值, 根据所述评分值对搜 索结果进行排序, 并向搜索客户端返回排序后的搜索结果; 所述一个或多个 成员搜索装置, 用于根据所述搜索请求搜索得到所述搜索结果; 并根据所述 兴趣模型计算所述搜索结果的所述评分值。
37.如权利要求 36所述的搜索***, 其特征在于, 所述搜索服务器还可 进一歩被配置为根据所述评分值及设定的阈值对搜索结果进行过滤, 向所述 搜索客户端返回过滤后的搜索结果。
38.—种搜索方法, 其特征在于, 包括:
接收搜索请求, 该搜索请求中包含一个或多个搜索关键字;
将所述一个或多个搜索关键字分发给搜索装置;
返回搜索结果后, 根据用户的个性化数据提取的用户兴趣模型计算的搜
索结果的评分值对搜索结果进行排序;
返回排序后的搜索结果。
39.如权利要求 38所述的搜索方法, 其特征在于, 所述接收搜索请求包 括接收搜索客户端的搜索请求, 或接收应用服务器传送的搜索请求。
40.如权利要求 38或 39所述的搜索方法, 其特征在于, 所述将所述一 个或多个搜索关键字分发给搜索装置包括: 将所述一个或多个关键字分发给 搜索服务器或分发给成员搜索装置或分发给成员搜索服务器。
41.如权利要求 38或 39所述的搜索方法, 其特征在于, 所述将所述一 个或多个搜索关键字分发给搜索装置包括: 携带用户兴趣模型将所述一个或 多个搜索关键字分发给成员搜索装置或成员搜索服务器或搜索服务器。
42.如权利要求 41所述的搜索方法, 其特征在于, 所述携带用户兴趣模 型将所述一个或多个搜索关键字分发给成员搜索装置或成员搜索服务器包 括: 指明采用统一算法对搜索结果进行个性化处理。
43.如权利要求 42所述的搜索方法, 其特征在于, 所述指明统一算法对 搜索结果进行个性化处理包括: 分发统一的评分算法标识 ID。
44.如权利要求 42所述的搜索方法, 其特征在于, 该方法还包括: 接收 所述搜索装置返回的根据所述统一算法对搜索结果进行计算得出个性化相关 性评分。
45.如权利要求 44所述的搜索方法, 其特征在于, 该方法还包括: 接收 所述搜索装置返回的根据所述个性化相关性评分对搜索结果进行排序的结 果。
46.如权利要求 44所述的搜索方法, 其特征在于, 该方法还包括: 根据 所述个性化相关性评分值和相关因素对搜索结果进行重新评分, 并根据所述 重新评分的评分值对搜索结果进行排序, 所述返回排序后的搜索结果为返回 根据所述重评分的评分值排序后的结果。
47.如权利要求 46所述的搜索方法, 其特征在于, 所述相关因素包括:
所述成员搜索装置或成员搜索服务器的级别信息, 和 /或价格排名信息。
48.如权利要求 41所述的搜索方法, 其特征在于, 该方法还包括: 接收 所述搜索装置返回的根据私有算法对搜索结果进行计算得出个性化相关性评 分。
49.如权利要求 48所述的搜索方法, 其特征在于, 该方法还包括: 接收 所述搜索装置返回的根据个性化相关性评分对搜索结果进行排序的结果。
50.如权利要求 48所述的搜索方法, 其特征在于, 该方法还包括: 根据 所述个性化相关性评分和相关因素对搜索结果进行重新评分, 根据所述重新 评分值对搜索结果进行排序。
51.如权利要求 50所述的搜索方法, 其特征在于, 所述相关因素包括成 员搜索装置级别信息, 所述根据所述成员搜索装置或成员搜索服务器的级别 信息进行重新评分包括: 计算级别评分信息, 计算公式如下:
P=rl*返回的评分值 +r2*级别因子
其中, P表示级别评分, rl表示返回的评分值占的权重, r2表示级别因 子占的权重, 所述返回的评分值为成员搜索装置返回的评分值, 所述级别因 子表示成员搜索装置级别。
52.如权利要求 51所述的搜索方法, 其特征在于, 所述根据所述成员搜 索装置或成员搜索服务器的级别信息进行重新评分包括: 计算综合评分信 息, 计算公式如下:
R=P+r3 *价格因素评分值
其中, R表示综合评分, r3表示价格因素评分权重, rl+r2+r3=l。
53.如权利要求 38所述的搜索方法, 其特征在于, 该方法还包括: 应用 服务器或搜索服务器或成员搜索服务器或成员搜索装置根据所述用户的个性 化数据提取所述用户的兴趣模型。
54.如权利要求 38所述的搜索方法, 其特征在于, 所述兴趣模型为: 用 户对 N个维度的评分值所组成的兴趣模型向量, N>=2。
55.如权利要求 38或 54所述的搜索方法, 其特征在于, 所述兴趣模型 向量为: 一个或多个静态兴趣模型与一个或多个动态兴趣模型的向量之和或 向量加权之和。
57.如权利要求 56所述的搜索方法, 其特征在于, 所述 pi取值可根据历 史搜索时间的推移而发生变化, 或根据用户的评价发生变化。
58.如权利要求 55所述的搜索方法, 其特征在于, 该方法还包括: 将所 述兴趣模型向量进行归一化处理。
59.如权利要求 38所述的搜索方法, 其特征在于, 所述兴趣模型包括用 户的个性化数据, 所述用户的个性化数据包括以下一种或多种: 静态用户档 案 User Profile 搜索历史、 位置信息或呈现信息。
60.—种搜索方法, 其特征在于, 包括:
接收搜索请求;
携带用户兴趣模型, 将所述搜索请求分发给搜索装置;
接收所述搜索装置返回的根据所述兴趣模型的个性化搜索结果, 并返回 所述个性化搜索结果。
61.如权利要求 60所述的搜索方法, 其特征在于, 所述搜索请求包括: 文本输入的搜索关键字, 和 /或语音识别出的搜索关键字。
62.如权利要求 60所述的搜索方法, 其特征在于, 该方法还包括: 接收 应用服务器根据用户个性化数据提取的所述用户兴趣模型或取出的预先存储 的提取好的所述用户兴趣模型。
