WO2018177303A1 - 媒体内容推荐方法、装置及存储介质 - Google Patents

媒体内容推荐方法、装置及存储介质 Download PDF

Info

Publication number
WO2018177303A1
WO2018177303A1 PCT/CN2018/080785 CN2018080785W WO2018177303A1 WO 2018177303 A1 WO2018177303 A1 WO 2018177303A1 CN 2018080785 W CN2018080785 W CN 2018080785W WO 2018177303 A1 WO2018177303 A1 WO 2018177303A1
Authority
WO
WIPO (PCT)
Prior art keywords
user
media content
trust
degree
users
Prior art date
Application number
PCT/CN2018/080785
Other languages
English (en)
French (fr)
Inventor
孟令民
Original Assignee
腾讯科技（深圳）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 腾讯科技（深圳）有限公司 filed Critical 腾讯科技（深圳）有限公司
Publication of WO2018177303A1 publication Critical patent/WO2018177303A1/zh
Priority to US15/929,122 priority Critical patent/US11182418B2/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
    • G06F16/40—Information retrieval; Database structures therefor; File system structures therefor of multimedia data, e.g. slideshows comprising image and additional audio data
    • G06F16/43—Querying
    • G06F16/435—Filtering based on additional data, e.g. user or group profiles
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • 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
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9535—Search customisation based on user profiles and personalisation

Definitions

  • the present application relates to the field of Internet technologies, and in particular, to a media content recommendation method, apparatus, and storage medium.
  • personalized recommendation systems have been widely used in online e-commerce platforms.
  • the so-called personalized recommendation is to recommend the information and products of interest to the user according to the user's interest characteristics and purchase behavior, thereby saving the time for the user to search for information or goods.
  • the application example provides a media content recommendation method, which is applied to an application server, and includes:
  • the application example further provides a media content recommendation device, including: a processor, a memory connected to the processor; and the memory readable instruction unit in the memory; the machine readable instruction unit includes:
  • a score record obtaining unit configured to obtain a score record of different users for different media content
  • a recommendation request receiving unit configured to receive a media content recommendation request of the first user among the plurality of users sent by the application client;
  • a trust determining unit configured to determine, according to the score record, a degree of trust of the first user to each of the plurality of users, wherein the first user has the trust degree of any second user Used to characterize the degree of recognition of the first user for the second user's rating of different media content;
  • a trust degree selection unit configured to select, in the determined trust degree of the first user to each second user, a plurality of trust degrees that reach a predetermined condition
  • a recommended media content determining unit configured to determine recommended media content provided to the first user from media content that is scored by the plurality of second users corresponding to the selected plurality of trust levels, and the determined recommended media content The link is sent to the application client.
  • the embodiment of the present application further provides a non-transitory computer readable storage medium, wherein the storage medium stores machine readable instructions, which are executable by a processor to perform the method described above .
  • the recommendation of the media content is more accurate.
  • FIG. 1 is a system architecture diagram related to an example of the present application
  • FIG. 2 is a flowchart of an example media content recommendation method of the present application
  • FIG. 3 is a schematic structural diagram of an example media content recommendation apparatus of the present application.
  • FIG. 4 is a schematic structural diagram of a computing device in an example of the present application.
  • the present application proposes an Internet-based media content recommendation method, which can be applied to the system architecture shown in FIG. 1.
  • the system architecture includes an application (APP) client 101, a media content push platform 102, and a media content provider client 106, which can communicate via the Internet 107, wherein the media content push platform 102 includes an application.
  • the server 103 the record database 104 in which the user accesses the media content, and the rating database 105 in which the user scores the media content.
  • the end user can access the application server 103 in the media content pushing platform 102 by using the application client 101, such as browsing a webpage or watching online video.
  • the application client 101 can report the user's access behavior to the application server 103, and the application server 103 saves the user's access behavior data in the user access record database 104.
  • the media content displayed on the application client 101 can also be scored.
  • a star rating option is displayed below the media content accessed by the user, and the user passes the The media content is scored by selecting a specific number of star ratings, and the application client obtains the corresponding rating of the media content by the user according to the 1-5 star rating given by the user to the media content.
  • the application client 101 can report the user's rating of the media content to the application server 103, and the application server 103 saves the user's rating of the media content in the rating database 105. While the application client 101 reports the user's access behavior and scoring behavior, the application client 101 can send a media content push request to the media content push platform 102, and the media content push platform 102 can match the media content push request.
  • the content is pushed to the application client 101.
  • the media content provider client 106 the media content provider can upload the material of the media content it wants to push to the media content push platform 102 to generate corresponding media content for push.
  • the system architecture shown in FIG. 1 may be a system architecture for implementing video recommendation.
  • the media content push platform 102 may be a video push platform, and the media content provider may be a video producer or an application client.
  • 101 is a video APP
  • the application server 103 is a video server.
  • the user views the video on the video APP
  • the video accessed by the user is reported to the video server, and the video server saves the video in the access record database 104.
  • the video APP scores the video of the user. It is reported to the video server, which saves it in the scoring database 105.
  • the video APP While the user is watching the video on the video APP, the video APP sends a video push request to the media content push platform 102, and the media content push platform 102 obtains the media content to be pushed according to some recommendation algorithms, and sends the link of the media content to the video APP.
  • the video app can obtain and display the video content corresponding to the selected video.
  • the recommended algorithm used is Collaborative Filtering (CF) recommendation algorithm.
  • CF Collaborative Filtering
  • the basic idea of CF is to recommend items to users according to their previous preferences and other users with similar interests.
  • Collaborative filtering is algorithmically divided into project-based collaborative filtering and user-based collaborative filtering.
  • Project-based collaborative filtering evaluates similarities between projects by rating users on different projects, and makes recommendations based on similarities between projects.
  • User-based collaborative filtering measures the similarity between users by scoring different items by users, and makes recommendations based on the similarity between users.
  • the user-based collaborative filtering algorithm is used to recommend the user, that is, the similarity between the users is evaluated by the scores of different users, and the recommendation is made based on the similarity between the users.
  • the degree of influence between the users calculated in this way is the same.
  • the similarity between the user A and the user B is calculated to be 0.7, and the influence of the user A on the user B and the user B to the user are performed when the recommendation is made.
  • the influence of A is the same, both are 0.7. This is obviously unscientific because the degree of influence and trust between people is asymmetric.
  • the present application provides a media content recommendation method, which is applicable to the application server 103 in the media content push platform 102. As shown in FIG. 2, the method includes the following steps:
  • Step 201 Acquire score records of different media content by multiple users.
  • the media content can be evaluated by scoring the accessed media content.
  • the application client uploads the user's rating of the media content to the application server 103, which saves it in the rating database 105.
  • the user's rating of the media content is shown in Table 1:
  • i 1 -i 10 represent media content
  • u 1 , u 2 , u 3 represent user u 1 , user u 2 , user u 3
  • W j, k represents the rating of the media content i k by the user u j .
  • the media content identified as 0 in Table 1 represents that the user has not scored the corresponding media content.
  • Step 202 Receive a media content recommendation request of the first user of the plurality of users sent by the application client.
  • the media content recommendation platform 102 sends a media content recommendation request.
