CN105302880A - Content correlation recommendation method and apparatus - Google Patents
Content correlation recommendation method and apparatus Download PDFInfo
- Publication number
- CN105302880A CN105302880A CN201510661729.8A CN201510661729A CN105302880A CN 105302880 A CN105302880 A CN 105302880A CN 201510661729 A CN201510661729 A CN 201510661729A CN 105302880 A CN105302880 A CN 105302880A
- Authority
- CN
- China
- Prior art keywords
- user
- matrix
- preference
- textual content
- content
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
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
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)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of internet and discloses a content correlation recommendation method and apparatus. The method comprises the steps of: acquiring access data of a user to non-text content; constructing a behavior matrix according to the access data; performing matrix decomposition on the behavior matrix to obtain a user interest vector and a content type vector; calculating the preference degree of the user to the non-text content according to the user interest vector and the content type vector; and performing correlation recommendation according to the preference degree. According to the technical scheme, the user interest vector and the content type vector with very high regularity are obtained through matrix decomposition of the access data of the user, so that relatively standard and rigorous preference degree calculation can be performed, clearly expressed, accurate and strict correlation between the user and the non-text content can be obtained, more accurate recommendation can be performed, and user experience is improved.
Description
Technical field
The present invention relates to Internet technical field, particularly a kind of relevance recommend method and device.
Background technology
The extensive development of internet makes each use freely can participate in the establishment of the network information per family with propagation, and this creates the big bang of the network information on the one hand, provide the user more information; Cause again the confusion of the network information on the other hand, for user finds that actual available information brings considerable hurdle.For network data, text message is the content with certain expression rule, than being easier to be associated so that user search or recommend; But other non-textual contents, such as picture, audio frequency, video etc., will there is very big change in the change that the randomness of its expression is very large and trickle viewed from data plane, there is no obvious expression rule, be difficult to the contact directly found each other, be also difficult to utilize interrelational form to carry out searching for or recommending.
Prior art is generally described by text or text label identifies non-textual content, to set up the association between non-textual content, but the artificial description through specialty or the content through a large amount of statistical study is only had just to there is comparatively detailed text description or label.In a practical situation, simple description or label are difficult to the actual features of precise expression non-textual content.Such as advertisement video, this kind of non-textual content needs stronger association popularization just to have significant effect, because ad material itself there is no too many Information Availability except trade information, and user not merely relies on trade information to distinguish to the preference of some specific product, therefore prior art is difficult to recommend accurately according to user preference.For example, certain user extremely likes Coca-Cola in the recent period, simultaneously also interested in panel computer, concerning this user, Coca-Cola advertisement is similar to panel computer advertisement, but prior art is difficult to advertisement larger for this two classes difference to be considered as similar and carries out association and promote.
Therefore analysis non-textual content being carried out to specialty is difficult in prior art, tagged manner simply can only find that whether the category of employment of two contents is similar, whether cannot meet the hobby of certain specific user in its inherence of accurate assurance, thus also cannot carry out search and the recommendation of non-textual content exactly.
Summary of the invention
Based on the defect of prior art, the object of this invention is to provide a kind of relevance recommend method and device, to set up non-textual content and associating between user efficiently and accurately.
According to an aspect of the present invention, provide a kind of relevance recommend method, comprise step:
Gather user to the visit data of non-textual content;
Behavioural matrix is built according to described visit data;
Described behavioural matrix is carried out matrix decomposition and obtain user interest vector sum content type vector;
According to described user interest vector sum, the described user of content type vector is to the preference of described non-textual content;
Correlation recommendation is carried out according to described preference.
Preferably, described method also comprises step:
After the described visit data of collection, obtain the behavior related information set of user and non-textual content according to described visit data; Wherein said behavior related information is by the access behavior marking also normalization of described user to described non-textual content.
Preferably, described matrix decomposition is carried out by svd.
Preferably, collaborative filtering is used to carry out described svd.
Preferably, the preference of user to non-textual content is calculated by calculating described user interest vector with the inner product of described content type vector.
According to another aspect of the present invention, additionally provide a kind of relevance recommendation apparatus, comprising:
Data acquisition module, for gathering the visit data of user to non-textual content;
Matrix builds module, for building behavioural matrix according to described visit data;
Matrix decomposition module, obtains user interest vector sum content type vector for described behavioural matrix being carried out matrix decomposition;
Computing module, for user described in content type vector calculation according to described user interest vector sum to the preference of described non-textual content;
Correlation recommendation module, for carrying out correlation recommendation according to described preference.
