WO2017197806A1 - 基于人工智能提供智能服务的方法、智能服务***及智能终端 - Google Patents

基于人工智能提供智能服务的方法、智能服务***及智能终端 Download PDF

Info

Publication number
WO2017197806A1
WO2017197806A1 PCT/CN2016/097292 CN2016097292W WO2017197806A1 WO 2017197806 A1 WO2017197806 A1 WO 2017197806A1 CN 2016097292 W CN2016097292 W CN 2016097292W WO 2017197806 A1 WO2017197806 A1 WO 2017197806A1
Authority
WO
WIPO (PCT)
Prior art keywords
service
user
information
weight
search term
Prior art date
Application number
PCT/CN2016/097292
Other languages
English (en)
French (fr)
Inventor
董大祥
张军
于佃海
Original Assignee
北京百度网讯科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 北京百度网讯科技有限公司 filed Critical 北京百度网讯科技有限公司
Publication of WO2017197806A1 publication Critical patent/WO2017197806A1/zh
Priority to US16/193,454 priority Critical patent/US11651002B2/en

Links

Images

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
    • G06F16/951—Indexing; Web crawling techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/3331—Query processing
    • G06F16/3332—Query translation
    • G06F16/3334—Selection or weighting of terms from queries, including natural language queries
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24—Querying
    • G06F16/248—Presentation of query results
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24—Querying
    • G06F16/245—Query processing
    • G06F16/2457—Query processing with adaptation to user needs
    • G06F16/24578—Query processing with adaptation to user needs using ranking
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/332—Query formulation
    • G06F16/3325—Reformulation based on results of preceding query
    • G06F16/3326—Reformulation based on results of preceding query using relevance feedback from the user, e.g. relevance feedback on documents, documents sets, document terms or passages
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F7/00—Methods or arrangements for processing data by operating upon the order or content of the data handled
    • G06F7/06—Arrangements for sorting, selecting, merging, or comparing data on individual record carriers
    • G06F7/14—Merging, i.e. combining at least two sets of record carriers each arranged in the same ordered sequence to produce a single set having the same ordered sequence
    • G06F7/16—Combined merging and sorting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Definitions