  • the application client uploads the user's rating of the accessed media content to the application server 103 in the media content push platform 102.
  • a score record of the user's accessed media content is stored in the rating database 105.
  • the video server may recommend the movie to the user according to the history score of the user's viewed movie and other user's historical movie score.
  • the degree of support of the first user to the second user is calculated.
  • the degree of support indicates the extent to which users can support each other's scores, and the support between users is different.
  • the degree of support between users is established by the number of ratings of media content that are shared by users in common, and the number of media content that the user has scored. Because each user has different numbers of media content scores, and the media content that the users have scored with each other has the same number of media content, the constructed user support is asymmetric, that is, user A
  • the degree of support for user B is different from that for user B for user A.
  • the value of support is a value between [0, 1].
  • the degree of trust of the first user to the second user may be determined according to the degree of support and the similarity.
  • Step 203 Determine, according to the score record, a trust degree of the first user to each of the plurality of users, where the trust level of the first user to any second user is used to represent the The degree of recognition by the first user of the second user's rating of different media content.
  • the trust degree of the first user to the second user determining, according to the rating of the first user on the media content in the score record, and the rating of the second user on each media content, determining that the first user is The degree of trust of the second user, wherein the degree of trust is used to characterize the degree of recognition of the first user's rating of each media content by the second user. The more similar the score result of the second user to each media content is, the more the first user's rating of each media content is, the higher the trust of the first user to the second user, and vice versa, the lower the trust degree.
  • the degree of trust of the first user to the second user is related to the proportion of the number of media content that the first user and the second user score close to the number of media content scored by the first user. Because the number of media content scored by the first user may be different from the number of media content scored by the second user, the trust degree of the first user to the second user and the trust degree of the second user to the first user may be different. It is a good example of the asymmetry of trust between people in reality. In addition, the degree of trust of the first user to the second user is also related to the degree of similarity between the first user and the second user.
  • Step 204 Select, in the determined trust degree of the first user for each second user, a plurality of trust degrees that reach a predetermined condition.
  • determining, in the determined trust degree of the first user for each second user, a plurality of trust degrees that reach a predetermined condition Specifically, among the trustworthiness of the first user obtained in step 203 for each second user, a plurality of trust degrees that reach a predetermined condition are selected.
  • the predetermined condition may be that a threshold is set, and the degree of trust exceeding the threshold is considered to have reached a predetermined condition.
  • the predetermined condition may also be to sort the calculated trust levels from high to low, and select the first N trust degrees.
  • Step 205 Determine recommended media content provided to the first user from media content that is scored by the plurality of second users corresponding to the selected plurality of trust levels, and send the determined link of the recommended media content to the The application client.
  • the media content may be further filtered, and the filtered media content is used as the recommended media content of the first user, for example, A plurality of media content with a higher second user rating is used as the recommended media content.
  • the first user's trust degree to the second user is determined by the first user's score record of each media content and the score record of any second user for each media content.
  • the trust degree of the first user to the second user is different from the trust degree of the second user to the first user, which well reflects the asymmetry of trust between people in reality.
  • the recommended media content pushed to the first user is selected in the media content scored by the second user with high trust, thereby avoiding the symmetry of similarity when the traditional similarity recommendation is adopted. The equivalent effect of sex on the recommendation results.
  • the media content recommendation method provided by the present application can also solve the problem that the traditional user-based similarity recommendation algorithm is unscientific because of the sparsity of the score.
  • the similarity between the user u 1 and the user u 2 is greater than the similarity between the user u 2 and the user u 3 using the Pearson coefficient similarity calculation formula.
  • the user u 2 and the user u 3 has a relatively large number of media content items in the scores of the media content that are scored together; and user u 1 and user u 2 score a media content together, the scores are all 5, although the scores are the same, but in other There is no similar media content on the media content.
  • the similarity between user u 2 and user u 3 is more reasonable than the similarity between user u 1 and user u 2 .
  • the main reason for calculating large deviations is that the user u 1 and user u 2 have too few common scoring items, and the sparsity of the data makes the calculated similarity with great chance.
  • the recommendation is based on the trust degree between the users, wherein the first user has the first user and the second user scores close to the first user and the second user scores.
  • the proportion of the number of media content scored by the user is related, and the trust degree of the first user to the second user and the trust degree of the second user to the first user may be different, and thus the above problem can be avoided.
  • step 203 when determining the trust degree of the first user to each second user, the following steps are included: for any second user, the following operations are performed:
  • Step S11 Determine, according to the score record, a first item number of media content included in a first media content set in which the difference between the score value of the second user and the first user is within a predetermined range.
  • the score value within the predetermined range means that the absolute value of the difference between the score values of the two users for the same media content is less than a defined threshold, where the difference between the score values is not more than ⁇ .
  • a set of media contents in which the user u k and the user u j are commonly scored and the score value is within a predetermined range is represented by the following formula (1).
  • i m representative of media content i m, W k, m u k represents the user rating of the media content i m
  • W j, m represents the user u j i m of media content rating.
  • ⁇ is defined as 2, and in other instances it may be defined as other values.
  • the media content in which the user u 1 and the user u 2 score value are within a predetermined range is the media content i 1
  • the user u 2 and the user u 3 score values are in a predetermined range.
  • the media content within is media content i 1 , i 4 , i 5 , i 6 , i 7 .
  • Step S12 Determine, according to the score record, the second item number of the media content included in the second media content set scored by the first user.
  • Adopted in this application Indicates the potential of the second media content collection that the user u k scored, that is, the number of items of the scored media content. As also shown in Table 1, the number of media content items scored by user u 1 is 3, the number of media content items scored by user u 2 is 6, and the number of media content items scored by user u 3 is 6.
  • Step S13 Determine, according to the first item number and the second item number, the degree of support of the first user to the second user.
  • the ratio of the first item number to the second item number is used as the first user's support for the second user.
  • u k is the first user, that is, the target user to whom the media content is to be recommended, and u j is another user.
  • the potential of the second set of media content scored by the user u k that is, the number of items of the media content in the second set of media content.
  • the media content push platform 102 pushes the media content to the user u j
  • the user u j is the target user, that is, the first user, and at this time, the support degree of the user first user u j to other users u k needs to be calculated, and the following formula is adopted ( 3) indicates the support of user u j to user u k :
  • Step S14 Determine, according to the score record, a similarity degree between the second user and the first user.
  • the similarity is used to indicate the degree of similarity between users, where the rating similarity indicates how similar the user's rating to different media content is. Users will rate the media content that they have visited and are interested in. The higher the score, the higher the value of interest. The more media content items that are commonly scored by users, the closer the scores to the same media content, indicating that the user's interests are more similar.
  • Commonly used algorithmic formulas for calculating similarity are cosine similarity, Pearson's correlation, and modified cosine similarity (Adjusted Cosine).
  • Step S15 Calculate the trust degree of the first user to the second user according to the support degree and the score similarity.