63.如权利要求 60所述的搜索方法, 其特征在于, 所述携带用户兴趣模 型, 将所述搜索请求分发给搜索装置还包括: 指明采用统一算法对搜索结果
做个性化处理。
64.如权利要求 63所述搜索方法, 其特征在于, 该方法还包括: 接收所 述搜索装置返回的根据所述统一算法对搜索结果个性化评分的评分值。
65.如权利要求 64所述的搜索方法, 其特征在于, 该方法还包括: 接收 所述搜索装置返回的根据所述个性化评分值排序的搜索结果。
66.如权利要求 60所述的搜索方法, 其特征在于, 该方法还包括: 根据 所述个性化相关性评分和 /或相关因素对搜索结果进行重新评分, 根据所述 重新评分值对搜索结果进行排序。
67.如权利要求 66所述的搜索方法, 其特征在于, 所述相关因素包括: 成员搜索装置或成员搜索服务器的级别信息, 和 /或价格排名信息。
68.如权利要求 67所述的搜索方法, 其特征在于, 所述根据所述成员搜 索装置或成员搜索服务器的级别信息进行重新评分包括: 计算级别评分信 息, 计算公式如下:
P=rl*返回的评分值 +r2*级别因子
其中, P表示级别评分, rl表示返回的评分值占的权重, r2表示级别因 子占的权重, 所述返回的评分值为成员搜索装置返回的评分值, 所述级别因 子表示成员搜索装置级别。
69.如权利要求 68所述的搜索方法, 其特征在于, 所述根据所述成员搜 索装置或成员搜索服务器的级别信息进行重新评分包括: 计算综合评分信 息, 计算公式如下:
R=P+r3*价格因素评分值
其中, R表示综合评分, r3表示价格因素评分权重, rl+r2+r3=l。
70.如权利要求 60所述的搜索方法, 其特征在于, 该方法还包括: 接收 所述搜索装置返回的根据私有算法对搜索结果进行个性化相关性评分。
71.如权利要求 70所述的搜索方法, 其特征在于, 该方法还包括: 接收
所述搜索装置返回的根据个性化相关性评分对搜索结果进行排序的结果。
72.—种搜索方法, 其特征在于, 包括:
接收搜索客户端的搜索请求;
根据用户个性化数据提取用户的兴趣模型或取出预先存储的所述用户的 兴趣模型;
携带所述用户的兴趣模型, 向搜索服务器发送所述搜索请求; 接收所述搜索服务器返回的根据所述用户的兴趣模型个性化的搜索结 果;
向所述搜索客户端返回所述个性化处理后的搜索结果。
73.如权利要求 72所述的搜索方法, 其特征在于, 所述兴趣模型为: 用 户对 N个维度的评分值所组成的兴趣模型向量, N>=2。
74.如权利要求 72或 73所述的搜索方法, 其特征在于, 所述兴趣模型 向量为: 一个或多个静态兴趣模型与一个或多个动态兴趣模型的向量之和或 向量加权之和。
76.如权利要求 75所述的搜索方法, 其特征在于, 所述 pi取值可根据历 史搜索时间的推移而发生变化, 或根据用户的评价发生变化。
77.如权利要求 72所述的搜索方法, 其特征在于, 所述兴趣模型包括用 户的个性化数据, 所述用户的个性化数据包括以下一种或多种: 静态用户档 案 User Profile 搜索历史、 位置信息或呈现信息。
78.—种搜索方法, 其特征在于, 该方法包括:
接收搜索请求, 所述搜索请求携带用户的兴趣模型;
根据所述搜索请求搜索得到搜索结果;
根据所述用户的兴趣模型对所述搜索结果进行个性化处理; 返回所述个性化处理后的搜索结果。
79.如权利要求 78所述的搜索方法, 其特征在于, 该方法还包括: 接收 指定的算法, 并根据所述指定的算法和所述用户的兴趣模型计算搜索结果的 评分值。
80.如权利要求 79所述的搜索方法, 其特征在于, 该方法还包括: 根据 所述评分值对所述搜索结果进行排序, 所述返回所述个性化处理后的搜索结 果包括返回排序后的搜索结果。
81.如权利要求 79或 80所述的搜索方法, 其特征在于, 该方法还包 括: 返回所述评分值。
82.如权利要求 79所述的搜索方法, 其特征在于, 所述接收指定的算法 为接收指定的算法标识 ID, 根据所述算法 ID确定所述算法。
83.如权利要求 78所述的搜索方法, 其特征在于, 该方法还包括: 利用 私有算法计算所述搜索结果的评分值。
84.如权利要求 83所述的搜索方法, 其特征在于, 该方法还包括: 据所 述评分值对所述搜索结果进行排序, 所述返回所述个性化处理后的搜索结果 包括返回排序后的搜索结果。
85.如权利要求 84所述的搜索方法, 其特征在于, 该方法还包括: 返回 所述评分值。
86.—种搜索装置, 其特征在于, 包括:
用于接收搜索请求工具;
用于将所述搜索请求分发给搜索装置的工具;
用于接收搜索装置返回的搜索结果的工具;
用于根据用户的个性化数据提取的用户兴趣模型计算搜索结果的评分值 的工具;
用于根据所述评分值对搜索结果进行排序的工具; 和
用于返回排序后的搜索结果的工具。
87.如权利要求 86所述的搜索装置, 其特征在于, 该装置还包括: 用于根据所述个性化相关性评分和相关因素对搜索结果进行重新评分的 工具;
用于根据所述重新评分值对搜索结果进行排序的工具。
88.如权利要求 87所述的搜索装置, 其特征在于, 所述用于根据所述个 性化相关性评分和相关因素对搜索结果进行重新评分的工具还包括: 用于计 算级别评分信息的工具, 计算公式如下:
P=rl*返回的评分值 +r2*级别因子
其中, P表示级别评分, rl表示返回的评分值占的权重, r2表示级别因 子占的权重, 所述返回的评分值为成员搜索装置返回的评分值, 所述级别因 子表示成员搜索装置级别。
89.如权利要求 88所述的搜索装置, 其特征在于, 所述用于根据所述个 性化相关性评分和相关因素对搜索结果进行重新评分的工具还包括: 用于计 算综合评分信息的工具, 计算公式如下:
R=P+r3*价格因素评分值
其中, R表示综合评分, r3表示价格因素评分权重, rl+r2+r3=l。
90.—种搜索装置, 其特征在于, 包括;
用于接收搜索请求的工具;
用于携带用户兴趣模型, 将所述搜索请求分发给搜索装置的工具; 用于接收所述搜索装置返回的根据所述兴趣模型的个性化搜索结果的工 具; 和
用于返回所述个性化搜索结果的工具。
91.一种搜索装置, 其特征在于, 包括:
用于接收搜索客户端的搜索请求的工具;
用于根据用户个性化数据提取用户的兴趣模型或取出预先存储的所述用
户的兴趣模型的工具;
用于携带所述用户的兴趣模型, 向搜索服务器发送所述搜索请求的工 具;
用于接收所述搜索服务器返回的根据所述用户的兴趣模型个性化的搜索 结果;
用于向所述搜索客户端返回所述个性化处理后的搜索结果的工具。
92.—种搜索装置, 其特征在于, 包括:
搜索请求接收模块, 用于接收搜索请求和用户的兴趣模型;
搜索处理模块, 用于根据所述搜索请求搜索得到搜索结果;
搜索结果个性化处理模块, 用于根据所述用户的兴趣模型对所述搜索结 果进行个性化处理; 和
搜索结果返回模块, 用于返回所述个性化处理后的搜索结果。
93.—种搜索***, 其特征在于, 包括: 搜索服务器, 可以与一个或多 个成员搜索装置进行搜索通信;
所述搜索服务器, 用于接收搜索请求, 携带用户兴趣模型, 将所述搜索 请求分发给所述一个或多个成员搜索装置;
所述一个或多个成员搜索装置, 用于根据所述搜索请求搜索得到搜索结 果, 并根据所述用户的兴趣模型计算该搜索结果的评分值, 并返回给所述搜 索服务器;
所述搜索服务器接收所述一个或多个成员搜索装置返回的根据所述兴趣 模型的个性化搜索结果, 并返回所述个性化搜索结果。
94.如权利要求 93所述的搜索***, 其特征在于, 还包括: 应用服务 器, 用于根据用户个性化数据提取用户的兴趣模型或取出预先存储的所述用 户的兴趣模型, 并发送给所述搜索服务器。
95.如权利要求 94所述的搜索***, 其特征在于, 所述搜索服务器还用 于根据所述个性化相关性评分值和相关因素对搜索结果进行重新评分, 并根
据所述重新评分的评分值对搜索结果进行排序。
96.如权利要求 95所述的搜索***, 其特征在于, 所述相关因素包括: 所述成员搜索装置或成员搜索服务器的级别信息, 和 /或价格排名信息。
97.如权利要求 95所述的搜索***, 其特征在于, 所述重新评分包括: 计算级别评分信息, 计算公式如下:
P=rl*返回的评分值 +r2*级别因子
其中, P表示级别评分, rl表示返回的评分值占的权重, r2表示级别因 子占的权重, 所述返回的评分值为成员搜索装置返回的评分值, 所述级别因 子表示成员搜索装置级别。
98.如权利要求 97所述的搜索***, 其特征在于, 所述重新评分包括: 计算综合评分信息, 计算公式如下:
R=P+r3*价格因素评分值
其中, R表示综合评分, r3表示价格因素评分权重, rl+r2+r3=l。
99.如权利要求 93所述的搜索***, 其特征在于, 所述一个或多个成员 搜索装置包括成员搜索服务器, 所述成员搜索服务器还连接有下一级成员引
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP09809240A EP2336905A4 (en) | 2008-08-26 | 2009-08-26 | SEARCH PROCESS AND SYSTEM |
US13/034,071 US20110145234A1 (en) | 2008-08-26 | 2011-02-24 | Search method and system |
Applications Claiming Priority (10)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN200810142177 | 2008-08-26 | ||
CN200810142177.X | 2008-08-26 | ||
CN2008101870196A CN101661475B (zh) | 2008-08-26 | 2008-12-15 | 一种搜索方法和*** |
CN200810187004.X | 2008-12-15 | ||
CN200810187004.XA CN101661474B (zh) | 2008-08-26 | 2008-12-15 | 一种搜索方法和*** |
CN200810187049.7 | 2008-12-15 | ||
CN200810187020A CN101661476A (zh) | 2008-08-26 | 2008-12-15 | 一种搜索方法和*** |
CN200810187049A CN101661477A (zh) | 2008-08-26 | 2008-12-15 | 一种搜索方法和*** |
CN200810187020.9 | 2008-12-15 | ||
CN200810187019.6 | 2008-12-15 |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US13/034,071 Continuation US20110145234A1 (en) | 2008-08-26 | 2011-02-24 | Search method and system |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2010022655A1 true WO2010022655A1 (zh) | 2010-03-04 |