  • the present application relates to the field of artificial intelligence, and more particularly, to a method for providing intelligent services based on artificial intelligence, an intelligent service system, and an intelligent terminal.
  • Artificial Intelligence abbreviated as AI in English. It is a new technical science that studies and develops theories, methods, techniques, and applications for simulating, extending, and extending human intelligence. Artificial intelligence is a branch of computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that responds in a manner similar to human intelligence. Research in this area includes robotics, speech recognition, image recognition, Natural language processing and expert systems.
  • Some intelligent service systems may have built-in feedback to collect user satisfaction and train and improve the system.
  • the collection of user feedback information and the training and improvement of the system by these intelligent service systems are all offline, and at least the problem lies in that the information collected offline is not real-time dynamic data.
  • the data learned and the learning are static, so that the first service result obtained by the user is also fixed, which makes the retrieval accuracy low, which greatly affects the user experience.
  • the embodiments of the present application are to provide a method, an intelligent service system, and an intelligent terminal for providing intelligent services based on artificial intelligence, so as to at least solve the problems existing in the prior art.
  • a method for providing an intelligent service based on artificial intelligence comprising: receiving a first service request of a user; determining a search term and a weight thereof for the first service request; according to the search term And weighting the first service result; and collecting feedback information of the user on the first service result, adjusting the search term for the first service request and/or its real-time according to the evaluation information in the feedback information Weights.
  • the above method can be implemented on a smart terminal or a remote server.
  • the evaluation information in the feedback information may be a score or positive and negative feedback.
  • the weight of the search term for the first service request is raised and saved.
  • the above method may further comprise: when the evaluation information in the feedback information is negative feedback or the score is lower than a predetermined threshold, according to the real-time adjusted search term and/or its weight, immediately provide a new first A service result.
  • the adjusting the search term and/or its weight in real time may include: determining that the first service is in the user when the evaluation information in the feedback information is negative feedback or the score is lower than a predetermined threshold Whether there is one or more other service requests other than the first service request within a predetermined time period before the request; and when the presence is determined, according to the negative feedback or the score, in combination with the one or more other service requests At least one of the search terms to adjust the search terms and/or their weights for the first service request in real time.
  • the determining the search term and the weight thereof may include: acquiring real-time status information and/or history information of the user; and content information according to the first service request and the acquired real-time status information of the user and/or Or historical information determines the terms and their weights.
  • the determining the search term and the weight thereof may include: performing an abstract spatial representation on the first service request according to the content information of the first service request and the acquired real-time status information and/or historical information of the user. Selecting a database retrieval manner for screening the first service result according to the first service request; and fusing an abstract spatial representation of the first service request with an abstract spatial representation of the database retrieval manner to obtain the Search terms and their weights.
  • the first service result is a list consisting of a plurality of first service result entries, and when the user's feedback on the at least one first service result entry in the list is less than a predetermined threshold or is In the case of negative feedback, the first service result entry in the list is reordered in real time according to the real-time adjusted search term and/or its weight, and the reordered list is immediately provided.
  • an artificial intelligence-based intelligent service system including: a request receiving module configured to receive a first service request of a user; and an enhanced learning module configured to be configured for the first service Determining a search term and its weight; a service processing module configured to provide a first service result based on the search term and its weight; and a feedback function module configured to collect feedback from the user on the first service result Information to obtain evaluation information in the feedback information; wherein the enhanced learning module is further configured to adjust a search term and/or a weight thereof for the first service request in real time according to the evaluation information in the feedback information.
  • the evaluation information in the feedback information may be a score or positive and negative feedback.
  • the enhanced learning module may be configured to: promote and save the search term for the first service request when the evaluation information in the feedback information is positive feedback or the score is greater than or equal to a predetermined threshold the weight of
  • the service processing module may be further configured to: according to the real-time adjusted search term and/or its weight when the evaluation information in the feedback information is negative feedback or the score is lower than a predetermined threshold , immediately provide a new first service result.
  • the enhanced learning module may be configured to: when the evaluation information in the feedback information Determining whether there is one or more other service requests other than the first service request within a predetermined time period before the user first service request when negative feedback or the score is below a predetermined threshold; and when it is determined to exist Retrieving the search term and/or its weight for the first service request in real time based on the negative feedback or score, in conjunction with the search term of at least one of the one or more other service requests.
  • the enhanced learning module may be further configured to: acquire real-time status information and/or history information of the user; and content information according to the first service request and the acquired real-time status information of the user and/or Or historical information determines the terms and their weights.
  • the enhanced learning module may be configured to perform a text abstract space representation on the first service request according to the content information of the first service request and the acquired real-time status information and/or historical information of the user. Selecting a database retrieval manner for screening the first service result according to the first service request; fusing the text abstract space representation with the abstract spatial representation of the database retrieval manner to obtain the search term and its weight .
  • said first service result may be a list of a plurality of first service result entries
  • said service processing module may be configured to: when said user is at least one of said first service result entries in said list
  • the score of the feedback is lower than the predetermined threshold or is negative feedback
  • the first service result item in the list is reordered in real time according to the real-time adjusted search term and/or its weight, and the reordered list is immediately provided.
  • an intelligent terminal based on artificial intelligence including an interface device that interacts with a user, and a compiling device connected to the interface device, the compiling device including: a memory; and the memory a processor coupled to receive a first service request of the user; determining a search term and its weight for the first service request; providing a first service result based on the search term and its weight; and collecting The feedback information of the first service result by the user adjusts the search term and/or its weight for the first service request in real time according to the evaluation information in the feedback information.
  • the method for providing an intelligent service based on artificial intelligence can collect feedback information of a user for a service result in real time (for example, including positive feedback information such as “like” and “dislike” This kind of negative feedback information), and adjust the system according to the feedback information in real time.
  • the system parameters get new search terms, new weights for the search terms, or both. This enables real-time training and improvement of the system, thereby improving retrieval accuracy.
  • FIG. 1 illustrates a flow diagram of a method 100 of providing intelligent services based on artificial intelligence, in accordance with one embodiment of the present application.
  • Figure 2 shows a schematic diagram of a specific implementation of determining a search term and its weight.
  • FIG. 3 illustrates a flow diagram of a method 300 of providing intelligent services based on artificial intelligence in accordance with another embodiment of the present application.
  • 4a-4c are diagrams showing an application example of a method of providing an intelligent service based on artificial intelligence according to another embodiment of the present application.
  • FIG. 5 is a schematic structural diagram of a physical architecture 500 of an artificial intelligence-based intelligent service system that implements an embodiment of the present application.
  • FIG. 6 is a block diagram of an artificial intelligence based intelligent terminal implementing an embodiment of the present application.
  • FIG. 7 is an interaction flow diagram for implementing the method embodiment shown in FIG.
  • FIG. 8 is a schematic structural diagram of an artificial intelligence-based intelligent service system according to an embodiment of the present application.
  • FIG. 9 is a block diagram showing the structure of a computer system suitable for implementing the terminal device or server of the embodiment of the present application.
  • the intelligent service system interacts with the user (and preferably with the server database or the smart terminal used by the user), parsing the content of the first service request entered by the user (and preferably, Combine the user's real-time status data and/or historical data, for example, the user's current status, the user's current geographic location, the user's gender, the user's age, the user's local time, the user's behavior log, etc.) to retrieve Words; collect feedback information of the user for the first service result in real time (for example, including positive feedback information such as "like” and negative feedback information such as "dislike"), and adjust the system parameters in real time according to the feedback information to obtain a new search.
  • Words new weights of terms, or both. This enables real-time training and improvement of the system, thereby improving retrieval accuracy.
  • the system may immediately provide a new first service result based on the new search term and/or its weight. In this way, when the user is dissatisfied with the retrieval result, the system can timely collect effective feedback information that is not satisfactory for the retrieval result, thereby immediately optimizing the first service result to improve the user experience.
  • FIG. 1 illustrates a flow diagram of a method 100 of providing intelligent services based on artificial intelligence, in accordance with one embodiment of the present application.
  • the method is implemented by an artificial intelligence based intelligent service system, which will be specifically described below.
  • all steps of the method can be implemented in a smart terminal such as a PC, laptop, smartphone, and tablet (eg, by an application App pre-installed therein).
  • all or part of the steps may also be implemented by a remote computer device (eg, a remote server) communicatively coupled to the smart terminal, while the search page displayed in the smart terminal or the search page displayed in the web browser may only As an interface or implementation part of the steps to interact with the user.
  • a remote computer device eg, a remote server
  • the method 100 includes the following steps:
  • a first service request from the user is received.
  • the first service request can be at least one of a text request input, a voice request input, and a picture request input.
  • the smart service system can also provide a list of services for the user to select, in which case the first service request can appear as a gesture of a certain service type in the list of services from the user's perspective.
  • the first service request may be presented as a further request for a service result provided by the previous system, for example, a gesture of clicking a link to "View Details" in the service result. This step is actually a process in which the user asks a question, for example, booking a plane ticket from Beijing to Shenzhen.
  • a search term and its weight are determined for the first service request of the received user.
  • the search terms mentioned herein cover the meaning different from the search terms in the prior art.
  • the search term is formed by dividing the input Chinese text according to a predetermined Chinese word segmentation algorithm, that is, the search term is extracted from the first service request.
  • Content information In the present application, the search term may not be limited to the content information of the first service request.
  • at least one of the user's real-time status information and historical information may be obtained prior to determining the search term and its weight.
  • the real-time status information of the user may include the geographical location where the user is currently located, the gender of the user, the age of the user, the local time of the user's location, and the like.
  • the intelligent service system can collect real-time status information of the user from the server database and/or the smart terminal currently used by the user, so as to more accurately predict the current needs of the user.
  • the user's history information may include the page that the user has browsed, the length of the page, the browsing time, the query log of the user, and the like.
  • the intelligent service system can collect historical information of the user from the smart terminal currently used by the user, so as to more accurately predict the user's interest preference. Thereby, the intelligent service system can determine the search term and its weight according to the content information of the first service request and the acquired real-time status information and/or historical information of the user. The specific implementation of determining the terms and their weights will be described in detail below.
  • a first service result is provided based on the search term and its weight.
  • the intelligent service system can use the search term and its weight to retrieve or perform corresponding processing to return the first service result.
  • the first service result has a different manifestation for different types of first service requests. For example, if the user asks "Which restaurants are nearby”, the first service result appears as a list of a plurality of restaurants near the geographic location of the user; if the user asks "How is the weather today", the first service result appears as the same day user Weather information for the location.
  • step 108 feedback information of the user on the first service result is collected, and the search term and/or its weight for the first service request is adjusted in real time according to the evaluation information in the feedback information.
  • User feedback on service results can be quantified by component values.
  • the feedback score can be the user to a certain
  • the scoring of the service result may also be a scoring of the behavior of the user for a certain service result according to a predetermined algorithm by the intelligent service system.
  • feedback can be simply divided into positive feedback and negative feedback.
  • positive feedback satisfactory for the search results
  • the feedback can be expressed as praise, voice (for example, the search results are too useful!), click on the search results (even if there is no order), click on the search results and place orders, etc. .
  • the feedback can be expressed as not operating the search results for a long time, closing a search result, closing the search service, closing the entire APP, and voice (for example, the search result is bad) !), conversion topics, etc.
  • User feedback on service results is retained in the artificial intelligence-based intelligent service system.
  • the intelligent service system can increase the weight of the search term.
  • the intelligent service system can reduce the weight of the current search term and enable the new search term.
  • the intelligent service system will continuously filter out the results that the user may like, and filter out the results of low quality. In this way, the intelligent service system can continuously improve the quality of service, thereby improving user satisfaction and increasing system stickiness.
  • the intelligent service system may determine the search term and its weight according to the input information of the service request and the acquired real-time status information and/or history information of the user. .
  • the intelligent service system needs to perform an abstract spatial representation 204 of the keywords extracted from the content information of the service request and the acquired real-time status information and/or historical information of the user when determining the search term and its weight.
  • the neural network can be used to represent the abstract space. The above keywords are transformed into the form of vectors after being represented by the abstract space.
  • a database retrieval mode 206 for screening service results can be selected based on the user's service request.
  • the retrieval method is a description of the way to filter the results in the database.
  • the system can set multiple retrieval methods in advance according to needs.
  • the abstract spatial representation 204 of the service request i.e., the input vector
  • the abstract space 208 representation of the database retrieval mode to obtain the search terms 212 required for the system retrieval, each of which may have a predetermined weight.
  • the system can adjust the search term and/or drop the search term according to the feedback. Right / promotion.
  • the feedback may be merged with an abstract spatial representation of the service request and an abstract spatial representation of the database retrieval method to obtain a real-time adjusted search term and/or its weight.
  • the system may update the parameters in the entire system by using a method of random gradient descent so that the feedback given by the user is the most positive.
  • FIG. 3 illustrates a flow diagram of a method 300 of providing intelligent services based on artificial intelligence in accordance with another embodiment of the present application.
  • the same steps as those of the method 100 shown in FIG. 1 are given the same reference numerals and will not be described again. Only the steps different from the method 100 are described below.