  • the degree of trust of the first user with the second user and the degree of similarity between the first user and the second user are used to construct the trust degree of the first user to the second user. Since the similarity between users is the same and the support between users is different, the calculated trust between users is different. That is, the trust degree of the first user to the second user is different from the trust degree of the second user to the first user, which well reflects the asymmetry of trust between people in reality. Subsequently, the first user is recommended to the media content by the trust degree of the first user to the different second users, thereby avoiding the equivalent influence of the symmetry of the similarity on the recommendation result when the media content is recommended by the traditional similarity.
  • step S14 when performing the determining the similarity degree between the second user and the first user, the Pearson coefficient similarity is used to calculate the similarity between the first user and the second user.
  • Degree mainly includes the following steps:
  • Step S21 searching, in the score record, a third media content set that the second user and the first user have scored, and the second user and the first user separately for each media in the third media content set The rating value of the content.
  • the media content that is commonly scored by the user u 2 and the user u 3 , the user u 2 and the user u 3 includes i 1 , i 4 , i 5 , i 6 , i 7 , and the formed The three media content collections are ⁇ i 1 , i 4 , i 5 , i 6 , i 7 ⁇ .
  • the user u 2 is scored for each media content in the collection: 5, 5, 4, 5, 4, and the user u 3 scores each media content in the collection respectively. For: 4, 4, 4, 3, 4.
  • Step S22 Calculate, according to the score record, an average rating value of the first user for the scored media content and an average rating value of the second user for the scored media content.
  • the media content scored by the user u 2 is the media content i 1 , i 4 , i 5 , i 6 , i 7 , i 8 , respectively, and the score values are 5 , 5 , 4 , 5 , 4 , respectively. 5, so that the average score of the media content scored by the user u 2 is 4.67.
  • the media content scored by the user u 3 is the media content i 1 , i 2 , i 4 , i 5 , i 6 , i 7 , respectively, and the score values are 4 , 5 , 4 , 4 , 3 , 4 , respectively, thereby obtaining the user u. 3 rated media content has an average rating of 4.
  • Step S23 The rating value of each media content in the third media content set, the average rating value of the first user, and the average rating of the second user, respectively, according to the second user and the first user a value, calculating a similarity degree of the second user and the first user.
  • the similarity of the second user to the first user is calculated by the following formula (4):
  • Sim(u k , u j ) represents the similarity between the user u k and the user u j
  • I k,j represents a third media content set that the user u k and the user u j have scored together
  • i represents the a media content in the third media content set
  • w k,i represents a rating value of the user u k for the media content i
  • w j,i represents a rating value of the user u j for the media content i
  • the average rating of the rated media content on behalf of the user u k Represents the user u j 's average rating of the rated media content.
  • the first user's trust in the second user is calculated by the following formula (5):
  • the trust degree of the user u j to the user u k is calculated by the following formula (6):
  • step 205 when the recommended media content provided to the first user is determined from the media content that is scored by the plurality of second users corresponding to the selected plurality of trust levels, It mainly includes the following steps:
  • Step S31 Select, among the media content that is scored by the plurality of second users corresponding to the selected plurality of trust levels, the media content that the first user has not accessed, to form a fourth media content set.
  • the media content that is accessed by the first user recorded in the record database 104 is determined, and the set of media content that the first user has not visited in the media content that is scored by the plurality of second users is determined. (Fourth media content collection).
  • the second user selects the second user whose trust degree is relatively high, and recommends the selected media content that is of interest to the second user, that is, the scored media content to the first user.
  • the media content accessed by the first user is no longer recommended, and only the media content that the first user has not accessed is recommended.
  • Step S32 Searching, in the score record, a score value of each of the plurality of second media content sets by the plurality of second users.
  • a score value of each of the plurality of second users for each of the media content in the fourth media content set is searched for in the score record.
  • Step S33 Determine, according to the score values of the plurality of second users for the media content in the fourth media content set, the recommendability of each media content.
  • the recommendability may be An average score of the plurality of second users for the media content.
  • the plurality of second users may weight the scores of the media content to obtain a recommendability of the media content, and the weight of the scores may be obtained according to actual experience. Other calculation methods can also be used to obtain the recommendability.
  • Step S34 The media content whose recommendability is up to a predetermined condition is used as the recommended media content of the first user.
  • the predetermined condition may be that a threshold is set, and the recommendability exceeding the threshold is considered to have reached a predetermined condition.
  • the predetermined condition may also be to sort the calculated recommendability from high to low, and select the top M recommendable degrees. The media content corresponding to the selected recommendability is used as the recommended media content of the first user.
  • the media content recommendation method provided by the present application, a good experimental result is obtained by testing on a movie score data set (movielength).
  • the accuracy of the recommended movie is about 30%, which is much higher than the accuracy of about 7% recommended based on the similarity.
  • the application also provides a media content recommendation device 300, which is applied to the application server 103 in the media content delivery platform 102.
  • the device mainly includes:
  • a score record obtaining unit 301 configured to acquire score records of different media content by a plurality of users
  • the recommendation request receiving unit 302 is configured to receive a media content recommendation request of the first user of the plurality of users sent by the application client;
  • the trust degree determining unit 303 is configured to determine, according to the score record, the trust degree of the first user to each of the plurality of users, wherein the first user trusts the second user Degree is used to characterize the degree of recognition of the first user to the second user for different media content rating results;
  • the device also includes:
  • the trust degree selection unit 304 is configured to select, in the determined trust degree of the first user for each second user, a plurality of trust degrees that reach a predetermined condition;
  • a recommended media content determining unit 305 configured to determine recommended media content provided to the first user from media content that is scored by the plurality of second users corresponding to the selected plurality of trust degrees, and the determined recommended media A link to the content is sent to the application client.
  • the first user's trust score for the second user is determined by the first user's score record of each media content and the score record of any second user for each media content, wherein The trust degree of the first user to the second user is different from the trust degree of the second user to the first user, which well reflects the asymmetry of trust between people in reality.
  • the media content scored by the second user with high trust is selected as the recommended media content of the first user, thereby avoiding the symmetry of the similarity when the traditional similarity recommendation is adopted. The equivalent effect of the recommendation results.
  • the media content recommendation device provided by the present application can also solve the problem that the traditional user-based similarity recommendation algorithm is unscientific because of the sparsity of the score.
  • the similarity between the user u 1 and the user u 2 is greater than the similarity between the user u 2 and the user u 3 using the Pearson coefficient similarity calculation formula.
  • the user u 2 and the user u 3 has a relatively large number of media content items in the scores of the media content that are scored together; and user u 1 and user u 2 score a media content together, the scores are all 5, although the scores are the same, but in other There is no similar media content on the media content.
  • the similarity between user u 2 and user u 3 is more reasonable than the similarity between user u 1 and user u 2 .
  • the main reason for calculating large deviations is that the user u 1 and user u 2 have too few common scoring items, and the sparsity of the data makes the calculated similarity with great chance.
  • the recommendation is based on the trust degree between the users, wherein the first user has the first user and the second user scores close to the first user and the second user scores.
  • the proportion of the number of media content scored by the user is related, and the trust degree of the first user to the second user and the trust degree of the second user to the first user may be different, and thus the above problem can be avoided.
  • the trust determination unit 303 is configured to:
  • the trust degree determining unit 303 is further configured to:
  • the second user is similar to the rating of the first user.