Preferably, described device also comprises:
Data processing module, for after the described visit data of collection, obtains the behavior related information set of user and non-textual content according to described visit data; Wherein said behavior related information is by the access behavior marking also normalization of described user to described non-textual content.
Preferably, described matrix decomposition module carries out described matrix decomposition by svd.
Preferably, described matrix decomposition module comprises:
Collaborative filtering module, carries out described svd for using collaborative filtering.
Preferably, described computing module comprises:
Preference computing module, for calculating the preference of user to non-textual content by calculating described user interest vector with the inner product of described content type vector.
Embodiments provide a kind of relevance recommend method and device, its technical scheme obtains regular extremely strong user interest vector sum content type vector by the matrix decomposition of user accesses data, thus comparatively specification and the calculating of rigorous preference can be carried out, obtain sake of clarity, associating between accurate, strict user with non-textual content, thus can recommend more accurately, promote user experience.
Accompanying drawing explanation
Fig. 1 is the basic procedure schematic diagram of relevance recommend method in one embodiment of the invention;
Fig. 2 is the modular structure schematic diagram of relevance recommendation apparatus in one embodiment of the invention.
Embodiment
For making the object, technical solutions and advantages of the present invention clearly understand, below in conjunction with embodiment also with reference to accompanying drawing, the present invention is described in more detail.Should be appreciated that, these describe just exemplary, and do not really want to limit the scope of the invention.In addition, in the following description, the description to known features and technology is eliminated, to avoid unnecessarily obscuring concept of the present invention.
Non-textual content is difficult to owing to lacking enough rule information analyze and associate, but often needs in practical application in numerous non-textual content, find that user may interested content recommend fast.Such as, in some application scenarios, the input of brand advertising, in order to cater to the hobby of user, wishes that the advertisement allowing this user see is the advertisement comparatively similar to its interest; Or in other scene, in ad click rate is estimated, the feature iting is desirable to find the fancy grade assessed value of active user and advertisement to be selected to be used as clicking rate to estimate.Therefore, expedite the emergence of out a kind of product demand, namely carry out content recommendation according to user preferences or find other contents higher concerning similarity user.But dependence prior art, directly find the computing method of user to the preference of non-textual content to be very difficult, is difficult to meet above-mentioned popularization demand.
The embodiment of the present invention proposes a kind of relevance recommend method based on matrix decomposition, utilizes user to calculate the preference of content to realize user the operation behavior matrix decomposition of non-textual content, thus realizes the extremely strong mixing recommendation of relevance.As shown in Figure 1, in embodiments of the present invention, relevance recommend method comprises step:
S1, gathers user to the visit data of non-textual content;
S2, builds behavioural matrix according to described visit data;
S3, carries out matrix decomposition and obtains user interest vector sum content type vector by described behavioural matrix;
S4, according to described user interest vector sum, user described in content type vector calculation is to the preference of described non-textual content;
S5, carries out correlation recommendation according to described preference.
Particularly, in a preferred embodiment of the invention, the visit data of step S1, by acquisition server background access log acquisition, also can utilize the mode such as plug-in unit, agency implanted in user side or intermediate equipment to gather.Preferably for advertisement, its technical scheme is described in the embodiment of the present invention, uses recent advertisement click logs and advertising display daily record data, integrate after first can carrying out data cleansing to both and add up.
In step S2, user builds behavioural matrix to the access behavior of advertisement to utilize visit data to determine.Suppose in visit data, to collect the access behavior of m user about n advertisement, build behavioural matrix A
m × n, wherein matrix element a
ijrepresent that i-th user gives a mark to the access behavior of a jth advertisement.Particularly, if user i clicks this advertisement j, then a
ijbe 1; If user i watches 1 time but do not click this advertisement j, then a
ijfor-1; If user i watches 2 times but do not click this advertisement j, then a
ijfor-2; If user i watches 3 times but do not click this advertisement j, then a
ijfor-3; If user i watches 4 times but do not click this advertisement j, then a
ijfor-4
Preferably, when daily record data is analyzed, identify and identify each user with the user cookie in daily record, identify and identify each advertisement with the advertisement id in daily record, then user associates with the behavior of advertisement and can be expressed as <cookie after extracting, id, the set of behavior marking >, wherein behavior marking is exactly above-mentioned matrix element a
ijvalue mode, more conveniently can obtain behavioural matrix quickly thus.