  • step 110 when the evaluation information in the feedback information is negative feedback or the score is lower than a predetermined threshold, the new first service result is immediately provided according to the real-time adjusted search term and/or its weight.
  • the intelligent service system will provide new service results.
  • the user may continue to provide feedback on new service results as the user receives a new service result.
  • the intelligent service system continuously adjusts the search terms and/or weights by continuously collecting user feedback, and constitutes a closed-loop training learning and continuous optimization process, which can continuously improve the quality of the returned service results, so as to provide the most users through training and learning. Good answer.
  • the intelligent service system greatly improves the probability of providing satisfactory service results in a short period of time, thereby greatly improving the user experience.
  • the intelligent service system determines the corresponding search term and its weight and Provide the corresponding service results.
  • the search term and/or its weight for the corresponding service request is adjusted in real time in conjunction with the search term for the at least one service request prior to the corresponding service request .
  • the intelligent service system can optimize the search terms and their weights for the user's current service request by combining the context of the user's dialogue with the system through training and learning.
  • the optimized search term may include keywords that are implicit but not present in the input information of the user's current service request.
  • the system can record the received user's service request and feedback as historical information for the user for use in future smart services. Or, the system can give the best of the user
  • the personalized neural network model is stored in the user's smart terminal or remote server to provide personalized intelligent services for the user in the future.
  • the system can collect service requests and feedback from multiple users, and update the database retrieval method in real time according to the collected service requests and feedback of multiple users.
  • the intelligent service system can be optimized overall as the number of uses increases, thereby continuously improving the user experience.
  • FIG. 4 is a diagram showing an application example of a method of providing an intelligent service based on artificial intelligence according to another embodiment of the present invention. As shown in Figures 4a and 4b, the user and the system make the following conversation through the interface of the App installed on the smart terminal:
  • A According to the dialogue process between the user and the system, extract the current key features, and use the existing neural network model of enhanced learning to score the real-time retrieval weights of the keywords. For example, keywords and their search weights: PM: 0.1, Shenzhen: 0.1, business trip: 0.3, ticket: 0.1, over: 0.01, weather: 0.5, location of the user (Beijing): 0.6.
  • the system After receiving negative feedback, the system updates the neural network model and re-scores the search terms. It should be noted that the search terms may be different each time, and the weights may be different, but the model is more inclined to predict context-related words as search terms.
  • the initial search term weight is higher.
  • the search terms, the weather, the location of the user (Beijing) will be strongly suppressed, and the weight of the search terms with lower weights will be less suppressed.
  • the enhanced learning model is used to predict the search term and its weight.
  • the search term and its weight are changed to: afternoon: 0.09, Shenzhen: 0.09, business trip: 0.2, ticket: 0.09, over: 0.009, weather: 0.1, User location (Beijing): 0.12.
  • the input of the whole reinforcement learning neural network model is: the key features of the user's state, such as the keyword of the user's dialogue with the system, the time and place at that time. In this example, afternoon, Shenzhen, please, help me, etc., can be input as features.
  • the output of the neural network model for enhanced learning is (before update): PM: 0.1, Shenzhen: 0.1, business trip: 0.3, ticket: 0.1, over: 0.01, weather: 0.5, location of the user (Beijing): 0.6.
  • Inputs and outputs are generally associated, but not necessarily identical. Weights are also given through the entire neural network.
  • the weight of the neural network to obtain the positive feedback of the set of predictive words is enhanced, thereby enhancing the effect of the next prediction.
  • FIG. 5 is a schematic structural diagram of a physical architecture of an artificial intelligence-based intelligent service system implementing an embodiment of the present application.
  • the system 500 includes a smart terminal 510 and one or more servers 530 that are connected by a network 520.
  • the smart terminal 510 is installed with a smart service App or a web browser.
  • the user accesses the smart service request interface by using the app or the web browser to make a service request to the system, receive the service result, and provide feedback on the service result.
  • the smart terminal 510 may be a smartphone, a PC, a tablet, a notebook, an intelligent robot, or the like.
  • FIG. 6 is a block diagram of an artificial intelligence based intelligent terminal 510 that implements an embodiment of the present invention.
  • the terminal 510 includes an interface device that interacts with the user, a compiling device that is coupled to the interface device, and a networking module 630 that is coupled to the compiling device.
  • the interface device that interacts with the user may be a touch screen 640, an audio output device 650 (including a speaker, a headset, etc.), a microphone 660; the compiling device may be a processor 610, a memory 620.
  • the processor 610 is configured to perform all or part of the steps of the method of the above described embodiments of the invention in conjunction with other elements.
  • the networking module 630 is configured to enable the terminal 510 and the server 530 Communication is between, for example, downloading service results from server 530, sending service requests to the server, and the like.
  • the memory 620 is configured to store information (eg, text, voice, pictures, etc.) of the service results downloaded from the server.
  • the touch screen 640 is configured to receive text input by the user, identify the user's gestures, and display the user's service request, system provided service results, and other related information.
  • the audio output device 650 is configured to play service results and system prompt information.
  • the microphone 660 is configured to collect voice information of the user.
  • the interaction flow for implementing the embodiment will be described below in conjunction with the method embodiment shown in FIG. Figure 7 is an interactive flow diagram for implementing this embodiment.
  • the interaction process involves the user 1, the smart terminal 510-1 used by the user 1, the smart terminal 510-2 used by the user 2, the user 2, the server 530, and other users other than the user 1, the user 2, and the smart terminal used by the user (not shown).
  • the functions implemented by the smart terminal 510-1 and the smart terminal 510-2 are distinguished here. It should be understood that these functions can be implemented on a smart terminal.
  • the interaction process includes the following steps:
  • Step 701 The user 1 starts the App by performing a tap gesture on the touch screen of the terminal 510-1.
  • Step 702 The terminal 510-1 starts an App.
  • Step 703 The terminal 510-1 displays a common service type list and a service request input interface.
  • Step 704 The user 1 selects a service type by performing a tap gesture on the touch screen of the terminal 510-1.
  • Step 705 The terminal 510-1 sends a service request corresponding to the selected service type to the server 530.
  • Step 706 The server 530 determines the search term and its weight by using a neural network model of the system
  • Step 707 The server 530 returns a service result to the terminal 510-1 according to the search term and its weight;
  • Step 708 The terminal 510-1 displays the service result.
  • Step 709 The user 1 performs a gesture of horizontally swiping on the touch screen of the terminal 510-1 to indicate that the service result is not satisfactory;
  • Step 710 The terminal 510-1 sends a negative feedback of the user 1 to the service result to the server 530.
  • Step 711 The server 530 updates the neural network model for the user 1 by reducing the weight of the current search term and replacing part of the current search term with the new search term.
  • Step 712 The server 530 returns a new service result to the terminal 510-1 according to the new search term and its weight.
  • Step 713 The terminal 510-1 displays a new service result.
  • Step 714 User 1 likes the new service result to indicate that the service result is satisfactory
  • Step 715 The terminal 510-1 sends a positive feedback of the user 1 to the new service result to the server 530.
  • Step 716 The server 530 increases the weight of the current search term.
  • Step 717 user 1 closes the App
  • Step 718 repeat steps 701-705;
  • Step 719 The server 530 returns the service result by using the neural network model after the previous update
  • Step 720 The terminal 510-1 displays the service result.
  • Step 721 The server 530 receives similar feedback from multiple users for the same service request, and updates the neural network model of the system;
  • Step 722 The user 2 launches the App by performing a tap gesture on the touch screen of the terminal 510-2.
  • Step 724 The terminal 510-2 displays a common service type list and a service request input interface.
  • Step 725 The user 2 selects the service type described above by performing a tap gesture on the touch screen of the terminal 510-2.
  • Step 726 The terminal 510-2 sends a service request corresponding to the selected service type to the server 530.
  • Step 727 The server 530 determines the search term and its weight by using the neural network model of the system updated in step 721.
  • Step 728 The server 530 returns the service result to the terminal 510-2 according to the search term and its weight;
  • Step 729 The terminal 510-2 displays the service result.
  • the present application provides an embodiment of an artificial intelligence based intelligent service system.
  • the description of the following system embodiments is similar to the above description of the method, and the beneficial effects of the same method are described. Said, do not repeat.
  • the modules included in the intelligent service system described below may all be implemented in the smart terminal, or may be implemented in the remote server, or may be implemented in the smart terminal, and the rest in the remote server. achieve.
  • the intelligent service system based on artificial intelligence includes:
  • the request receiving module 802 is configured to receive a first service request of the user
  • the enhanced learning module 804 is configured to determine a search term and its weight for the first service request
  • a service processing module 806 configured to provide a first service result based on the search term and its weight
  • the feedback function module 808 is configured to collect feedback information of the user on the first service result to obtain evaluation information in the feedback information.
  • the enhanced learning module 804 can be further configured to adjust the search terms and/or their weights for the first service request in real time based on the rating information in the feedback information.
  • the evaluation information in the feedback information may be a score or positive or negative feedback.
  • the service processing module may be further configured to provide a new one immediately based on the real-time adjusted search term and/or its weight when the evaluation information in the feedback information is negative feedback or the score is below a predetermined threshold. A service result.
  • the service processing module may promote and save the weight of the search term for the first service request.
  • the enhanced learning module may determine whether there is a first service request within a predetermined time period before the first service request One or more other service requests, when determined to exist, adjust the request for the first service in real time based on the negative feedback or the score, in conjunction with the search term of at least one of the one or more other service requests Search for words and / or their weights.
  • the enhanced learning module may be further configured to obtain real-time status information and/or historical information of the user, and according to the content information of the first service request and the acquired real-time user
  • the status information and/or history information determines the terms and their weights.
  • the enhanced learning module is configured to perform a textual abstract spatial representation of the first service request according to the content information of the first service request and the acquired real-time status information and/or historical information of the user; Selecting a database retrieval method for screening the first service result; fusing the text abstract space representation with an abstract spatial representation of the database retrieval method to obtain the search term and its weight.
  • the first service result may be a list of a plurality of first service result entries
  • the service processing module may be configured to when the user is at least one first service result in the list
  • the score of the feedback of the item is lower than a predetermined threshold or is negative feedback
  • the first service result item in the list is reordered in real time according to the real-time adjusted search term and/or its weight, and the reordering is provided immediately. list of.
  • the intelligent service system can further include a history information recording module (not shown) configured to record the first service request and the feedback of the user as historical information of the user.
  • a history information recording module (not shown) configured to record the first service request and the feedback of the user as historical information of the user.
  • the intelligent service system can further include: an update module (not shown) configured to collect first service requests and feedback of the plurality of users, and according to the first service request of the plurality of collected users and Feedback, update the database retrieval method.
  • an update module (not shown) configured to collect first service requests and feedback of the plurality of users, and according to the first service request of the plurality of collected users and Feedback, update the database retrieval method.
  • FIG. 9 is a block diagram showing the structure of a computer system suitable for implementing the terminal device or server of the embodiment of the present application.
  • computer system 900 includes a central processing unit (CPU) 901 that can be loaded into a program in random access memory (RAM) 903 according to a program stored in read only memory (ROM) 902 or from storage portion 908. And perform various appropriate actions and processes.
  • RAM random access memory
  • ROM read only memory
  • various programs and data required for the operation of the system 900 are also stored.
  • the CPU 901, the ROM 902, and the RAM 903 are connected to each other through a bus 904.
  • An input/output (I/O) interface 905 is also coupled to bus 904.
  • the following components are connected to the I/O interface 905: an input portion 906 including a keyboard, a mouse, etc.; an output portion 907 including, for example, a cathode ray tube (CRT), a liquid crystal display (LCD), and the like, and a storage portion 908 including a hard disk or the like. And include network interfaces such as LAN cards, modems, etc.
  • the communication portion 909 of the card The communication section 909 performs communication processing via a network such as the Internet.
  • Driver 910 is also connected to I/O interface 905 as needed.
  • a removable medium 911 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory or the like is mounted on the drive 910 as needed so that a computer program read therefrom is installed into the storage portion 908 as needed.
  • an embodiment of the present disclosure includes a computer program product comprising a computer program tangibly embodied on a machine readable medium, the computer program comprising program code for executing the method illustrated in the flowchart.
  • the computer program can be downloaded and installed from the network via the communication portion 909, and/or installed from the removable medium 911.
  • each block of the flowchart or block diagrams can represent a module, a program segment, or a portion of code that includes one or more logic for implementing the specified.
  • Functional executable instructions can also occur in a different order than that illustrated in the drawings. For example, two successively represented blocks may in fact be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending upon the functionality involved.
  • each block of the block diagrams and/or flowcharts, and combinations of blocks in the block diagrams and/or flowcharts can be implemented in a dedicated hardware-based system that performs the specified function or operation. Or it can be implemented by a combination of dedicated hardware and computer instructions.
  • the disclosed apparatus and method can be implemented in other ways.
  • the device embodiments described above are merely illustrative.
  • the division of the unit is only a logical function division.
  • there may be another division manner such as: multiple units or components may be combined, or Can be integrated into another system, or some features can be ignored or not executed.
  • the coupling, or direct coupling, or communication connection of the components shown or discussed may be indirect coupling or communication connection through some interfaces, devices or units, and may be electrical, mechanical or other forms. of.
  • the above described units as separate components may or may not be physically separated as The components displayed by the unit may or may not be physical units, that is, may be located in one place or distributed to multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the solution in this embodiment. .
  • each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may be separately used as one unit, or two or more units may be integrated into one unit;
  • the unit can be implemented in the form of hardware or in the form of hardware plus software functional units.
  • the foregoing program may be stored in a computer readable storage medium, and the program is executed when executed.
  • the foregoing storage device includes the following steps: the foregoing storage medium includes: a mobile storage device, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.
  • ROM read-only memory
  • RAM random access memory
  • magnetic disk or an optical disk.
  • optical disk A medium that can store program code.
  • the above-described integrated unit of the present invention may be stored in a computer readable storage medium if it is implemented in the form of a software function module and sold or used as a standalone product.
  • the technical solution of the embodiments of the present invention may be embodied in the form of a software product in essence or in the form of a software product stored in a storage medium, including a plurality of instructions.
  • a computer device (which may be a personal computer, server, or network device, etc.) is caused to perform all or part of the methods described in various embodiments of the present invention.
  • the foregoing storage medium includes various media that can store program codes, such as a mobile storage device, a ROM, a RAM, a magnetic disk, or an optical disk.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