Family
ID=41789505
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2009/073534 WO2010022655A1 (zh) | 2008-08-26 | 2009-08-26 | 一种搜索方法和*** |
Country Status (4)
Country | Link |
---|---|
US (1) | US20110145234A1 (zh) |
EP (1) | EP2336905A4 (zh) |
CN (4) | CN101661474B (zh) |
WO (1) | WO2010022655A1 (zh) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102542012A (zh) * | 2010-12-15 | 2012-07-04 | 微软公司 | 分类搜索查询的结果 |
Families Citing this family (68)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9495460B2 (en) * | 2009-05-27 | 2016-11-15 | Microsoft Technology Licensing, Llc | Merging search results |
CN102256217A (zh) * | 2010-05-17 | 2011-11-23 | 华为技术有限公司 | 问题的提交、答案的返回、信息的推送方法及装置 |
CN101853308A (zh) * | 2010-06-11 | 2010-10-06 | 中兴通讯股份有限公司 | 一种个性化元搜索的方法及其应用终端 |
CN101894157A (zh) * | 2010-07-14 | 2010-11-24 | 中兴通讯股份有限公司 | 网页呈现方法和装置 |
CN101930475A (zh) * | 2010-09-14 | 2010-12-29 | 中兴通讯股份有限公司 | 网页显示方法及浏览器 |
US20120084247A1 (en) * | 2010-10-02 | 2012-04-05 | Microsoft Corporation | Affecting user experience based on assessed state |
WO2012045205A1 (zh) * | 2010-10-07 | 2012-04-12 | Liu Alion | 基于因特网的植物信息电子查询方法及其*** |
CN102467553A (zh) * | 2010-11-18 | 2012-05-23 | 中兴通讯股份有限公司 | 一种智能推送数据的方法和装置 |
CN102486781A (zh) * | 2010-12-03 | 2012-06-06 | 阿里巴巴集团控股有限公司 | 一种搜索排序方法和装置 |
US20120158765A1 (en) * | 2010-12-15 | 2012-06-21 | Microsoft Corporation | User Interface for Interactive Query Reformulation |
WO2012083560A1 (zh) * | 2010-12-24 | 2012-06-28 | 百度在线网络技术(北京)有限公司 | 一种用于基于强广义第二价格进行推广排序的方法和设备 |
CN102737027B (zh) * | 2011-04-01 | 2016-08-31 | 深圳市世纪光速信息技术有限公司 | 个性化搜索方法及*** |
CN102780573A (zh) * | 2011-05-11 | 2012-11-14 | 百度在线网络技术(北京)有限公司 | 一种基于网络流量进行实验的方法与设备 |
US9262513B2 (en) | 2011-06-24 | 2016-02-16 | Alibaba Group Holding Limited | Search method and apparatus |
CN105975632B (zh) * | 2011-06-24 | 2019-11-19 | 阿里巴巴集团控股有限公司 | 一种搜索方法、关系建立方法及相关设备 |
CN102915311B (zh) * | 2011-08-03 | 2016-04-27 | 腾讯科技(深圳)有限公司 | 搜索方法及*** |
US9177341B2 (en) | 2011-08-23 | 2015-11-03 | Amazon Technologies, Inc. | Determining search relevance from user feedback |
US8903817B1 (en) * | 2011-08-23 | 2014-12-02 | Amazon Technologies, Inc. | Determining search relevance from user feedback |
US20130086036A1 (en) * | 2011-09-01 | 2013-04-04 | John Rizzo | Dynamic Search Service |
CN102982035B (zh) | 2011-09-05 | 2015-10-07 | 腾讯科技(深圳)有限公司 | 一种社区用户的搜索排序方法及*** |