The embodiment of the present invention gives just to divide to effective click behavior, negative point that gives in various degree to frustrating behavior simultaneously, thus can analyze the fancy grade of user to advertisement more accurately, even can find detest degree further and shield.Certainly, in actual computation process, for ease of computer understanding and process, before calculating, normalization process can be done to score value, all be converted to nonnegative number by score value, such as by above-mentioned a
ijscore value is unified adds 4.Therefore, relevant technical staff in the field is appreciated that above-mentioned marking and account form not unique embodiment of the present invention, should not regard as the restriction specifically implemented the present invention.
In step S3, after acquisition user is to the behavioural matrix of advertisement, obtain user and advertisement vector representation separately by matrix decomposition.Carry out matrix decomposition preferably by svd (SVD, SingularValueDecomposition), above-mentioned behavioural matrix is broken down into
wherein k can be regarded as user interest or other classification quantity of commercial paper, usually chooses the value much smaller than m and n, such as k=10; Σ
k × kbe a diagonal matrix, what diagonal line stored is singular value orderly from big to small; U
m × kfor user's matrix;
for advertisement matrix.Further, i-th row vector u of matrix U
ifor the interest vector of user i represents, the column vector that vector length (i.e. interest dimension) is k, U is called as left singular vector; A jth row vector v of matrix V
jfor the categorization vector of advertisement j represents, the column vector that vector length (i.e. classification dimension) is k, V is called as right singular vector.
Preferably, collaborative filtering Collaborativefiltering is used to carry out svd to matrix.But Collaborativefiltering directly carries out mathematical SVD decomposition to very large matrix, but uses ALS algorithm to carry out iteration optimization object vector indirectly to carry out matrix decomposition.Can obtain the model result of Collaborativefiltering after decomposition, this result comprises two parts, and Part I is the vector representation of each user, i.e. the part of matrix U, and Part II is the vector representation of each advertisement, i.e. the part of matrix V.
In step S4, the calculating of preference can be carried out according to the user interest vector sum content type vector obtained.Particularly, the embodiment of the present invention utilizes user interest vector sum content type vector to calculate the preference of user to content, preferably calculates preference by compute vector inner product, by user interest vector u
idot product content (advertisement) categorization vector v
jwhat obtain is exactly the preference-score of user i to content (advertisement) j.In addition, the embodiment of the present invention also can utilize the vectorial similarity calculated between user of user interest, utilize the similarity that content type vector comes between Computed-torque control, preferably calculate similarity by the COS distance between compute vector or Euclidean distance; Similarity can be further used for deep correlation recommendation, and such as user, in viewing or when clicking a certain advertisement, shows the link of similar ad further.
In step S5, carry out correlation recommendation according to after preference sequence.The correlation recommendation of the embodiment of the present invention is preferably associating of user and advertisement, such as user open webpage/application, search or viewing Online Video time, recommend the advertisement that preference is high.The quantity of recommending can take default value or free setting, such as will carry out the recommendation of 4 advertisements in the stand-by period of user's Switch Video, can take out and recommend with the advertisement of this user preference degree Top4.In addition, correlation recommendation also can be advertisement and the associating of advertisement, and such as recommends when user plays/click a certain advertisement other advertisements that similarity is high.
As shown in Figure 2, the embodiment of the present invention also provides a kind of data characteristics formatting mechanism 1 simultaneously, comprising:
Data acquisition module 101, for gathering the visit data of user to non-textual content;
Matrix builds module 102, for building behavioural matrix according to described visit data;
Matrix decomposition module 103, obtains user interest vector sum content type vector for described behavioural matrix being carried out matrix decomposition;
Computing module 104, for user described in content type vector calculation according to described user interest vector sum to the preference of described non-textual content;
Correlation recommendation module 105, for carrying out correlation recommendation according to described preference.