一种基于人工智能提供智能服务的方法、智能服务***及智能终端。该方法包括：收到用户的第一服务请求（102）；针对所述第一服务请求确定检索词及其权重（104）；根据所述检索词及其权重提供第一服务结果（106）；以及收集所述用户对所述第一服务结果的反馈信息，根据所述反馈信息中的评价信息实时调整所述针对第一服务请求的检索词和/或其权重（108）。

Description

基于人工智能提供智能服务的方法、智能服务***及智能终端
相关申请的交叉引用
本申请要求于2016年5月17日提交的中国专利申请号为“201610327170.X”的优先权，其全部内容作为整体并入本申请中。
技术领域
本申请涉及人工智能领域，更具体地，涉及基于人工智能提供智能服务的方法、智能服务***及智能终端。
背景技术
人工智能(Artificial Intelligence)，英文缩写为AI。它是研究、开发用于模拟、延伸和扩展人的智能的理论、方法、技术及应用***的一门新的技术科学。人工智能是计算机科学的一个分支，它企图了解智能的实质，并生产出一种新的能以人类智能相似的方式做出反应的智能机器，该领域的研究包括机器人、语言识别、图像识别、自然语言处理和专家***等。
当前，在诸如PC、笔记本电脑、智能手机、平板电脑之类的智能终端的应用程序App领域，涌现出各种各样的基于人工智能的智能服务***(通常又称为智能个人助手、智能个人助理等等)，用户可以用问答对话的方式与其进行交互，例如，用户可以输入询问附近有哪些餐馆、询问近期有哪些电影上映、订外卖等等的请求，而***通常通过检索(包括相关性计算、人工策略等)为用户提供服务。
某些智能服务***可能内置有反馈功能，以便收集用户的使用满意度，从而对***进行训练和改进。然而，这些智能服务***对用户反馈信息的收集以及对***的训练和改进都是线下进行的，其存在的问题至少在于：线下收集的信息不是实时的动态数据。也就是说，在相当长的时间间隔内，用于***训练 和学习的数据都是一成不变的，从而导致用户得到的第一服务结果也是固定不变的，使得检索准确率低，大大地影响了用户的使用体验。
在背景技术中公开的上述信息仅用于加强对本发明的背景的理解，因此其可能包含没有形成为本领域普通技术人员所知晓的现有技术的信息。
发明内容
有鉴于此，本申请实施例希望提供一种基于人工智能提供智能服务的方法、智能服务***及智能终端，以至少解决现有技术中存在的问题。
本申请实施例的技术方案是这样实现的：
根据本申请的一个实施例，提供一种基于人工智能提供智能服务的方法，包括：收到用户的第一服务请求；针对所述第一服务请求确定检索词及其权重；根据所述检索词及其权重提供第一服务结果；以及收集所述用户对所述第一服务结果的反馈信息，根据所述反馈信息中的评价信息实时调整所述针对第一服务请求的检索词和/或其权重。
优选地，上述方法可以在智能终端或远程服务器上实现。
优选地，所述反馈信息中的评价信息可以为分值或正、负反馈。
优选地，当所述反馈信息中的评价信息为正反馈或所述分值大于或等于预定阈值时，提升并保存所述针对第一服务请求的检索词的权重。
优选地，上述方法可以进一步包括，当所述反馈信息中的评价信息为负反馈或所述分值低于预定阈值时，根据实时调整后的检索词和/或其权重，立即提供新的第一服务结果。
优选地，所述实时调整所述检索词和/或其权重可以包括：当所述反馈信息中的评价信息为负反馈或所述分值低于预定阈值时，确定在所述用户第一服务请求之前的预定时段内是否存在除第一服务请求之外的一个或多个其它服务请求；以及当确定存在时，根据所述负反馈或分值，结合所述一个或多个其它服务请求中的至少一个的检索词，来实时调整针对第一服务请求的检索词和/或其权重。
优选地，所述确定检索词及其权重可以包括：获取所述用户的实时状态信息和/或历史信息；以及根据所述第一服务请求的内容信息以及所获取的用户的实时状态信息和/或历史信息确定检索词及其权重。
优选地，所述确定检索词及其权重可以包括：根据所述第一服务请求的内容信息以及所获取的用户的实时状态信息和/或历史信息，对所述第一服务请求进行抽象空间表示；根据所述第一服务请求选择用于筛选第一服务结果的数据库检索方式；以及将所述第一服务请求的抽象空间表示与所述数据库检索方式的抽象空间表示进行融合，以得到所述检索词及其权重。
优选地，所述第一服务结果为由多个第一服务结果条目构成的列表，并且当所述用户对该列表中的至少一个第一服务结果条目的反馈的分值低于预定阈值或为负反馈时，根据实时调整后的检索词和/或其权重，对该列表中的第一服务结果条目进行实时重新排序，并立即提供重新排序后的列表。
根据本申请的另一实施例，提供一种基于人工智能的智能服务***，包括：请求接收模块，被配置为接收用户的第一服务请求；增强学习模块，被配置为针对所述第一服务请求确定检索词及其权重；服务处理模块，被配置为根据所述检索词及其权重提供第一服务结果；以及反馈功能模块，被配置为收集所述用户对所述第一服务结果的反馈信息，以得到所述反馈信息中的评价信息；其中，所述增强学习模块进一步被配置为根据所述反馈信息中的评价信息实时调整针对第一服务请求的检索词和/或其权重。
优选地，所述反馈信息中的评价信息可以为分值或正、负反馈。
优选地，所述增强学习模块可以被配置为：当所述反馈信息中的评价信息为正反馈或所述分值大于或等于预定阈值时，提升并保存所述针对第一服务请求的检索词的权重
优选地，所述服务处理模块可以进一步被配置为当所述反馈信息中的评价信息为负反馈或所述分值低于预定阈值时，根据实时调整后的所述检索词和/或其权重，立即提供新的第一服务结果。
优选地，所述增强学习模块可以被配置为：当所述反馈信息中的评价信息 为负反馈或所述分值低于预定阈值时，确定在所述用户第一服务请求之前的预定时段内是否存在除第一服务请求之外的一个或多个其它服务请求；以及当确定存在时，根据所述负反馈或分值，结合所述一个或多个其它服务请求中的至少一个的检索词，来实时调整针对第一服务请求的检索词和/或其权重。
优选地，所述增强学习模块可以进一步被配置为：获取所述用户的实时状态信息和/或历史信息；以及根据所述第一服务请求的内容信息以及所获取的用户的实时状态信息和/或历史信息确定检索词及其权重。
优选地，所述增强学习模块可以被配置为：根据所述第一服务请求的内容信息以及所获取的用户的实时状态信息和/或历史信息，对所述第一服务请求进行文本抽象空间表示；根据所述第一服务请求选择用于筛选第一服务结果的数据库检索方式；将所述文本抽象空间表示与所述数据库检索方式的抽象空间表示进行融合，以得到所述检索词及其权重。
优选地，所述第一服务结果可以为由多个第一服务结果条目构成的列表，并且所述服务处理模块可以被配置为当所述用户对该列表中的至少一个第一服务结果条目的反馈的分值低于预定阈值或为负反馈时，根据实时调整后的检索词和/或其权重，对该列表中的第一服务结果条目进行实时重新排序，并立即提供重新排序后的列表。
根据本申请的又一实施例，提供一种基于人工智能的智能终端，包括与用户互动的接口设备以及与所述接口设备连接的编译设备，所述编译设备包括：存储器；以及与所述存储器连接的处理器；所述处理器被配置为接收用户的第一服务请求；针对所述第一服务请求确定检索词及其权重；根据所述检索词及其权重提供第一服务结果；以及收集所述用户对所述第一服务结果的反馈信息，根据所述反馈信息中的评价信息实时调整所述针对第一服务请求的检索词和/或其权重。
根据本申请实施例的提供基于人工智能的智能服务的方法、智能服务***及智能终端，能够实时收集用户针对服务结果的反馈信息(例如，包括“喜欢”这种正反馈信息和“不喜欢”这种负反馈信息)，并根据该反馈信息实时调整系 统参数得到新的检索词、检索词的新权重、或两者。这样实现了对***的实时训练和改进，从而提高了检索准确率。
这部分旨在提供对本专利申请的主题的概述。这部分并非旨在提供本发明的排他性的或详尽的说明。本文包括了详细的描述，以提供关于本专利申请的进一步信息。
附图说明
在附图中(这些附图不一定是按照比例绘制的)，相同的数字能够描述不同视图中的相似部件。具有不同字母后缀的相同数字能够表示相似部件的不同示例。附图通过示例而非限制的方式概括地示例了本申请中讨论的各个实施例。
图1示出了根据本申请一个实施例的基于人工智能提供智能服务的方法100的流程图。
图2示出了确定检索词及其权重的具体实现方式的示意图。
图3示出了根据本申请另一实施例的基于人工智能提供智能服务的方法300的流程图。
图4a-4c示出了根据本申请另一实施例的基于人工智能提供智能服务的方法的一个应用示例的示意图。
图5为实现本申请实施例的基于人工智能的智能服务***的物理架构500的结构示意图。
图6为实现本申请实施例的基于人工智能的智能终端的框图。
图7为实现图3所示的方法实施例的交互流程图。
图8为根据本申请实施例的基于人工智能的智能服务***的结构示意图。
图9示出了适于用来实现本申请实施例的终端设备或服务器的计算机***的结构示意图。
具体实施方式
在下文中，仅简单地描述了某些示例性实施例。正如本领域技术人员可能 认识的那样，在不脱离本发明的精神或范围的情况下，可通过各种不同方式修改所描述的实施例。因此，附图和描述被认为本质上是示例性的而非限制性的。
在本申请的各个实施例中，智能服务***与用户进行交互(以及优选地，与服务器数据库或用户所使用的智能终端进行交互)，解析用户输入的第一服务请求的内容(以及优选地，结合用户的实时状态数据和/或历史数据，例如，用户当前状态，用户当前所处的地理位置，用户的性别，用户的年龄，用户所在地的当地时间，用户的行为日志等等)以得到检索词；实时收集用户针对第一服务结果的反馈信息(例如，包括“喜欢”这种正反馈信息和“不喜欢”这种负反馈信息)，并根据该反馈信息实时调整***参数得到新的检索词、检索词的新权重、或两者。这样实现了对***的实时训练和改进，从而提高了检索准确率。优选地，在收集到负反馈信息并且用户仍在当前检索服务界面的情况下，***可以根据新的检索词和/或其权重，立即提供新的第一服务结果。如此一来，在用户不满意检索结果时，***能够及时收集对检索结果不满意的有效反馈信息，从而立即对第一服务结果进行优化，以提升用户的使用体验。
图1示出了根据本申请一个实施例的基于人工智能提供智能服务的方法100的流程图。该方法由将在下文中具体描述的基于人工智能的智能服务***来实现。在一个示例中，该方法的全部步骤可以在诸如PC、笔记本电脑、智能手机以及平板电脑之类的智能终端中(例如，通过预先安装于其中的应用程序App)实现。在另一示例中，也可以由与智能终端可通信地连接的远程计算机设备(例如，远程服务器)实现全部或部分步骤，而智能终端中安装的App或上网浏览器中显示的检索页面可以仅作为与用户交互的界面或实现部分步骤。
如图1所示，该方法100包括如下步骤：
在步骤102，收到用户的第一服务请求。