  • the trust degree determining unit 303 calculates the score similarity of the second user and the first user by using formula (4):
  • Sim(u k , u j ) represents the similarity between the user u k and the user u j
  • I k,j represents a third media content set that the user u k and the user u j have scored together
  • i represents the a media content in the third media content set
  • w k,i represents a rating value of the user u k for the media content i
  • w j,i represents a rating value of the user u j for the media content i
  • the average rating of the rated media content on behalf of the user u k Represents the user u j 's average rating of the rated media content.
  • the trust degree determining unit 303 is further configured to:
  • the ratio of the first item number to the second item number is used as the first user's support for the second user.
  • the trust determination unit 303 calculates the trust of the first user to the second user by formula (5):
  • the recommended media content determining unit 305 is configured to:
  • the media content whose recommendability is up to a predetermined condition is used as the recommended media content of the first user.
  • the embodiment of the present application further provides a non-transitory computer readable storage medium, wherein the storage medium stores machine readable instructions, which are executable by a processor to perform the method described above .
  • FIG. 4 shows a compositional diagram of a computing device in which the media content recommendation device 300 is located.
  • the computing device includes one or more processors (CPUs) 402, communication modules 404, memory 406, user interface 410, and a communication bus 408 for interconnecting these components.
  • processors CPUs
  • communication modules 404 memory 406, user interface 410
  • communication bus 408 for interconnecting these components.
  • the processor 402 can receive and transmit data through the communication module 404 to effect network communication and/or local communication.
  • User interface 410 includes one or more output devices 412 that include one or more speakers and/or one or more visual displays.
  • User interface 410 also includes one or more input devices 414 including, for example, a keyboard, a mouse, a voice command input unit or loudspeaker, a touch screen display, a touch sensitive tablet, a gesture capture camera or other input button or control, and the like.
  • the memory 406 can be a high speed random access memory such as DRAM, SRAM, DDRRAM, or other random access solid state storage device; or a non-volatile memory such as one or more magnetic disk storage devices, optical disk storage devices, flash memory devices, or Other non-volatile solid-state storage devices.
  • a high speed random access memory such as DRAM, SRAM, DDRRAM, or other random access solid state storage device
  • non-volatile memory such as one or more magnetic disk storage devices, optical disk storage devices, flash memory devices, or Other non-volatile solid-state storage devices.
  • the memory 406 stores a set of instructions executable by the processor 402, the set of instructions stored in the memory being configured to be executed by the processor to implement the steps in the media content recommendation method of the present application described above, while also implementing the media content of the present application.
  • the function of each module in the device is recommended.
  • the memory 406 includes:
  • Operating system 416 including programs for processing various basic system services and for performing hardware related tasks
  • the application 418 includes various applications for media content recommendation, such an application being able to implement the processing flow in each of the above examples, such as may include some or all of the units or modules in the media content recommendation device 300. At least one of the units or modules in the media content recommendation device 300 may store machine executable instructions.