Relevant technical staff in the field be appreciated that with said method correspondingly, also there is each functional module corresponding with various method steps in the device of the embodiment of the present invention, this is no longer going to repeat them simultaneously.In actual applications, said apparatus can be independently computing equipment, also can be the separate functional unit loaded by computing equipment, can also be computing equipment directly realize virtual/solid element.Equally, each module in device all can by being arranged in the central processor CPU of computing equipment, microprocessor MPU, the realization such as digital signal processor DSP or on-site programmable gate array FPGA, and the realization rate of said apparatus and module should not be considered as the restriction to the specific embodiment of the invention.
The specific implementation of some gordian technique of technical solution of the present invention is further illustrated below by a typical application scenarios.
In this scene, first the extraction advertisement click logs of two weeks in the past and advertising display daily record are as input.The MapReduce program of Hadoop is utilized to do Data Integration cleaning and statistical work.Map end is to the user cookie and the advertisement Id that take out every bar record, legitimate verification is carried out to both, do not abandoned by the record of checking, the output key of Map end is the combination of cookie and advertisement Id: cookie_ advertisement Id, whether export value is be click advertisement and identifier, if this record is click logs, be then designated 1, otherwise be designated 0.Reduce termination by from Map end with cookie_ advertisement Id for key, whether click the input that identification sets is combined into value, check and add up this mark set, mark is clicked if existed, be 5 then to this key marking, if non-number of clicks is 1, then marking is 4, non-number of clicks is 2, marking is 3, and non-number of clicks is 3, and marking is 2, non-number of clicks is more than 4 times, and marking is 1.Finally to export with cookie_ advertisement Id as key, give a mark as value data output on HDFS.
Matrix decomposition uses the collaborativefiltering in sparkmllib assembly to come.Every bar record input table is shown as a Rating object by the collaborativefiltering of Spark, Rating object by a user, a product, rating composition, advertisement Id i.e. product here, the marking of calculating i.e. rating here.After obtaining Rating set, set several important parameter, rank=10, i.e. k value, iterations=20, i.e. iterations, lambda=0.01, i.e. normalization factor, then uses ALS algorithm to complete decomposition with the form of iteration optimization to matrix, obtains last U matrix and V matrix saves as model.
Embodiments provide a kind of relevance recommend method and device, its technical scheme obtains regular extremely strong user interest vector sum content type vector by the matrix decomposition of user accesses data, thus comparatively specification and the calculating of rigorous preference can be carried out, obtain sake of clarity, associating between accurate, strict user with non-textual content, thus can recommend more accurately, promote user experience.
Should be understood that, above-mentioned embodiment of the present invention only for exemplary illustration or explain principle of the present invention, and is not construed as limiting the invention.Therefore, any amendment made when without departing from the spirit and scope of the present invention, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.In addition, claims of the present invention be intended to contain fall into claims scope and border or this scope and border equivalents in whole change and modification.
Claims (10)
1. a relevance recommend method, is characterized in that, described method comprises step:
Gather user to the visit data of non-textual content;
Behavioural matrix is built according to described visit data;
Described behavioural matrix is carried out matrix decomposition and obtain user interest vector sum content type vector;
According to described user interest vector sum, user described in content type vector calculation is to the preference of described non-textual content;
Correlation recommendation is carried out according to described preference.
2. method according to claim 1, is characterized in that, described method also comprises step:
After the described visit data of collection, obtain the behavior related information set of user and non-textual content according to described visit data; Wherein said behavior related information is by the access behavior marking also normalization of described user to described non-textual content.
3. method according to claim 1, is characterized in that, carries out described matrix decomposition by svd.
4. method according to claim 3, is characterized in that, uses collaborative filtering to carry out described svd.
5. method according to claim 1, is characterized in that, calculates the preference of user to non-textual content by calculating described user interest vector with the inner product of described content type vector.
6. a relevance recommendation apparatus, is characterized in that, described device comprises:
Data acquisition module, for gathering the visit data of user to non-textual content;
Matrix builds module, for building behavioural matrix according to described visit data;
Matrix decomposition module, obtains user interest vector sum content type vector for described behavioural matrix being carried out matrix decomposition;
Computing module, for user described in content type vector calculation according to described user interest vector sum to the preference of described non-textual content;
Correlation recommendation module, for carrying out correlation recommendation according to described preference.
7. device according to claim 6, is characterized in that, described device also comprises:
Data processing module, for after the described visit data of collection, obtains the behavior related information set of user and non-textual content according to described visit data; Wherein said behavior related information is by the access behavior marking also normalization of described user to described non-textual content.