在一个示例中，该第一服务请求可以是文本请求输入、语音请求输入和图片请求输入中的至少一种。在一个示例中，智能服务***也可以提供服务列表供用户选择，在这种情况下，从用户的角度来看，该第一服务请求可以表现为点击服务列表中的某个服务类型的手势。在一个示例中，从用户的角度来看， 该第一服务请求可以表现为针对之前***提供的服务结果的进一步请求，例如，点击服务结果中的“查看详情”的链接的手势。该步骤实际上是用户提出问题的过程，例如，订北京到深圳的飞机票。
在步骤104，针对所接收的用户的第一服务请求确定检索词及其权重。
这里所提及的检索词涵盖了区别于现有技术中的检索词的含义。以中文文本输入的第一服务请求为例，在现有技术中，检索词是根据预定的中文分词算法将输入的中文文本切分而成的，也就是说，检索词提取自第一服务请求的内容信息。而在本申请中，检索词可以不仅仅限于第一服务请求的内容信息。在一个示例中，在确定检索词及其权重之前，可以获取用户的实时状态信息和历史信息中的至少一者。用户的实时状态信息可以包括用户当前所处的地理位置，用户的性别，用户的年龄，用户所在地的当地时间等等。智能服务***可以从服务器数据库和/或用户当前使用的智能终端中收集用户的实时状态信息，以便更准确地预测用户的当前需求。用户的历史信息可以包括用户浏览过的页面、页面长度、浏览时间、用户的查询日志等等。智能服务***可以从用户当前使用的智能终端中收集用户的历史信息，以便更准确地预测用户的兴趣偏好。由此，智能服务***可以根据第一服务请求的内容信息以及所获取的用户的实时状态信息和/或历史信息来确定检索词及其权重。将在下文中对确定检索词及其权重的具体实现方式进行详细描述。
在步骤106，根据所述检索词及其权重提供第一服务结果。智能服务***可以利用检索词及其权重进行检索或执行相应处理，以返回第一服务结果。
针对不同类型的第一服务请求，第一服务结果具有不同的表现形式。例如，如果用户询问“附近有哪些餐馆”，则第一服务结果表现为用户所在地理位置附近的多个餐馆的列表；如果用户询问“今天天气怎么样”，则第一服务结果表现为当天用户所在地理位置的天气信息。
在步骤108，收集所述用户对所述第一服务结果的反馈信息，根据所述反馈信息中的评价信息实时调整所述针对第一服务请求的检索词和/或其权重。
用户对服务结果的反馈可以被量化成分值。反馈的分值可以是用户对某个 服务结果的评分，也可以是智能服务***根据预定算法对用户针对某个服务结果的行为的打分。在某些情况下，反馈可以简单地划分为正反馈和负反馈。在正反馈(对检索结果满意)的情况下，该反馈可以表现为点赞、语音(例如，检索结果太有用了！)、点击检索结果(即便没有下单)、点击检索结果并下单等。在负反馈(对检索结果不满意)的情况下，该反馈可以表现为长时间不对检索结果进行操作、关闭某个检索结果、关闭该检索服务、关闭整个APP、语音(例如：检索结果好烂！)、转换话题等。
用户对服务结果的反馈(例如，喜欢/不喜欢)会保留在基于人工智能的智能服务***中。简单来说，当用户对服务结果满意时，智能服务***可以提高检索词的权重。当用户对服务结果不满意时，智能服务***可以降低当前检索词的权重、启用新的检索词。通过对用户反馈信息的学习，实时调整检索词和/或其权重，智能服务***会不断筛选出用户可能喜欢的结果，而将质量不高的结果过滤掉。通过这一方式，智能服务***可以不断提升服务质量，从而提高用户满意度，提高***粘性。
图2示出了图1所示的方法100中确定检索词及其权重的具体实现方式的示意图。如上所述，为了提高检索的准确度，当用户输入服务请求202时，智能服务***可以根据服务请求的输入信息以及所获取的用户的实时状态信息和/或历史信息来确定检索词及其权重。智能服务***在确定检索词及其权重时，需要对从服务请求的内容信息以及所获取的用户的实时状态信息和/或历史信息提取的关键词进行抽象空间表示204。可以利用神经网络来对其进行抽象空间表示，上述关键词经抽象空间表示后转化为向量的形式。此外，可以根据用户的服务请求选择用于筛选服务结果的数据库检索方式206。检索方式即为对数据库中的结果进行筛选的方式的刻画，通常***可以根据需要预先设置多种检索方式。然后，将服务请求的抽象空间表示204(即输入向量)与数据库检索方式的抽象空间208表示进行融合210，以得到***检索所需要的检索词212，每个检索词可具有预定的权重。当用户对根据上述检索词及其权重提供的服务结果做出反馈214时，***可以根据反馈实时调整检索词和/或对检索词进行降 权/升权。具体地，可以将所述反馈与所述服务请求的抽象空间表示以及所述数据库检索方式的抽象空间表示进行融合，以得到实时调整后的检索词和/或其权重。具体地，***可利用随机梯度下降的方法，对整个***中的参数进行更新，以使得用户给予的反馈最正向。
图3示出了根据本申请另一实施例的基于人工智能提供智能服务的方法300的流程图。在本实施例中，与图1所示的方法100相同的步骤使用相同的附图标记，并且在此不再赘述。下面仅描述与方法100不同的步骤。
在步骤110，当反馈信息中的评价信息为负反馈或分值低于预定阈值时，根据实时调整后的检索词和/或其权重，立即提供新的第一服务结果。
例如，如果用户对当前的服务结果不满意，用户可以在智能终端上执行横向刷屏的手势，刷掉当前的服务结果。此时，如果用户仍在当前的检索界面下，智能服务***会提供新的服务结果。在一个示例中，在用户接收到新的服务结果时，用户可以继续对新的服务结果做出反馈。智能服务***通过持续地收集用户反馈，不断调整检索词和/或权重，构成一个闭环的训练学习和不断优化的流程，可以不断提升返回的服务结果的质量，以便通过训练和学习向用户提供最佳的答案。通过这种交互方式，智能服务***极大地提高了在短时间内提供令人满意的服务结果的概率，从而大大提升了用户的体验。
在某些情况下，存在如下的使用场景：用户在某个时间段内向智能服务***提出了多个服务请求；相应地，针对每个服务请求，智能服务***确定相应的检索词及其权重并提供相应的服务结果。当用户对多个服务结果中的一个的反馈为负反馈时，结合针对在对应的服务请求之前的至少一个服务请求的检索词，实时调整针对该对应的服务请求的检索词和/或其权重。简言之，智能服务***通过训练和学习，可以结合用户与***对话的上下文来优化针对用户当前的服务请求的检索词及其权重。例如，经优化后的检索词中可能包括隐含但并未显现在用户当前的服务请求的输入信息中的关键词。
在一个示例中，***可以记录接收的用户的服务请求及反馈作为该用户的历史信息，以便用于以后的智能服务中。或者，***可以将针对该用户的经优 化的神经网络模型存储在该用户的智能终端或者远程服务器中，以便以后为该用户提供个性化的智能服务。
在一个示例中，***可以收集多个用户的服务请求及反馈，并且根据所收集的多个用户的服务请求及反馈，实时更新数据库检索方式。以这样的方式，可以使智能服务***随着使用次数的增多在总体上得到优化，从而不断提升用户的使用体验。
图4示出了根据本发明另一实施例的基于人工智能提供智能服务的方法的一个应用示例的示意图。如图4a和4b所示，用户与***通过安装于智能终端上的App的界面进行了如下对话：
用户：我下午要去深圳出差，请帮我订一下机票。
***：....(完成订机票的动作)
用户：下午那边的天气如何？
***：北京天气，雾转霾。
用户：(通过手势划掉结果，表示负反馈)
如图4c所示，***在收到负反馈之后，立即提供了新的服务结果：
***：已为您实时调整结果。
***：深圳，白天阴转晴，夜晚有雨。
整个***的工作机制如下：
A:根据用户与***的对话过程，提取当前的关键特征，并利用已有的增强学习的神经网络模型对关键词进行实时的检索权重打分。例如关键词及其检索权重：下午:0.1、深圳:0.1、出差:0.3、机票:0.1、那边:0.01、天气:0.5、用户所处地点(北京):0.6。
B:由于检索词的预测权重倾向于询问北京天气，***回复了北京天气。
C:***收到负反馈之后，对神经网络模型进行更新，重新对检索词进行打分。需要注意的是，每次的检索词是可以不一样的，权重可能也不一样，但模型更倾向于预测与上下文相关的词作为检索词。
D:通过神经网络的调权，如果用户给出负反馈，那么最初检索词权重较高 的检索词，天气、用户所处地点(北京)会受到强烈打压，而权重较低的检索词权重受到的打压比较小。这时候再利用增强学习模型对检索词及其权重进行一次预测，检索词及其权重变为：下午:0.09、深圳:0.09、出差:0.2、机票:0.09、那边:0.009、天气:0.1、用户所处地点(北京):0.12。
E:***根据最新预测的检索词及其权重进行回复，深圳今天白天阴转晴，夜晚有小雨。
其中，整个增强学习的神经网络模型的输入是：用户所处状态的关键特征，例如用户与***对话的关键词，当时的时间，地点等。本例子中，下午、深圳、请、帮我等，都可以作为特征进行输入。增强学习的神经网络模型的输出是(更新前)：下午:0.1、深圳:0.1、出差:0.3、机票:0.1、那边:0.01、天气:0.5、用户所处地点(北京):0.6。输入和输出一般比较关联，但不一定完全一致。权重也是通过整个神经网络给出的。
此外，当用户提供正反馈时，神经网络对获得正反馈的这组预测词的权重会被加强，从而强化下一次预测的效果。
图5为实现本申请实施例的基于人工智能的智能服务***的物理架构的结构示意图。参照图5，其示出一个基于人工智能的智能服务***。该***500包括智能终端510以及一个或多个服务器530，这些智能终端510和服务器530通过网络520连接。智能终端510中安装有智能服务App或网页浏览器，用户通过使用该App或网页浏览器访问智能服务请求界面来向***提出服务请求，接收服务结果，并对服务结果做出反馈。
智能终端510可以是智能手机、PC、平板电脑、笔记本电脑、智能机器人等等。图6为实现本发明实施例的基于人工智能的智能终端510的框图。该终端510包括与用户互动的接口设备，与接口设备连接的编译设备，以及与编译设备连接的联网模块630。其中，与用户互动的接口设备可以是触摸屏640、音频输出设备650(包括扬声器、耳机等)、麦克风660；编译设备可以是处理器610、存储器620。处理器610被配置为结合其他元件执行上述本发明实施例的方法的全部或部分步骤。联网模块630被配置为能使该终端510与服务器530 之间通信，例如从服务器530下载服务结果，将服务请求发送到服务器上等等。存储器620被配置为存储从服务器下载的服务结果的信息(例如，文本、语音、图片等等)。触摸屏640被配置为接收用户的文本输入，识别用户的手势，并显示用户的服务请求、***提供的服务结果以及其他相关信息。音频输出设备650被配置为播放服务结果及***提示信息。麦克风660被配置为采集用户的语音信息。