  • the processor 402 can implement the functions of at least one of the above-described units or modules by executing machine-executable instructions in at least one of the units or modules in the memory 406.
  • the hardware modules in the embodiments may be implemented in a hardware manner or a hardware platform plus software.
  • the above software includes machine readable instructions stored in a non-volatile storage medium.
  • embodiments can also be embodied as software products.
  • the hardware may be implemented by specialized hardware or hardware that executes machine readable instructions.
  • the hardware can be a specially designed permanent circuit or logic device (such as a dedicated processor such as an FPGA or ASIC) for performing a particular operation.
  • the hardware may also include programmable logic devices or circuits (such as including general purpose processors or other programmable processors) that are temporarily configured by software for performing particular operations.
  • each instance of the present application can be implemented by a data processing program executed by a data processing device such as a computer.
  • the data processing program constitutes the present application.
  • a data processing program usually stored in a storage medium is executed by directly reading a program out of a storage medium or by installing or copying the program to a storage device (such as a hard disk and or a memory) of the data processing device. Therefore, such a storage medium also constitutes the present application, and the present application also provides a non-volatile storage medium in which a data processing program is stored, which can be used to execute any of the above-mentioned method examples of the present application. An example.
  • the machine readable instructions corresponding to the modules of FIG. 4 may cause an operating system or the like operating on a computer to perform some or all of the operations described herein.
  • the non-transitory computer readable storage medium may be inserted into a memory provided in an expansion board within the computer or written to a memory provided in an expansion unit connected to the computer.
  • the CPU or the like installed on the expansion board or the expansion unit can perform part and all of the actual operations according to the instructions.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Computational Linguistics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)

Abstract

本申请公开了一种媒体内容推荐方法，包括：获取多个用户对各媒体内容的评分记录；接收应用客户端发送的第一用户的媒体内容推荐请求，根据所述评分记录确定该第一用户对所述多个用户中的各个第二用户的信任度，其中，该第一用户对任一第二用户的所述信任度用于表征该第一用户对于该第二用户对各媒体内容评分结果的认可程度；在确定的该第一用户对各第二用户的信任度中选取达到预定条件的多个信任度；从所述选取的多个信任度对应的多个第二用户评分过的媒体内容中确定提供给该第一用户的推荐媒体内容，将所述确定的推荐媒体内容的链接发送给所述应用客户端。本申请还提出了相应的媒体内容推荐装置及存储介质。

Description

媒体内容推荐方法、装置及存储介质
本申请要求于2017年3月28日提交中国专利局、申请号为201710191230.4、申请名称为“媒体内容推荐方法及装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及互联网技术领域，尤其涉及媒体内容推荐方法、装置及存储介质。
背景技术
随着互联网技术的发展，人们在互联网上进行的活动越来越多，不仅是简单的浏览网页，还可以在互联网上进行即时通讯、购物、广告宣传和网络游戏等。随着互联网技术的普及，个性化推荐系统已经广泛地应用于在线电子商务平台。所谓个性化推荐即根据用户的兴趣特点和购买行为，向用户推荐用户感兴趣的信息和商品，从而可节省用户搜索信息或商品的时间。
技术内容
本申请实例提供了一种媒体内容推荐方法，应用于应用服务器，包括：
获取多个用户对不同媒体内容的评分记录；
接收应用客户端发送的所述多个用户中第一用户的媒体内容推荐请求；
根据所述评分记录确定该第一用户对所述多个用户中的各个第二 用户的信任度，其中，该第一用户对任一第二用户的所述信任度用于表征该第一用户对于该第二用户对不同媒体内容评分结果的认可程度；
在确定的该第一用户对各第二用户的信任度中选取达到预定条件的多个信任度；
从所述选取的多个信任度对应的多个第二用户评分过的媒体内容中确定提供给该第一用户的推荐媒体内容，将所述确定的推荐媒体内容的链接发送给所述应用客户端。
本申请实例还提供了一种媒体内容推荐装置，包括：处理器，与所述处理器相连接的存储器；所述存储器中存储有机器可读指令单元；所述机器可读指令单元包括：
评分记录获取单元，用于获取多个用户对不同媒体内容的评分记录；
推荐请求接收单元，用于接收应用客户端发送的所述多个用户中第一用户的媒体内容推荐请求；
信任度确定单元，用于根据所述评分记录确定该第一用户对所述多个用户中的各个第二用户的信任度，其中，该第一用户对任一第二用户的所述信任度用于表征该第一用户对于该第二用户对不同媒体内容评分结果的认可程度；