8. device according to claim 6, is characterized in that, described matrix decomposition module carries out described matrix decomposition by svd.
9. device according to claim 8, is characterized in that, described matrix decomposition module comprises:
Collaborative filtering module, carries out described svd for using collaborative filtering.
10. device according to claim 6, is characterized in that, described computing module comprises:
Preference computing module, for calculating the preference of user to non-textual content by calculating described user interest vector with the inner product of described content type vector.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510661729.8A CN105302880A (en) | 2015-10-14 | 2015-10-14 | Content correlation recommendation method and apparatus |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510661729.8A CN105302880A (en) | 2015-10-14 | 2015-10-14 | Content correlation recommendation method and apparatus |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105302880A true CN105302880A (en) | 2016-02-03 |
Family
ID=55200150
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510661729.8A Pending CN105302880A (en) | 2015-10-14 | 2015-10-14 | Content correlation recommendation method and apparatus |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105302880A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106851396A (en) * | 2016-09-29 | 2017-06-13 | 创意引晴(开曼)控股有限公司 | To be inserted and player method in the automatic advertising that computer vision basis is realized |
| CN107220303A (en) * | 2017-05-10 | 2017-09-29 | 努比亚技术有限公司 | Recommendation method, device and the computer-readable medium of a kind of application |
| CN107748998A (en) * | 2017-10-13 | 2018-03-02 | 北京奇艺世纪科技有限公司 | The method and system that advertisement is recommended |
| CN107993084A (en) * | 2016-10-27 | 2018-05-04 | 北京酷我科技有限公司 | A kind of advertisement sending method |
| CN108388624A (en) * | 2018-02-12 | 2018-08-10 | 科大讯飞股份有限公司 | Multimedia messages recommend method and device |
| WO2019128311A1 (en) * | 2017-12-29 | 2019-07-04 | 广东神马搜索科技有限公司 | Advertisement similarity processing method and apparatus, calculation device, and storage medium |
| CN111078858A (en) * | 2018-10-19 | 2020-04-28 | 阿里巴巴集团控股有限公司 | Article searching method and device and electronic equipment |
| CN111724183A (en) * | 2019-03-20 | 2020-09-29 | ***股份有限公司 | Merchant recommendation method and merchant recommendation system |
| CN116089732A (en) * | 2023-04-11 | 2023-05-09 | 江西时刻互动科技股份有限公司 | User preference identification method and system based on advertisement click data |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060259344A1 (en) * | 2002-08-19 | 2006-11-16 | Choicestream, A Delaware Corporation | Statistical personalized recommendation system |
| CN101853463A (en) * | 2009-03-30 | 2010-10-06 | 北京邮电大学 | Collaborative filtering recommending method and system based on client characteristics |
| CN102289509A (en) * | 2011-08-31 | 2011-12-21 | 南京大学 | Method for obtaining, migrating and using website data |
| CN102306178A (en) * | 2011-08-25 | 2012-01-04 | 盛乐信息技术（上海）有限公司 | Video recommendation method and device |
| CN103034706A (en) * | 2012-12-07 | 2013-04-10 | 合一网络技术(北京)有限公司 | Video recommended list generating device and method based on information network |
| CN103246672A (en) * | 2012-02-09 | 2013-08-14 | 中国科学技术大学 | Method and device for performing personalized recommendation on users |
| CN103279552A (en) * | 2013-06-06 | 2013-09-04 | 浙江大学 | Collaborative filtering recommendation method based on user interest groups |
-
2015
- 2015-10-14 CN CN201510661729.8A patent/CN105302880A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060259344A1 (en) * | 2002-08-19 | 2006-11-16 | Choicestream, A Delaware Corporation | Statistical personalized recommendation system |
| CN101853463A (en) * | 2009-03-30 | 2010-10-06 | 北京邮电大学 | Collaborative filtering recommending method and system based on client characteristics |
| CN102306178A (en) * | 2011-08-25 | 2012-01-04 | 盛乐信息技术（上海）有限公司 | Video recommendation method and device |
| CN102289509A (en) * | 2011-08-31 | 2011-12-21 | 南京大学 | Method for obtaining, migrating and using website data |
| CN103246672A (en) * | 2012-02-09 | 2013-08-14 | 中国科学技术大学 | Method and device for performing personalized recommendation on users |