下面结合图3所示的方法实施例来描述实现该实施例的交互流程。图7为实现该实施例的交互流程图。该交互流程涉及用户1、用户1使用的智能终端510-1，用户2、用户2使用的智能终端510-2、服务器530以及除了用户1、用户2之外的其它用户及其使用的智能终端(未示出)。为了便于描述，这里对智能终端510-1和智能终端510-2实现的功能进行了区分。应该理解，这些功能均可以在一个智能终端上实现。该交互流程包括以下步骤：
步骤701、用户1通过在终端510-1的触摸屏上执行轻敲手势来启动App；
步骤702、终端510-1启动App；
步骤703、终端510-1显示常用服务类型列表和服务请求输入接口；
步骤704、用户1通过在终端510-1的触摸屏上执行轻敲手势来选择一种服务类型；
步骤705、终端510-1向服务器530发送与所选择的服务类型对应的服务请求；
步骤706、服务器530通过***的神经网络模型确定检索词及其权重；
步骤707、服务器530根据检索词及其权重向终端510-1返回服务结果；
步骤708、终端510-1显示服务结果；
步骤709、用户1在终端510-1的触摸屏上执行横向刷屏的手势以表示对该服务结果不满意；
步骤710、终端510-1向服务器530发送用户1对该服务结果的负反馈；
步骤711、服务器530以降低当前检索词的权重、用新的检索词替换掉部分当前检索词的方式更新针对用户1的神经网络模型；
步骤712、服务器530根据新的检索词及其权重向终端510-1返回新的服务结果；
步骤713、终端510-1显示新的服务结果；
步骤714、用户1为该新的服务结果点赞以表示对该服务结果满意；
步骤715、终端510-1向服务器530发送用户1对该新的服务结果的正反馈；
步骤716、服务器530提升当前检索词的权重；
步骤717、用户1关闭App；
步骤718、重复步骤701-705；
步骤719、服务器530利用前次更新后的神经网络模型返回服务结果；
步骤720、终端510-1显示服务结果；
步骤721：服务器530针对同一服务请求接收到来自多个用户的相似反馈，更新***的神经网络模型；
步骤722：用户2通过在终端510-2的触摸屏上执行轻敲手势来启动App；
步骤723、终端510-2启动App；
步骤724、终端510-2显示常用服务类型列表和服务请求输入接口；
步骤725、用户2通过在终端510-2的触摸屏上执行轻敲手势来选择上述的服务类型；
步骤726、终端510-2向服务器530发送与所选择的服务类型对应的服务请求；
步骤727、服务器530通过在步骤721中更新后的***的神经网络模型确定检索词及其权重，
步骤728、服务器530根据检索词及其权重向终端510-2返回服务结果；
步骤729、终端510-2显示该服务结果。
这里需要指出的是，以上描述的交互流程仅仅是一个示例，实际的交互流程并不仅限于此。
基于上面方法实施例，本申请提供了基于人工智能的智能服务***的实施例。以下***实施例的描述，与上述方法描述是类似的，同方法的有益效果描 述，不做赘述。对于本申请***实施例中未披露的技术细节，请参照本申请方法实施例的描述。需要注意的是，以下所述的智能服务***中所包括的模块，可以全部在智能终端中实现，也可以全部在远程服务器中实现，或者可以一部分在智能终端中实现，其余部分在远程服务器中实现。
如图8所示，所述基于人工智能的智能服务***包括：
请求接收模块802，被配置为接收用户的第一服务请求；
增强学习模块804，被配置为针对第一服务请求确定检索词及其权重；
服务处理模块806，被配置为根据检索词及其权重提供第一服务结果；以及
反馈功能模块808，被配置为收集用户对第一服务结果的反馈信息，以得到反馈信息中的评价信息。
所述增强学习模块804可以进一步被配置为根据反馈信息中的评价信息实时调整针对第一服务请求的检索词和/或其权重。
在一个示例中，所述反馈信息中的评价信息可以为分值或正、负反馈。
在一个示例中，服务处理模块可以进一步被配置为当反馈信息中的评价信息为负反馈或分值低于预定阈值时，根据实时调整后的检索词和/或其权重，立即提供新的第一服务结果。
在一个示例中，当所述反馈信息中的评价信息为正反馈或所述分值大于或等于预定阈值时，服务处理模块可以提升并保存所述针对第一服务请求的检索词的权重。
在一个示例中，当所述反馈信息中的评价信息为负反馈或所述分值低于预定阈值时，增强学习模块可以确定在第一服务请求之前的预定时段内是否存在除第一服务请求之外的一个或多个其它服务请求，当确定存在时，根据负反馈或分值，结合所述一个或多个其它服务请求中的至少一个的检索词，来实时调整针对第一服务请求的检索词和/或其权重。
在一个示例中，增强学习模块可以进一步被配置为获取用户的实时状态信息和/或历史信息，并且根据第一服务请求的内容信息以及所获取的用户的实时 状态信息和/或历史信息确定检索词及其权重。
在一个示例中，增强学习模块被配置为根据第一服务请求的内容信息以及所获取的用户的实时状态信息和/或历史信息，对第一服务请求进行文本抽象空间表示；根据第一服务请求选择用于筛选第一服务结果的数据库检索方式；将所述文本抽象空间表示与数据库检索方式的抽象空间表示进行融合，以得到检索词及其权重。
在一个示例中，所述第一服务结果可以为由多个第一服务结果条目构成的列表，并且所述服务处理模块可以被配置为当所述用户对该列表中的至少一个第一服务结果条目的反馈的分值低于预定阈值或为负反馈时，根据实时调整后的检索词和/或其权重，对该列表中的第一服务结果条目进行实时重新排序，并立即提供重新排序后的列表。
在一个示例中，该智能服务***可以进一步包括：历史信息记录模块(未示出)，被配置为记录所述用户的所述第一服务请求及所述反馈作为该用户的历史信息。
在一个示例中，该智能服务***可以进一步包括：更新模块(未示出)，被配置为收集多个用户的第一服务请求及反馈，以及根据所收集的多个用户的第一服务请求及反馈，更新所述数据库检索方式。
图9示出了适于用来实现本申请实施例的终端设备或服务器的计算机***的结构示意图。
如图9所示，计算机***900包括中央处理单元(CPU)901，其可以根据存储在只读存储器(ROM)902中的程序或者从存储部分908加载到随机访问存储器(RAM)903中的程序而执行各种适当的动作和处理。在RAM 903中，还存储有***900操作所需的各种程序和数据。CPU 901、ROM 902以及RAM903通过总线904彼此相连。输入/输出(I/O)接口905也连接至总线904。
以下部件连接至I/O接口905：包括键盘、鼠标等的输入部分906；包括诸如阴极射线管(CRT)、液晶显示器(LCD)等以及扬声器等的输出部分907；包括硬盘等的存储部分908；以及包括诸如LAN卡、调制解调器等的网络接口 卡的通信部分909。通信部分909经由诸如因特网的网络执行通信处理。驱动器910也根据需要连接至I/O接口905。可拆卸介质911，诸如磁盘、光盘、磁光盘、半导体存储器等等，根据需要安装在驱动器910上，以便于从其上读出的计算机程序根据需要被安装入存储部分908。
特别地，根据本公开的实施例，上文参考流程图描述的过程可以被实现为计算机软件程序。例如，本公开的实施例包括一种计算机程序产品，其包括有形地包含在机器可读介质上的计算机程序，所述计算机程序包含用于执行流程图所示的方法的程序代码。在这样的实施例中，该计算机程序可以通过通信部分909从网络上被下载和安装，和/或从可拆卸介质911被安装。
附图中的流程图和框图，图示了按照本申请各种实施例的***、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段、或代码的一部分，所述模块、程序段、或代码的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。也应当注意，在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个接连地表示的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图和/或流程图中的每个方框、以及框图和/或流程图中的方框的组合，可以用执行规定的功能或操作的专用的基于硬件的***来实现，或者可以用专用硬件与计算机指令的组合来实现。
在本申请所提供的几个实施例中，应该理解到，所揭露的设备和方法可以通过其它的方式实现。以上所描述的设备实施例仅仅是示意性的，例如，所述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，如：多个单元或组件可以结合，或可以集成到另一个***，或一些特征可以忽略，或不执行。另外，所显示或讨论的各组成部分相互之间的耦合、或直接耦合、或通信连接可以是通过一些接口，设备或单元的间接耦合或通信连接，可以是电性的、机械的或其它形式的。
上述作为分离部件说明的单元可以是、或也可以不是物理上分开的，作为 单元显示的部件可以是、或也可以不是物理单元，即可以位于一个地方，也可以分布到多个网络单元上；可以根据实际的需要选择其中的部分或全部单元来实现本实施例方案的目的。
另外，在本发明各实施例中的各功能单元可以全部集成在一个处理单元中，也可以是各单元分别单独作为一个单元，也可以两个或两个以上单元集成在一个单元中；上述集成的单元既可以采用硬件的形式实现，也可以采用硬件加软件功能单元的形式实现。
本领域普通技术人员可以理解：实现上述方法实施例的全部或部分步骤可以通过程序指令相关的硬件来完成，前述的程序可以存储于一计算机可读取存储介质中，该程序在执行时，执行包括上述方法实施例的步骤；而前述的存储介质包括：移动存储设备、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、磁碟或者光盘等各种可以存储程序代码的介质。
或者，本发明上述集成的单元如果以软件功能模块的形式实现并作为独立的产品销售或使用时，也可以存储在一个计算机可读取存储介质中。基于这样的理解，本发明实施例的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机、服务器、或者网络设备等)执行本发明各个实施例所述方法的全部或部分。而前述的存储介质包括：移动存储设备、ROM、RAM、磁碟或者光盘等各种可以存储程序代码的介质。
以上所述，仅为本发明的具体实施方式，但本发明的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本发明揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本发明的保护范围之内。因此，本发明的保护范围应以所述权利要求的保护范围为准。