US9348479B2 (en) | 2011-12-08 | 2016-05-24 | Microsoft Technology Licensing, Llc | Sentiment aware user interface customization |
US9378290B2 (en) | 2011-12-20 | 2016-06-28 | Microsoft Technology Licensing, Llc | Scenario-adaptive input method editor |
CN103246670B (zh) | 2012-02-09 | 2016-02-17 | 深圳市腾讯计算机***有限公司 | 微博排序、搜索、展示方法和*** |
JP5905638B2 (ja) * | 2012-04-11 | 2016-04-20 | インテル コーポレイション | ユーザ・インターフェース・コンテンツ個人別最適化システム |
CN103425643B (zh) * | 2012-05-14 | 2018-07-31 | 深圳市世纪光速信息技术有限公司 | 一种相关搜索请求串推荐方法和*** |
CN104428734A (zh) | 2012-06-25 | 2015-03-18 | 微软公司 | 输入法编辑器应用平台 |
EP2682910A1 (en) | 2012-07-06 | 2014-01-08 | Axel Springer Digital TV Guide GmbH | Device and method for automatic filter adjustment |
US8935255B2 (en) * | 2012-07-27 | 2015-01-13 | Facebook, Inc. | Social static ranking for search |
CN103578469A (zh) * | 2012-08-08 | 2014-02-12 | 百度在线网络技术(北京)有限公司 | 一种展示语音识别结果的方法及装置 |
EP2891078A4 (en) | 2012-08-30 | 2016-03-23 | Microsoft Technology Licensing Llc | CHOICE OF CANDIDATE BASED ON CHARACTERISTICS |
CN102937974A (zh) * | 2012-10-17 | 2013-02-20 | 北京奇虎科技有限公司 | 一种搜索服务器及搜索方法 |
CN102937975B (zh) * | 2012-10-17 | 2016-06-29 | 奇飞翔艺(北京)软件有限公司 | 一种网页搜索设备和方法 |
CN102968454B (zh) * | 2012-10-26 | 2016-08-03 | 北京百度网讯科技有限公司 | 一种用于获取推广对象搜索结果的方法和设备 |
US8886644B1 (en) * | 2012-11-01 | 2014-11-11 | Google Inc. | User control of search filter bubble |
US20140136517A1 (en) * | 2012-11-10 | 2014-05-15 | Chian Chiu Li | Apparatus And Methods for Providing Search Results |
CN103838773A (zh) * | 2012-11-26 | 2014-06-04 | 百度在线网络技术(北京)有限公司 | 搜索结果用户评判方法及装置 |
CN103902449B (zh) * | 2012-12-28 | 2018-05-25 | 百度在线网络技术(北京)有限公司 | 一种用于生成搜索引擎相关性排序测试用例的方法与设备 |
US9576053B2 (en) | 2012-12-31 | 2017-02-21 | Charles J. Reed | Method and system for ranking content of objects for search results |
CN104052765A (zh) * | 2013-03-12 | 2014-09-17 | 蓝燕君 | 媒体信息传播方法及*** |
US20140330770A1 (en) * | 2013-05-03 | 2014-11-06 | Gface Gmbh | Context-aware implicit and explicit search |
EP3030982A4 (en) | 2013-08-09 | 2016-08-03 | Microsoft Technology Licensing Llc | INPUT PROCESSORS EDITOR WITH LANGUAGE SUPPORT |
CN104462146A (zh) * | 2013-09-24 | 2015-03-25 | 北京千橡网景科技发展有限公司 | 用于信息检索的方法和设备 |
CN103617241B (zh) * | 2013-11-26 | 2017-06-06 | 北京奇虎科技有限公司 | 搜索信息处理方法、浏览器终端与服务器 |
CN103646092B (zh) * | 2013-12-18 | 2017-07-04 | 孙燕群 | 基于用户参与的搜索引擎排序方法 |
CN105874531B (zh) * | 2014-01-06 | 2020-06-26 | 株式会社Ntt都科摩 | 终端设备、服务器设备以及计算机可读记录介质 |