| CN103034706A (en) * | 2012-12-07 | 2013-04-10 | 合一网络技术(北京)有限公司 | Video recommended list generating device and method based on information network |
| CN103279552A (en) * | 2013-06-06 | 2013-09-04 | 浙江大学 | Collaborative filtering recommendation method based on user interest groups |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106851396A (en) * | 2016-09-29 | 2017-06-13 | 创意引晴(开曼)控股有限公司 | To be inserted and player method in the automatic advertising that computer vision basis is realized |
| CN107993084A (en) * | 2016-10-27 | 2018-05-04 | 北京酷我科技有限公司 | A kind of advertisement sending method |
| CN107220303A (en) * | 2017-05-10 | 2017-09-29 | 努比亚技术有限公司 | Recommendation method, device and the computer-readable medium of a kind of application |
| CN107748998A (en) * | 2017-10-13 | 2018-03-02 | 北京奇艺世纪科技有限公司 | The method and system that advertisement is recommended |
| WO2019128311A1 (en) * | 2017-12-29 | 2019-07-04 | 广东神马搜索科技有限公司 | Advertisement similarity processing method and apparatus, calculation device, and storage medium |
| CN108388624A (en) * | 2018-02-12 | 2018-08-10 | 科大讯飞股份有限公司 | Multimedia messages recommend method and device |
| CN111078858A (en) * | 2018-10-19 | 2020-04-28 | 阿里巴巴集团控股有限公司 | Article searching method and device and electronic equipment |
| CN111078858B (en) * | 2018-10-19 | 2023-06-09 | 阿里巴巴集团控股有限公司 | Article searching method and device and electronic equipment |
| CN111724183A (en) * | 2019-03-20 | 2020-09-29 | ***股份有限公司 | Merchant recommendation method and merchant recommendation system |
| CN116089732A (en) * | 2023-04-11 | 2023-05-09 | 江西时刻互动科技股份有限公司 | User preference identification method and system based on advertisement click data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105302880A (en) | Content correlation recommendation method and apparatus | |
| Karimi et al. | News recommender systems–Survey and roads ahead | |
| TWI601088B (en) | Topic management network public opinion evaluation management system and method | |
| US20190197416A1 (en) | Information recommendation method, apparatus, and server based on user data in an online forum | |
| Lipizzi et al. | Extracting and evaluating conversational patterns in social media: A socio-semantic analysis of customers’ reactions to the launch of new products using Twitter streams | |
| Barford et al. | Adscape: Harvesting and analyzing online display ads | |
| CN105247507B (en) | Method, system and storage medium for the influence power score for determining brand | |
| Patil et al. | Detecting experts on Quora: by their activity, quality of answers, linguistic characteristics and temporal behaviors | |
| TWI408560B (en) | A method, system and apparatus for recommending items or people of potential interest to users in a computer-based network | |
| US9032290B2 (en) | Tool for creating structured objects and actions on a social networking system | |
| CN110543598B (en) | Information recommendation method and device and terminal | |
| US9367603B2 (en) | Systems and methods for behavioral segmentation of users in a social data network | |
| US8793593B2 (en) | Integrating structured objects and actions generated on external systems into a social networking system | |
| US8909515B2 (en) | Dynamic sentence formation from structured objects and actions in a social networking system | |
| US11061990B2 (en) | Generating feedback for a target content item based on published content items | |
| CN105183925A (en) | Content association recommending method and content association recommending device | |
| US20150081725A1 (en) | System and method for actively obtaining social data | |
| US20140143329A1 (en) | Discovering signature of electronic social networks | |
| US20130073632A1 (en) | Structured objects and actions on a social networking system | |
| US20140278939A1 (en) | Advertisement extraction device and advertisement extraction method | |
| US20140095308A1 (en) | Advertisement distribution apparatus and advertisement distribution method | |
| CN107818166A (en) | A kind of information recommends method, apparatus, server and system | |
| US8965867B2 (en) | Measuring and altering topic influence on edited and unedited media | |
| CN112104642B (en) | Abnormal account number determination method and related device | |
| US9411860B2 (en) | Capturing intentions within online text |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160203 |
|
| RJ01 | Rejection of invention patent application after publication |