Claims (20)

  1. 一种基于人工智能提供智能服务的方法，包括：
    收到用户的第一服务请求；
    针对所述第一服务请求确定检索词及其权重；
    根据所述检索词及其权重提供第一服务结果；以及
    收集所述用户对所述第一服务结果的反馈信息，根据所述反馈信息中的评价信息实时调整所述针对第一服务请求的检索词和/或其权重。
  2. 根据权利要求1所述的方法，其中，所述方法在智能终端或与所述智能终端可通信地连接的远程计算机设备上实现。
  3. 根据权利要求1所述的方法，其中，所述反馈信息中的评价信息为分值或正、负反馈。
  4. 根据权利要求3所述的方法，其中，当所述反馈信息中的评价信息为正反馈或所述分值大于或等于预定阈值时，提升并保存所述针对第一服务请求的检索词的权重。
  5. 根据权利要求3所述的方法，进一步包括，当所述反馈信息中的评价信息为负反馈或所述分值低于预定阈值时，根据实时调整后的检索词和/或其权重，立即提供新的第一服务结果。
  6. 根据权利要求5所述的方法，其中，所述实时调整所述检索词和/或其权重包括：
    当所述反馈信息中的评价信息为负反馈或所述分值低于预定阈值时，确定在所述用户第一服务请求之前的预定时段内是否存在除第一服务请求之外的一个或多个其它服务请求；以及
    当确定存在时，根据所述负反馈或分值，结合所述一个或多个其它服务请求中的至少一个的检索词，来实时调整针对第一服务请求的检索词和/或其权重。
  7. 根据权利要求1所述的方法，其中，所述确定检索词及其权重包括：
    获取所述用户的实时状态信息和/或历史信息；以及
    根据所述第一服务请求的内容信息以及所获取的用户的实时状态信息和/或历史信息确定检索词及其权重。
  8. 根据权利要求7所述的方法，其中，所述确定检索词及其权重包括：
    根据所述第一服务请求的内容信息以及所获取的用户的实时状态信息和/或历史信息，对所述第一服务请求进行抽象空间表示；
    根据所述第一服务请求选择用于筛选第一服务结果的数据库检索方式；以及
    将所述第一服务请求的抽象空间表示与所述数据库检索方式的抽象空间表示进行融合，以得到所述检索词及其权重。
  9. 根据权利要求5所述的方法，其中，所述第一服务结果为由多个第一服务结果条目构成的列表，并且
    其中，当所述用户对该列表中的至少一个第一服务结果条目的反馈的分值低于预定阈值或为负反馈时，根据实时调整后的检索词和/或其权重，对该列表中的第一服务结果条目进行实时重新排序，并立即提供重新排序后的列表。
  10. 一种基于人工智能的智能服务***，包括：
    请求接收模块，被配置为接收用户的第一服务请求；
    增强学习模块，被配置为针对所述第一服务请求确定检索词及其权重；
    服务处理模块，被配置为根据所述检索词及其权重提供第一服务结果；以及
    反馈功能模块，被配置为收集所述用户对所述第一服务结果的反馈信息，以得到所述反馈信息中的评价信息；
    其中，所述增强学习模块进一步被配置为根据所述反馈信息中的评价信息实时调整针对第一服务请求的检索词和/或其权重。
  11. 根据权利要求10所述的***，其中，所述反馈信息中的评价信息为分值或正、负反馈。
  12. 根据权利要求11所述的***，其中，所述增强学习模块被配置为：
    当所述反馈信息中的评价信息为正反馈或所述分值大于或等于预定阈值时，提升并保存所述针对第一服务请求的检索词的权重。
  13. 根据权利要求11所述的***，其中，所述服务处理模块进一步被配置为当所述反馈信息中的评价信息为负反馈或所述分值低于预定阈值时，根据实时调整后的所述检索词和/或其权重，立即提供新的第一服务结果。
  14. 根据权利要求13所述的***，其中，所述增强学习模块被配置为：
    当所述反馈信息中的评价信息为负反馈或所述分值低于预定阈值时，确定在所述用户第一服务请求之前的预定时段内是否存在除第一服务请求之外的一个或多个其它服务请求；以及
    当确定存在时，根据所述负反馈或分值，结合所述一个或多个其它服务请求中的至少一个的检索词，来实时调整针对第一服务请求的检索词和/或其权重。
  15. 根据权利要求10所述的***，其中，所述增强学习模块进一步被配置为：
    获取所述用户的实时状态信息和/或历史信息；以及
    根据所述第一服务请求的内容信息以及所获取的用户的实时状态信息和/或历史信息确定检索词及其权重。
  16. 根据权利要求15所述的***，其中，所述增强学习模块被配置为：
    根据所述第一服务请求的内容信息以及所获取的用户的实时状态信息和/或历史信息，对所述第一服务请求进行文本抽象空间表示；
    根据所述第一服务请求选择用于筛选第一服务结果的数据库检索方式；
    将所述文本抽象空间表示与所述数据库检索方式的抽象空间表示进行融合，以得到所述检索词及其权重。
  17. 根据权利要求13所述的***，其中，所述第一服务结果为由多个第一服务结果条目构成的列表，并且
    其中，所述服务处理模块被配置为当所述用户对该列表中的至少一个第一服务结果条目的反馈的分值低于预定阈值或为负反馈时，根据实时调整后的检 索词和/或其权重，对该列表中的第一服务结果条目进行实时重新排序，并立即提供重新排序后的列表。
  18. 一种基于人工智能的智能终端，包括与用户互动的接口设备以及与所述接口设备连接的编译设备，所述编译设备包括：
    存储器；以及
    与所述存储器连接的处理器；
    所述处理器被配置为接收用户的第一服务请求；针对所述第一服务请求确定检索词及其权重；根据所述检索词及其权重提供第一服务结果；以及收集所述用户对所述第一服务结果的反馈信息，根据所述反馈信息中的评价信息实时调整所述针对第一服务请求的检索词和/或其权重。
  19. 一种设备，包括：
    处理器；和
    存储器，
    所述存储器中存储有能够被所述处理器执行的计算机可读指令，在所述计算机可读指令被执行时，所述处理器执行基于人工智能提供智能服务的方法，所述方法包括：
    收到用户的第一服务请求；
    针对所述第一服务请求确定检索词及其权重；
    根据所述检索词及其权重提供第一服务结果；以及
    收集所述用户对所述第一服务结果的反馈信息，根据所述反馈信息中的评价信息实时调整所述针对第一服务请求的检索词和/或其权重。
  20. 一种非易失性计算机存储介质，所述计算机存储介质存储有能够被处理器执行的计算机可读指令，当所述计算机可读指令被处理器执行时，所述处理器执行基于人工智能提供智能服务的方法，所述方法包括：
    收到用户的第一服务请求；
    针对所述第一服务请求确定检索词及其权重；
    根据所述检索词及其权重提供第一服务结果；以及
    收集所述用户对所述第一服务结果的反馈信息，根据所述反馈信息中的评价信息实时调整所述针对第一服务请求的检索词和/或其权重。
PCT/CN2016/097292 2016-05-17 2016-08-30 基于人工智能提供智能服务的方法、智能服务***及智能终端 WO2017197806A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US16/193,454 US11651002B2 (en) 2016-05-17 2018-11-16 Method for providing intelligent service, intelligent service system and intelligent terminal based on artificial intelligence

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201610327170.XA CN106021463B (zh) 2016-05-17 2016-05-17 基于人工智能提供智能服务的方法、智能服务***及智能终端
CN201610327170.X 2016-05-17

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US16/193,454 Continuation US11651002B2 (en) 2016-05-17 2018-11-16 Method for providing intelligent service, intelligent service system and intelligent terminal based on artificial intelligence

Publications (1)

Publication Number Publication Date
WO2017197806A1 true WO2017197806A1 (zh) 2017-11-23

Family

ID=57097346

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2016/097292 WO2017197806A1 (zh) 2016-05-17 2016-08-30 基于人工智能提供智能服务的方法、智能服务***及智能终端

Country Status (3)

Country Link
US (1) US11651002B2 (zh)
CN (1) CN106021463B (zh)
WO (1) WO2017197806A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112199470A (zh) * 2020-10-09 2021-01-08 康佳集团股份有限公司 一种基于会话的客诉服务方法、智能终端及存储介质