RU2580516C2 (ru) | 2014-08-19 | 2016-04-10 | Общество С Ограниченной Ответственностью "Яндекс" | Способ формирования персонализированной модели ранжирования, способ формирования модели ранжирования, электронное устройство и сервер |
CN104239440B (zh) * | 2014-09-01 | 2017-08-25 | 百度在线网络技术(北京)有限公司 | 搜索结果的展现方法和装置 |
US9940409B2 (en) | 2014-10-31 | 2018-04-10 | Bank Of America Corporation | Contextual search tool |
US9922117B2 (en) | 2014-10-31 | 2018-03-20 | Bank Of America Corporation | Contextual search input from advisors |
US9785304B2 (en) | 2014-10-31 | 2017-10-10 | Bank Of America Corporation | Linking customer profiles with household profiles |
CN104462357B (zh) * | 2014-12-08 | 2017-11-17 | 百度在线网络技术(北京)有限公司 | 实现个性化搜索的方法和装置 |
CN106708817B (zh) * | 2015-07-17 | 2020-11-06 | 腾讯科技(深圳)有限公司 | 信息搜索方法及装置 |
CN105933413B (zh) * | 2016-04-21 | 2019-01-11 | 深圳大数点科技有限公司 | 一种基于用户声音交互的个性化实时内容推送*** |
CN106021602B (zh) * | 2016-06-15 | 2018-07-06 | 腾讯科技(深圳)有限公司 | 一种搜索结果排序的方法及装置 |
CN106294661B (zh) * | 2016-08-04 | 2019-09-20 | 百度在线网络技术(北京)有限公司 | 一种扩展搜索方法与装置 |
CN107729336B (zh) * | 2016-08-11 | 2021-07-27 | 阿里巴巴集团控股有限公司 | 数据处理方法、设备及*** |
US11256726B2 (en) * | 2016-12-19 | 2022-02-22 | Trellist Marketing and Technology | Interacting with objects based on geolocation |
CN106844744B (zh) * | 2017-02-15 | 2020-10-16 | 腾讯科技(深圳)有限公司 | 点击模型应用方法、装置及搜索*** |
CN107526807B (zh) * | 2017-08-22 | 2020-01-31 | 中国联合网络通信集团有限公司 | 信息推荐方法及装置 |
CN108415903B (zh) * | 2018-03-12 | 2021-09-07 | 武汉斗鱼网络科技有限公司 | 判断搜索意图识别有效性的评价方法、存储介质和设备 |
US11841854B2 (en) | 2018-07-24 | 2023-12-12 | MachEye, Inc. | Differentiation of search results for accurate query output |
US11651043B2 (en) | 2018-07-24 | 2023-05-16 | MachEye, Inc. | Leveraging analytics across disparate computing devices |
US11853107B2 (en) | 2018-07-24 | 2023-12-26 | MachEye, Inc. | Dynamic phase generation and resource load reduction for a query |
US11816436B2 (en) * | 2018-07-24 | 2023-11-14 | MachEye, Inc. | Automated summarization of extracted insight data |
US11341126B2 (en) | 2018-07-24 | 2022-05-24 | MachEye, Inc. | Modifying a scope of a canonical query |
CN111259129B (zh) * | 2020-01-19 | 2024-04-30 | 中国平安人寿保险股份有限公司 | 数据报表搜索方法、装置、电子设备及存储介质 |
CN111737542B (zh) * | 2020-07-17 | 2020-12-11 | 平安国际智慧城市科技股份有限公司 | 药品实体信息的搜索方法及存储介质 |
CN112434711B (zh) * | 2020-11-27 | 2023-10-13 | 杭州海康威视数字技术股份有限公司 | 一种数据管理方法、装置及电子设备 |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1811780A (zh) * | 2006-03-03 | 2006-08-02 | ***通信集团公司 | 基于个性化信息的搜索***及搜索方法 |