信任度选取单元，用于在确定的该第一用户对各第二用户的信任度中选取达到预定条件的多个信任度；
推荐媒体内容确定单元，用于从所述选取的多个信任度对应的多个第二用户评分过的媒体内容中确定提供给该第一用户的推荐媒体内容，将所述确定的推荐媒体内容的链接发送给所述应用客户端。
本申请实施例还提供了一种非易失性计算机可读存储介质，其中所述存储介质中存储有机器可读指令，所述机器可读指令可以由处理器执 行以完成上述所述的方法。
采用本申请提供的上述方案，对媒体内容的推荐更加准确。
附图说明
为了更清楚地说明本申请实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1是是本申请实例涉及的系统构架图；
图2是本申请一实例媒体内容推荐方法的流程图；
图3是本申请一实例媒体内容推荐装置的结构示意图；以及
图4为本申请实例中的计算设备组成结构示意图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
本申请提出了一种基于互联网的媒体内容推荐方法，该方法可应用于图1所示的系统构架中。如图1所示，该系统构架包括：应用(APP)客户端101、媒体内容推送平台102和媒体内容提供方客户端106，这些实体可以通过互联网107进行通信，其中媒体内容推送平台102包括应用服务器103、用户访问媒体内容的记录数据库104以及用户对媒体 内容进行评分的评分数据库105。
终端用户可以使用应用客户端101访问媒体内容推送平台102中的应用服务器103，比如：浏览网页或者观看在线视频等。当用户使用应用客户端101访问应用服务器103时，应用客户端101可以将用户的访问行为上报给应用服务器103，应用服务器103将用户的访问行为数据保存在用户访问记录数据库104中。用户使用应用客户端101访问应用服务器103时，也可以对应用客户端101上展示的媒体内容进行评分，例如，在应用客户端上，用户访问的媒体内容下方展示有星级评分选项，用户通过选择具体的星级个数对媒体内容进行评分，应用客户端根据用户对该媒体内容给出的1-5星评价获得该用户对该媒体内容的相应评分。应用客户端101可以将用户对媒体内容的评分上报给应用服务器103，应用服务器103将用户对媒体内容的评分保存在评分数据库105中。应用客户端101在上报用户的访问行为及评分行为的同时，应用客户端101可以向媒体内容推送平台102发出媒体内容推送请求，媒体内容推送平台102可以将与该媒体内容推送请求相匹配的媒体内容推送给应用客户端101。通过媒体内容提供方客户端106，媒体内容提供方可以将其要推送的媒体内容的素材上传到媒体内容推送平台102，以生成相应的用于推送的媒体内容。
当上述媒体内容为视频时，图1所示的系统构架可以为实现视频推荐的系统构架，其中，媒体内容推送平台102可以为视频推送平台，媒体内容提供方可以为视频制作者，应用客户端101为视频APP，应用服务器103为视频服务器。用户在视频APP上观看视频时，将用户访问过的视频上报给视频服务器，视频服务器将其保存在访问记录数据库104中，当用户对观看的视频进行评分时，视频APP将用户对视频的评分上报给视频服务器，服务器将其保存在评分数据库105中。用户在视频 APP上观看视频的同时，视频APP向媒体内容推送平台102发送视频推送请求，媒体内容推送平台102根据一些推荐算法获得待推送的媒体内容，并将该媒体内容的链接发送给视频APP，视频APP能够获取到所选择的视频对应的视频内容并展示。
在一些实例中，采用的推荐算法是协同过滤(Collaborative Filtering，简称CF)推荐算法，CF的基本思想是根据用户之前的喜好以及其他兴趣相近的用户的选择来给用户推荐项目。协同过滤从算法上分为基于项目的协同过滤及基于用户的协同过滤。基于项目的协同过滤通过用户对不同项目的评分来评测项目之间的相似性，基于项目之间的相似性做出推荐。基于用户的协同过滤是通过不同用户对项目的评分来评测用户之间的相似性，基于用户之间的相似性做出推荐。
在一些实例中，采用基于用户的协同过滤算法对用户进行推荐，即通过不同用户对项目的评分来评测用户之间的相似性，基于用户之间的相似性做出推荐。
但这样计算出来的用户之间的影响程度是相同的，比如计算出来用户A和用户B之间的相似度为0.7，在进行推荐的时候，用户A对用户B的影响力和用户B对用户A的影响力是相同的，都是0.7。这显然是不科学的，因为人与人的之间影响力和信任程度是不对称的。
基于上述技术问题，本申请提出一种媒体内容推荐方法，该方法可应用于媒体内容推送平台102中的应用服务器103，如图2所示，该方法包括以下步骤：
步骤201：获取多个用户对不同媒体内容的评分记录。
用户在应用客户端101上访问媒体内容时，可以通过对访问的媒体内容进行打分来对该媒体内容进行评价。应用客户端将用户对媒体内容的打分上传到应用服务器103，应用服务器将其保存在评分数据库105 中。用户对媒体内容的评分如表1所示：
Figure PCTCN2018080785-appb-000001
表1
其中i 1-i 10代表媒体内容，u 1、u 2、u 3代表用户u 1、用户u 2、用户u 3，W j，k代表用户u j对媒体内容i k的评分。表1中标识为0的媒体内容代表用户对对应的媒体内容没有进行评分。
步骤202：接收应用客户端发送的所述多个用户中第一用户的媒体内容推荐请求。
用户在应用客户端101上访问媒体内容时，会向媒体内容推送平台102发送媒体内容推荐请求。在用户的历史访问媒体内容过程中，应用客户端将用户对访问过的媒体内容的评分上传给媒体内容推送平台102中的应用服务器103。在评分数据库105中保存有该用户对访问过的媒体内容的评分记录。例如对于电影媒体内容来说，当用户在视频APP上对观看过的影片进行评分时，视频APP将用户对影片的评分上传给视频服务器，视频服务器将其保存在评分数据库中，评分数据库中还保存有其他用户对影片的评分。当上述用户再次登录视频APP观看影片时，视频服务器可以根据该用户对观看过的影片的历史评分以及其他用户的历史影片评分向上述用户推荐影片。
对于所述评分记录中的多个用户中的任一第二用户，计算该第一用户对该第二用户的支持度。支持度表示用户间对评分相互能够支撑的程度，用户相互之间的支持度是不同的。通过用户相互之间共同的评分过的媒体内容中评分相近的个数，以及用户自身的评分过的媒体内容个数，构建用户之间的支持度。因为每个用户对媒体内容评分的个数不同， 而用户相互之间都评分过的媒体内容中评分相近的媒体内容个数相同，因此构建出来的用户支持度是相互不对称的，即用户A对用户B的支持度与用户B对用户A的支持度是不同的。支持度的取值是介于[0，1]之间的一个值。根据支持度与相似度可以确定第一用户对第二用户的信任度。
步骤203：根据所述评分记录确定该第一用户对所述多个用户中的各个第二用户的信任度，其中，该第一用户对任一第二用户的所述信任度用于表征该第一用户对于该第二用户对不同媒体内容评分结果的认可程度。
在确定第一用户对某一个第二用户的信任度时，根据所述评分记录中第一用户对各媒体内容的评分情况及该第二用户对各媒体内容的评分情况确定第一用户对该第二用户的信任度，其中信任度用以表征第一用户对于该第二用户对各媒体内容评分结果的认可程度。第二用户对各媒体内容的评分结果与第一用户对各媒体内容的评分结果越相似，第一用户对第二用户的信任度越高，反之，则信任度越低。第一用户对第二用户的信任度与第一用户与第二用户评分接近的媒体内容个数占第一用户评分过的媒体内容个数的比例有关。因为第一用户评分过的媒体内容个数与第二用户评分过的媒体内容个数可能不同，因而第一用户对第二用户的信任度与第二用户对第一用户的信任度可能不同，很好地体现了现实中人与人相互之间信任的不对称性。此外，第一用户对第二用户的信任度还与第一用户与第二用户之间的评分相似度有关。
步骤204：在确定的该第一用户对各第二用户的信任度中选取达到预定条件的多个信任度。
在确定得到的该第一用户对各第二用户的信任度中选取达到预定条件的多个信任度。具体地，在步骤203中得到的第一用户对各第二用 户的信任度中，选取达到预定条件的多个信任度。该预定条件可以是设置一阈值，超过该阈值的信任度认为是达到了预定条件。在一些实例中，该预定条件也可以是将计算得到的信任度从高到低进行排序，选取前N个信任度。
步骤205：从所述选取的多个信任度对应的多个第二用户评分过的媒体内容中确定提供给该第一用户的推荐媒体内容，将所述确定的推荐媒体内容的链接发送给所述应用客户端。
从所述选取的多个信任度对应的多个第二用户评分过的媒体内容中确定提供给该第一用户的推荐媒体内容，将所述确定的推荐媒体内容的链接发送给所述应用客户端。首先，确定选取的多个信任度对应的多个第二用户，进而将所述多个第二用户评分过的媒体内容作为该第一用户的推荐媒体内容，媒体内容推送平台102将确定的媒体内容的链接发送给第一用户所在客户端。在一些实例中，在上述确定多个第二用户评分过的媒体内容后，还可以对媒体内容进行再一步的过滤，过滤后的媒体内容作为第一用户的推荐媒体内容，例如，将所述多个第二用户评分比较高的媒体内容作为推荐媒体内容。