Families Citing this family (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106897378A (zh) * 2017-01-20 2017-06-27 浙江大学 一种面向三维cad模型的语义参数检索动态调节方法
CN106919702B (zh) * 2017-02-14 2020-02-11 北京时间股份有限公司 基于文档的关键词推送方法及装置
CN107169586A (zh) * 2017-03-29 2017-09-15 北京百度网讯科技有限公司 基于人工智能的资源组合优化方法、装置及存储介质
EP3446469B1 (en) * 2017-06-23 2021-09-29 Beijing Didi Infinity Technology And Development Co., Ltd. Method of user behavior based service dispatch
US11721090B2 (en) * 2017-07-21 2023-08-08 Samsung Electronics Co., Ltd. Adversarial method and system for generating user preferred contents
CN107688950B (zh) * 2017-08-31 2022-03-11 广东美的制冷设备有限公司 家电售后咨询方法、电子设备和计算机可读存储介质
US11698936B2 (en) * 2017-10-09 2023-07-11 Home Depot Product Authority, Llc System and methods for search engine parameter tuning using genetic algorithm
US10740339B2 (en) * 2017-12-18 2020-08-11 Microsoft Technology Licensing, Llc Query term weighting
CN110110049A (zh) * 2017-12-29 2019-08-09 深圳市优必选科技有限公司 服务咨询方法、装置、***、服务机器人及存储介质
CN109074502A (zh) * 2018-07-26 2018-12-21 深圳前海达闼云端智能科技有限公司 训练人工智能模型的方法、装置、存储介质及机器人
CN109816321A (zh) * 2018-11-27 2019-05-28 深圳市汇邦企业服务有限公司 一种服务管理方法、装置、设备和计算机可读存储介质
CN110008220B (zh) * 2019-01-24 2024-03-15 平安科技（深圳）有限公司 黑名单传导图谱衰减系数获取方法、装置和计算机设备
CN111797852A (zh) * 2019-04-09 2020-10-20 Oppo广东移动通信有限公司 算法模型的调整方法、装置、存储介质及电子设备
CN111368158A (zh) * 2020-03-31 2020-07-03 中国建设银行股份有限公司 基于人工智能平台的服务查找方法及装置
US11775533B2 (en) * 2021-01-14 2023-10-03 Capital One Services, Llc Customizing search queries for information retrieval
CN115396389B (zh) * 2022-07-07 2023-09-22 深圳市华汇数据服务有限公司 一种基于情绪能量感知的物联网信息技术客服服务***
CN115454171A (zh) * 2022-08-30 2022-12-09 海尔优家智能科技（北京）有限公司 智能设备及其除雾方法、设备及介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101334796A (zh) * 2008-02-29 2008-12-31 浙江师范大学 一种个性化及协同化融合的网上多媒体检索与查询方法
CN101685454A (zh) * 2008-09-28 2010-03-31 华为技术有限公司 人机交互方法及***
US20100153094A1 (en) * 2008-12-11 2010-06-17 Electronics And Telecommunications Research Institute Topic map based indexing and searching apparatus
CN102387207A (zh) * 2011-10-21 2012-03-21 华为技术有限公司 基于用户反馈信息的推送方法和推送***
CN104699708A (zh) * 2013-12-09 2015-06-10 ***通信集团北京有限公司 一种客服机器人的自学习方法及装置

Family Cites Families (30)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030132944A1 (en) * 2001-10-03 2003-07-17 Sun Microsystems, Inc. User control of generalized semantic zooming
US20040008828A1 (en) * 2002-07-09 2004-01-15 Scott Coles Dynamic information retrieval system utilizing voice recognition
US7610306B2 (en) * 2003-06-30 2009-10-27 International Business Machines Corporation Multi-modal fusion in content-based retrieval
US7580909B2 (en) * 2003-08-26 2009-08-25 Northrop Grumman Corporation Visual representation tool for structured arguments
US20050198026A1 (en) * 2004-02-03 2005-09-08 Dehlinger Peter J. Code, system, and method for generating concepts
US7788248B2 (en) * 2005-03-08 2010-08-31 Apple Inc. Immediate search feedback
US7720792B2 (en) * 2005-04-05 2010-05-18 Content Analyst Company, Llc Automatic stop word identification and compensation
US20070168344A1 (en) * 2006-01-19 2007-07-19 Brinson Robert M Jr Data product search using related concepts
US8150827B2 (en) * 2006-06-07 2012-04-03 Renew Data Corp. Methods for enhancing efficiency and cost effectiveness of first pass review of documents
CN100388282C (zh) * 2006-09-14 2008-05-14 浙江大学 基于多模态信息融合分析的跨媒体检索方法
CN100565515C (zh) * 2006-11-30 2009-12-02 腾讯科技（深圳）有限公司 一种中文自动应答方法及***
US9122743B2 (en) * 2008-01-30 2015-09-01 International Business Machines Corporation Enhanced search query modification
US8677285B2 (en) * 2008-02-01 2014-03-18 Wimm Labs, Inc. User interface of a small touch sensitive display for an electronic data and communication device
US8301437B2 (en) * 2008-07-24 2012-10-30 Yahoo! Inc. Tokenization platform
US20110004588A1 (en) * 2009-05-11 2011-01-06 iMedix Inc. Method for enhancing the performance of a medical search engine based on semantic analysis and user feedback
US8988356B2 (en) * 2009-12-31 2015-03-24 Google Inc. Touch sensor and touchscreen user input combination
US8838169B2 (en) * 2010-01-19 2014-09-16 Ringcentral, Inc. Provisioning interfaces for accessing virtual private branch exchange services through a mobile device
US9263034B1 (en) * 2010-07-13 2016-02-16 Google Inc. Adapting enhanced acoustic models
EP2541473A1 (en) * 2011-06-27 2013-01-02 Amadeus S.A.S. Method and system for a pre-shopping reservation system with increased search efficiency
US8918389B2 (en) * 2011-07-13 2014-12-23 Yahoo! Inc. Dynamically altered search assistance
US9864781B1 (en) * 2013-11-05 2018-01-09 Western Digital Technologies, Inc. Search of NAS data through association of errors
EP3198477B1 (en) * 2014-09-23 2022-07-27 InterDigital Madison Patent Holdings, SAS A method and apparatus for search query formulation
US20160165472A1 (en) * 2014-12-09 2016-06-09 Futurewei Technologies, Inc. Analytics assisted self-organizing-network (SON) for coverage capacity optimization (CCO)
US20160224569A1 (en) * 2015-02-03 2016-08-04 General Electric Company System and method for automatically publishing a web form from a semantic query
US10156978B2 (en) * 2015-03-31 2018-12-18 Lg Electronics Inc. Terminal and operating method thereof
EP3298431A1 (en) * 2015-05-18 2018-03-28 Lasermotive Inc. Light curtain safety system
CN105045692B (zh) * 2015-07-22 2018-11-30 京东方科技集团股份有限公司 触摸屏测试装置及***、触摸屏测试控制装置
AU2016312969A1 (en) * 2015-08-24 2018-03-15 Codafication Pty Ltd Method and electronic system for generating job bookings
US20170092153A1 (en) * 2015-09-30 2017-03-30 Mark Horbal Assistive Grid For Mobile Devices
US11334800B2 (en) * 2016-05-12 2022-05-17 International Business Machines Corporation Altering input search terms

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101334796A (zh) * 2008-02-29 2008-12-31 浙江师范大学 一种个性化及协同化融合的网上多媒体检索与查询方法
CN101685454A (zh) * 2008-09-28 2010-03-31 华为技术有限公司 人机交互方法及***
US20100153094A1 (en) * 2008-12-11 2010-06-17 Electronics And Telecommunications Research Institute Topic map based indexing and searching apparatus
CN102387207A (zh) * 2011-10-21 2012-03-21 华为技术有限公司 基于用户反馈信息的推送方法和推送***
CN104699708A (zh) * 2013-12-09 2015-06-10 ***通信集团北京有限公司 一种客服机器人的自学习方法及装置

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112199470A (zh) * 2020-10-09 2021-01-08 康佳集团股份有限公司 一种基于会话的客诉服务方法、智能终端及存储介质
CN112199470B (zh) * 2020-10-09 2023-12-05 康佳集团股份有限公司 一种基于会话的客诉服务方法、智能终端及存储介质

Also Published As

Publication number Publication date
CN106021463A (zh) 2016-10-12
US11651002B2 (en) 2023-05-16
US20190087472A1 (en) 2019-03-21
CN106021463B (zh) 2019-07-09

Similar Documents

Publication Publication Date Title
WO2017197806A1 (zh) 基于人工智能提供智能服务的方法、智能服务***及智能终端
US11868732B2 (en) System for minimizing repetition in intelligent virtual assistant conversations
US11971936B2 (en) Analyzing web pages to facilitate automatic navigation
CN108700987B (zh) 移动设备上的模拟超链接
US10536580B2 (en) Recommendations based on feature usage in applications
CN112163165B (zh) 信息推荐方法、装置、设备及计算机可读存储介质
US10733197B2 (en) Method and apparatus for providing information based on artificial intelligence
CN112313680A (zh) 助理***中手势输入的自动完成
US20180011931A1 (en) Master content summaries for variant content
WO2021135562A1 (zh) 特征有效性评估方法、装置、电子设备及存储介质
CN109471978B (zh) 一种电子资源推荐方法及装置
CN112380331A (zh) 信息推送的方法和装置
US11967177B2 (en) Method for managing item recommendation using degree of association between language unit and usage history
KR20200009117A (ko) 텍스트 데이터 수집 및 분석을 위한 시스템
US20180143986A1 (en) Generating persuasive electronic messages using interplay of message content features
WO2016138349A1 (en) Systems and methods of structuring reviews with auto-generated tags
CN117708428A (zh) 推荐信息预测方法、装置以及电子设备
CN109710751A (zh) 法律文件的智能推荐方法、装置、设备及存储介质
US11734588B2 (en) Managing domain competence during a computing session
CN114265948A (zh) 图像推送方法和装置
US20240232272A1 (en) Analyzing web pages to facilitate automatic navigation
CN116226548A (zh) 双向匹配方法、双向匹配装置、存储介质及电子设备
WO2024049851A1 (en) System and method for searching media or data based on contextual weighted keywords
CN116975083A (zh) 信息搜索方法、装置、计算机设备和存储介质
CN116151884A (zh) 确定推荐信息资源的方法和装置、计算设备及存储介质

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 16902186

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 16902186

Country of ref document: EP

Kind code of ref document: A1