US20080114756A1 (en) * | 1999-12-28 | 2008-05-15 | Levino | Authomatic, personalized online information and product services |
Family Cites Families (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6493702B1 (en) * | 1999-05-05 | 2002-12-10 | Xerox Corporation | System and method for searching and recommending documents in a collection using share bookmarks |
US6327590B1 (en) * | 1999-05-05 | 2001-12-04 | Xerox Corporation | System and method for collaborative ranking of search results employing user and group profiles derived from document collection content analysis |
GB2395808A (en) * | 2002-11-27 | 2004-06-02 | Sony Uk Ltd | Information retrieval |
US20050071328A1 (en) * | 2003-09-30 | 2005-03-31 | Lawrence Stephen R. | Personalization of web search |
US20060064411A1 (en) * | 2004-09-22 | 2006-03-23 | William Gross | Search engine using user intent |
US20060074883A1 (en) * | 2004-10-05 | 2006-04-06 | Microsoft Corporation | Systems, methods, and interfaces for providing personalized search and information access |
US7693901B2 (en) * | 2005-08-10 | 2010-04-06 | Microsoft Corporation | Consumer-focused results ordering |
US7664746B2 (en) * | 2005-11-15 | 2010-02-16 | Microsoft Corporation | Personalized search and headlines |
KR101019548B1 (ko) * | 2006-08-31 | 2011-03-09 | 콸콤 인코포레이티드 | 사용자-기반 바이어스들을 사용하여 검색 결과들을 획득 또는 제공하는 방법 및 장치 |
US9798807B2 (en) * | 2008-05-06 | 2017-10-24 | Excalibur Ip, Llc | Algorithmically generated topic pages |
US8731995B2 (en) * | 2008-05-12 | 2014-05-20 | Microsoft Corporation | Ranking products by mining comparison sentiment |
-
2008
- 2008-12-15 CN CN200810187004.XA patent/CN101661474B/zh active Active
- 2008-12-15 CN CN200810187020A patent/CN101661476A/zh active Pending
- 2008-12-15 CN CN2008101870196A patent/CN101661475B/zh active Active
- 2008-12-15 CN CN200810187049A patent/CN101661477A/zh active Pending
-
2009
- 2009-08-26 EP EP09809240A patent/EP2336905A4/en not_active Withdrawn
- 2009-08-26 WO PCT/CN2009/073534 patent/WO2010022655A1/zh active Application Filing
-
2011
- 2011-02-24 US US13/034,071 patent/US20110145234A1/en not_active Abandoned
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080114756A1 (en) * | 1999-12-28 | 2008-05-15 | Levino | Authomatic, personalized online information and product services |
CN1811780A (zh) * | 2006-03-03 | 2006-08-02 | ***通信集团公司 | 基于个性化信息的搜索***及搜索方法 |
Non-Patent Citations (1)