采用本申请提供的媒体内容推荐方法，通过第一用户对各媒体内容的评分记录及任一第二用户对各媒体内容的评分记录，确定该第一用户对该第二用户的信任度。其中，第一用户对第二用户的信任度与第二用户对第一用户的信任度不同，很好地体现了现实中人与人相互之间信任的不对称性。根据第一用户对各第二用户的信任度，在信任度高的第二用户评分过的媒体内容中选取向第一用户推送的推荐媒体内容，避免了采用传统相似度推荐时，相似度的对称性对推荐结果的等同的影响。
此外，采用本申请提供的媒体内容推荐方法，还能解决传统的基于用户的相似度的推荐算法由于评分的稀疏性造成的相似度计算不科学的 问题。如上述表1中的示例，利用皮尔森系数相似性计算公式得到用户u 1和用户u 2的相似度要大于用户u 2和用户u 3的相似度，由上表可见，用户u 2和用户u 3在共同评分的媒体内容的评分值接近的媒体内容项数比较多；而用户u 1和用户u 2共同评分了一项媒体内容，分值都是5，虽然评分值相同，但是在其他媒体内容上却没有评分相近的媒体内容。显然，用户u 2和用户u 3的相似度大于用户u 1和用户u 2的相似度更合理些。计算出现较大偏差的主要原因在于用户u 1和用户u 2的共同评分项目太少，数据的稀疏性会使得计算出的相似度带有很大的偶然性。而采用本申请提供的媒体内容推荐方法，基于用户之间的信任度进行推荐，其中第一用户对第二用户的信任度与第一用户与第二用户评分接近的媒体内容个数占第一用户评分过的媒体内容个数的比例有关，并且第一用户对第二用户的信任度与第二用户对第一用户的信任度可能不同，因而能够避免上述问题。
在一些实例中，在上述步骤203中，在执行确定该第一用户对各个第二用户的信任度时，包括以下步骤：对于任一第二用户，执行以下操作：
步骤S11：根据所述评分记录，确定该第二用户与该第一用户的评分值之差在预定范围内的第一媒体内容集合中包含的媒体内容的第一项数。
本申请中，采用
Figure PCTCN2018080785-appb-000002
表示用户u k和用户u j共同评分并且评分值在预定范围内的媒体内容的集合。评分值在预定范围内是指两个用户对同一个媒体内容的评分值的差值的绝对值小于一个定义的阈值，这里使用评分值的差值不超过δ来表示。采用以下公式(1)表示用户u k和用户u j共同评分并且评分值在预定范围内的媒体内容的集合。
Figure PCTCN2018080785-appb-000003
其中，i m代表媒体内容i m，W k，m代表用户u k对媒体内容i m的评分，W j，m代表用户u j对媒体内容i m的评分。在一些实例中δ定义为2，在其他实例中还可以定义为其他值。例如，在δ定义为2的情况下，在表1中，用户u 1与用户u 2评分值在预定范围内的媒体内容为媒体内容i 1，用户u 2与用户u 3评分值在预定范围内的媒体内容为媒体内容i 1、i 4、i 5、i 6、i 7。获得用户评分值在预定范围内的第一媒体内容集合后，进而确定该第一媒体内容集合中的媒体内容的项数。
步骤S12：根据所述评分记录，确定该第一用户评分过的第二媒体内容集合中包含的媒体内容的第二项数。
在本申请中采用
Figure PCTCN2018080785-appb-000004
表示用户u k评分过的第二媒体内容集合的势，即评分过的媒体内容的项数。还如在表1中，用户u 1评分过的媒体内容项数为3，用户u 2评分过的媒体内容项数为6，用户u 3评分过的媒体内容项数为6。
步骤S13：根据所述第一项数与所述第二项数确定该第一用户对该第二用户的支持度。
在一些实例中，将所述第一项数与所述第二项数的比值作为该第一用户对该第二用户的支持度。
采用以下公式(2)表示用户u k对用户u j的支持度Sup(u k,u j)
Figure PCTCN2018080785-appb-000005
此时，u k为上述第一用户，即要向其推荐媒体内容时的目标用户，u j 为其他用户，
Figure PCTCN2018080785-appb-000006
表示用户u k与用户u j的评分值在预定范围内的第一媒体内容集合中媒体内容的项数。
Figure PCTCN2018080785-appb-000007
表示用户u k评分过的第二媒体内容集合的势，即该第二媒体内容集合中媒体内容的项数。
媒体内容推送平台102当向用户u j推送媒体内容时，用户u j为目标用户，即第一用户，此时需要计算用户第一用户u j对其他用户u k的支持度，采用以下公式(3)表示用户u j对用户u k的支持度：
Figure PCTCN2018080785-appb-000008
对比公式(2)及公式(3)会发现，由于分子相同但是分母不同，因此用户u k对用户u j的支持度与用户u j对用户u k的支持度不同。因而得到的用户之间的支持度矩阵不像用户间的相似度矩阵一样是对阵矩阵了，支持度矩阵是非对称的。
步骤S14：根据所述评分记录，确定该第二用户与该第一用户的评分相似度。
相似度用于表示用户之间的相似程度，在这里评分相似度表示用户对不同媒体内容的评分的相似程度。用户会对访问过的、感兴趣的媒体内容进行评分，评分越高，兴趣值越高。用户之间共同评分过的媒体内容项数越多、对同一媒体内容的评分值越接近，表示用户的兴趣越相似。常用的计算相似度的算法公式有余弦(Cosine)相似性、皮尔森系数相似性(Pearson’s correlation)以及修正的余弦相似性(Adjusted Cosine)。
步骤S15：根据所述支持度及所述评分相似度计算该第一用户对该第二用户的信任度。
采用上述所述的第一用户对第二用户的支持度，以及第一用户与第二用户之间的评分相似度来构建第一用户对第二用户的信任度。由于用 户之间的相似度相同，而用户之间支持度不同，因此计算出来的用户相互之间的信任度也是不同的。即第一用户对第二用户的信任度与第二用户对第一用户的信任度不同，很好的体现了现实中人与人相互之间信任的不对称性。后续通过第一用户对不同第二用户的信任度来对第一用户进行推荐媒体内容，避免了采用传统相似度推荐媒体内容时，相似度的对称性对推荐结果的等同的影响。
在一些实例中，在上述步骤S14中，在执行所述确定该第二用户与该第一用户的评分相似度时，采用皮尔森系数相似性计算第一用户与第二用户之间的评分相似度，主要包括以下步骤：
步骤S21：在所述评分记录中查找该第二用户与该第一用户都评分过的第三媒体内容集合以及该第二用户及该第一用户分别对所述第三媒体内容集合中各媒体内容的评分值。
还如表1中的示例所示，对应用户u 2及用户u 3，用户u 2及用户u 3共同评分的媒体内容包括i 1、i 4、i 5、i 6、i 7，形成的第三媒体内容集合为{i 1、i 4、i 5、i 6、i 7}。同时在表1中查找用户u 2对该集合中的每一项媒体内容的评分分别为：5、5、4、5、4，用户u 3对该集合中的每一项媒体内容的评分分别为：4、4、4、3、4。
步骤S22：根据所述评分记录计算该第一用户对评分过的媒体内容的平均评分值以及该第二用户对评分过的媒体内容的平均评分值。
还如上例所示，用户u 2评分过的媒体内容分别为媒体内容i 1、i 4、i 5、i 6、i 7、i 8，评分值分别为5、5、4、5、4、5，从而得到用户u 2评分过的媒体内容的平均评分值为4.67。用户u 3评分过的媒体内容分别为媒体内容i 1、i 2、i 4、i 5、i 6、i 7，评分值分别为4、5、4、4、3、4，从而得到用户u 3评分过的媒体内容的平均评分值为4。
步骤S23：根据该第二用户及该第一用户分别对所述第三媒体内容 集合中各媒体内容的评分值、该第一用户的所述平均评分值及该第二用户的所述平均评分值，计算该第二用户与该第一用户的评分相似度。
根据在步骤S21中得到的第二用户及第一用户对都评分过的第三媒体内容集合中各媒体内容的评分值，以及步骤S22中第一用户对评分过的媒体内容的平均评分值以及该第二用户对评分过的媒体内容的平均评分值计算第一用户与第二用户之间的评分相似度。
在一些实例中，通过以下公式(4)计算该第二用户与该第一用户的评分相似度：
Figure PCTCN2018080785-appb-000009
其中，Sim(u k，u j)代表用户u k及用户u j之间的评分相似度，I k，j代表用户u k及用户u j共同评分过的第三媒体内容集合，i代表所述第三媒体内容集合中的一项媒体内容，w k，i代表用户u k对媒体内容i的评分值，w j，i代表用户u j对媒体内容i的评分值，
Figure PCTCN2018080785-appb-000010
代表用户u k对评分过的媒体内容的平均评分，
Figure PCTCN2018080785-appb-000011
代表用户u j对评分过的媒体内容的平均评分。
在一些实例中，通过以下公式(5)计算该第一用户对该第二用户的信任度：
Figure PCTCN2018080785-appb-000012
其中Sup(u k,u j)是用户u k对用户u j的支持度，计算方式参加公式(2)，而Sim(u k,u j)是用户u k和用户u j之间的评分相似度，计算方式参见公式(4)。信任度的取值是介于0到1之间的一个值。在这里用户u k是第一用户，用户u j为第二用户。