Title |
---|
YANG XIANFENG, RESEARCH ON THE TECHNOLOGY OF PERSONALIZED INFORMATION RETRIEVAL IN SEARCH ENGINE, 31 December 2007 (2007-12-31), pages 17, XP008155125 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102542012A (zh) * | 2010-12-15 | 2012-07-04 | 微软公司 | 分类搜索查询的结果 |
CN102542012B (zh) * | 2010-12-15 | 2015-10-28 | 微软技术许可有限责任公司 | 分类搜索查询的结果 |
US9251185B2 (en) | 2010-12-15 | 2016-02-02 | Girish Kumar | Classifying results of search queries |
Also Published As
Publication number | Publication date |
---|---|
CN101661477A (zh) | 2010-03-03 |
CN101661475B (zh) | 2013-04-24 |
CN101661474B (zh) | 2012-07-04 |
EP2336905A4 (en) | 2011-11-02 |
CN101661475A (zh) | 2010-03-03 |
EP2336905A1 (en) | 2011-06-22 |
US20110145234A1 (en) | 2011-06-16 |
CN101661474A (zh) | 2010-03-03 |
CN101661476A (zh) | 2010-03-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2010022655A1 (zh) | 一种搜索方法和*** | |
Zamani et al. | Situational context for ranking in personal search | |
US10270791B1 (en) | Search entity transition matrix and applications of the transition matrix | |
US9589277B2 (en) | Search service advertisement selection | |
US9449271B2 (en) | Classifying resources using a deep network | |
CN103339623B (zh) | 涉及因特网搜索的方法和设备 | |
US8290927B2 (en) | Method and apparatus for rating user generated content in search results | |
WO2018209954A1 (zh) | 用于推送信息的方法和装置 | |
CN104428767B (zh) | 用于识别相关实体的方法、***和装置 | |
US7877404B2 (en) | Query classification based on query click logs | |
US8099412B2 (en) | Query statistics provider | |
US9934293B2 (en) | Generating search results | |
US20210365984A1 (en) | Image recognition based content item selection | |
CN110597962B (zh) | 搜索结果展示方法、装置、介质及电子设备 | |
WO2018040069A1 (zh) | 信息推荐***及方法 | |
JP2013506189A (ja) | クエリの一般属性に基づく情報の検索 | |
WO2011054245A1 (zh) | 移动搜索方法、装置和*** | |
CN106663100B (zh) | 多域查询补全 | |
US8224693B2 (en) | Advertisement selection based on key words | |
WO2010096986A1 (zh) | 移动搜索方法及装置 | |
WO2017136295A1 (en) | Adaptive seeded user labeling for identifying targeted content | |
CN110147494A (zh) | 信息搜索方法、装置,存储介质及电子设备 | |
CN104077327A (zh) | 核心词重要性识别方法和设备及搜索结果排序方法和设备 | |
US7716209B1 (en) | Automated advertisement publisher identification and selection | |
WO2020133187A1 (zh) | 一种针对内容的智能搜索推荐方法、存储介质及终端 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 09809240 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2009809240 Country of ref document: EP |