当对用户u j推荐媒体内容时，用户u j为第一用户，用户u k是第二用户，那么用户u j对用户u k的信任度通过以下公式(6)计算：
Figure PCTCN2018080785-appb-000013
其中，Sup(u j,u k)是用户u j对用户u k的支持度，计算方式参加公式(3)，而Sim(u j,u k)与Sim(u k,u j)相同，是用户u k和用户u j之间的评分相似度，计算方式参见公式(4)。
对比公式(5)和公式(6)，由于Sup(u k,u j)和Sup(u j,u k)不一定相等，所以用户u k对用户u j的信任度和用户u j对用户u k的信任度也不一定相等。也就是说，对计算信任度的两个用户之间，两个用户的信任度是不一样的。根据公式(5)或(6)得到的用户信任度矩阵比用户相似度矩阵更贴近客观现实。
在一些实例中，在上述步骤205中，在执行所述从所述选取的多个信任度对应的多个第二用户评分过的媒体内容中确定提供给该第一用户的推荐媒体内容时，主要包括以下步骤：
步骤S31：在所述选取的多个信任度对应的多个第二用户评分过的媒体内容中，选取所述第一用户没有访问过的媒体内容，形成第四媒体内容集合。
根据确定的多个第二用户，确定该多个第二用户评分过的媒体内容，同时在所述媒体内容中选取第一用户没有访问过的媒体内容形成第四媒体内容集合。在这里可以根据用户访问记录数据库104中记录的第一用户访问过的媒体内容，确定所述多个第二用户评分过的所述媒体内容中所述第一用户没有访问过的媒体内容的集合(第四媒体内容集合)。所述第二用户为选取的所述第一用户对其信任度比较高的第二用户，将 该选取的第二用户感兴趣的媒体内容，即评分过的媒体内容推荐给该第一用户。同时在向第一用户推荐媒体内容时，第一用户访问过的媒体内容不再推荐，只推荐第一用户没有访问过的媒体内容。
步骤S32：在所述评分记录中查找所述多个第二用户分别对所述第四媒体内容集合中每一项媒体内容的评分值。
对于上步骤中确定的媒体内容集合，在评分记录中查找所述多个第二用户中每一个第二用户分别对所述第四媒体内容集合中每一项媒体内容的评分值。
步骤S33：根据所述多个第二用户对所述第四媒体内容集合中各项媒体内容的评分值，确定各项媒体内容的可推荐度。
对于所述第四媒体内容集合中任一项媒体内容，根据所述多个第二用户对该媒体内容的评分值确定该媒体内容的可推荐度，在一些实例中该可推荐度可以为所述多个第二用户对该媒体内容的平均评分。在另一些实例中，可以将所述多个第二用户对该媒体内容的评分加权相加后得到该媒体内容的可推荐度，评分的权重可以根据实际经验得到。还可以采用其他的计算方式获得可推荐度。
步骤S34：将所述可推荐度达到预定条件的媒体内容作为所述第一用户的推荐媒体内容。
该预定条件可以是设置一阈值，超过该阈值的可推荐度认为是达到了预定条件。在一些实例中，该预定条件也可以是将计算得到的可推荐度从高到低进行排序，选取前M个可推荐度。将选取的可推荐度对应的媒体内容作为所述第一用户的推荐媒体内容。
采用本申请提供的媒体内容推荐方法，通过在电影评分数据集(movielength)上测试，取到了很好的实验结果。根据准确率评价标准，推荐的电影的准确率大概在30％左右，远高于基于相似度进行推荐的 7％左右的准确率。
本申请还提供了一种媒体内容推荐装置300，应用于媒体内容推送平台102中的应用服务器103，如图3所示，该装置主要包括：
评分记录获取单元301，用于获取多个用户对不同媒体内容的评分记录；
推荐请求接收单元302，用于接收应用客户端发送的所述多个用户中第一用户的媒体内容推荐请求；
信任度确定单元303，用于根据所述评分记录确定该第一用户对所述多个用户中的各个第二用户的信任度，其中，该第一用户对任一第二用户的所述信任度用于表征该第一用户对于该第二用户对不同媒体内容评分结果的认可程度；
所述装置还包括：
信任度选取单元304，用于在确定的该第一用户对各第二用户的信任度中选取达到预定条件的多个信任度；
推荐媒体内容确定单元305，用于从所述选取的多个信任度对应的多个第二用户评分过的媒体内容中确定提供给该第一用户的推荐媒体内容，将所述确定的推荐媒体内容的链接发送给所述应用客户端。
采用本申请提供的媒体内容推荐装置，通过第一用户对各媒体内容的评分记录及任一第二用户对各媒体内容的评分记录，确定该第一用户对该第二用户的信任度，其中，第一用户对第二用户的信任度与第二用户对第一用户的信任度不同，很好地体现了现实中人与人相互之间信任的不对称性。根据第一用户对各第二用户的信任度，选取信任度高的第二用户评分过的媒体内容作为第一用户的推荐媒体内容，避免了采用传统相似度推荐时，相似度的对称性对推荐结果的等同的影响。
此外，采用本申请提供的媒体内容推荐装置，还能解决传统的基于用 户的相似度的推荐算法由于评分的稀疏性造成的相似度计算不科学的问题。如上述表1中的示例，利用皮尔森系数相似性计算公式得到用户u 1和用户u 2的相似度要大于用户u 2和用户u 3的相似度，由上表可见，用户u 2和用户u 3在共同评分的媒体内容的评分值接近的媒体内容项数比较多；而用户u 1和用户u 2共同评分了一项媒体内容，分值都是5，虽然评分值相同，但是在其他媒体内容上却没有评分相近的媒体内容。显然，用户u 2和用户u 3的相似度大于用户u 1和用户u 2的相似度更合理些。计算出现较大偏差的主要原因在于用户u 1和用户u 2的共同评分项目太少，数据的稀疏性会使得计算出的相似度带有很大的偶然性。而采用本申请提供的媒体内容推荐方法，基于用户之间的信任度进行推荐，其中第一用户对第二用户的信任度与第一用户与第二用户评分接近的媒体内容个数占第一用户评分过的媒体内容个数的比例有关，并且第一用户对第二用户的信任度与第二用户对第一用户的信任度可能不同，因而能够避免上述问题。
在一些实例中，所述信任度确定单元303，用于：
对于任一第二用户，执行以下操作：
根据所述评分记录，确定该第二用户与该第一用户的评分值之差在预定范围内的第一媒体内容集合中包含的媒体内容的第一项数；
根据所述评分记录，确定该第一用户评分过的第二媒体内容集合中包含的媒体内容的第二项数；
根据所述第一项数与所述第二项数确定该第一用户对该第二用户的支持度；
根据所述评分记录，确定该第二用户与该第一用户的评分相似度；及
根据所述支持度及所述评分相似度计算该第一用户对该第二用户的信任度。
在一些实例中，所述信任度确定单元303还用于：
在所述评分记录中查找该第二用户与该第一用户都评分过的第三媒体内容集合，以及该第二用户及该第一用户分别对所述第三媒体内容集合中各媒体内容的评分值；
根据所述评分记录计算该第一用户对评分过的媒体内容的平均评分值以及该第二用户对评分过的媒体内容的平均评分值；
根据该第二用户及该第一用户分别对所述第三媒体内容集合中各媒体内容的评分值、该第一用户的所述平均评分值及该第二用户的所述平均评分值，计算该第二用户与该第一用户的评分相似度。
在一些实例中，所述信任度确定单元303通过公式(4)计算该第二用户与该第一用户的评分相似度：
Figure PCTCN2018080785-appb-000014
其中，Sim(u k，u j)代表用户u k及用户u j之间的评分相似度，I k，j代表用户u k及用户u j共同评分过的第三媒体内容集合，i代表所述第三媒体内容集合中的一项媒体内容，w k，i代表用户u k对媒体内容i的评分值，w j，i代表用户u j对媒体内容i的评分值，
Figure PCTCN2018080785-appb-000015
代表用户u k对评分过的媒体内容的平均评分，
Figure PCTCN2018080785-appb-000016
代表用户u j对评分过的媒体内容的平均评分。
在一些实例中，所述信任度确定单元303，还用于：
将所述第一项数与所述第二项数的比值作为该第一用户对该第二用户的支持度。
在一些实例中，所述信任度确定单元303通过公式(5)计算该第 一用户对该第二用户的信任度：
Figure PCTCN2018080785-appb-000017
其中Sup(u k,u j)是用户u k对用户u j的支持度，Sim(u k,u j)是用户u k和用户u j之间的评分相似度，用户u k是第一用户，用户u j为第二用户。
在一些实例中，所述推荐媒体内容确定单元305，用于：
在所述选取的多个信任度对应的多个第二用户评分过的媒体内容中，选取所述第一用户没有访问过的第四媒体内容集合；
在所述评分记录中查找所述多个第二用户分别对所述第四媒体内容集合中每一项媒体内容的评分值；
根据所述多个第二用户对所述第四媒体内容集合中各项媒体内容的评分值，确定各项媒体内容的可推荐度；
将所述可推荐度达到预定条件的媒体内容作为所述第一用户的推荐媒体内容。
本申请实施例还提供了一种非易失性计算机可读存储介质，其中所述存储介质中存储有机器可读指令，所述机器可读指令可以由处理器执行以完成上述所述的方法。
图4示出了媒体内容推荐装置300所在的计算设备的组成结构图。如图4所示，该计算设备包括一个或者多个处理器(CPU)402、通信模块404、存储器406、用户接口410，以及用于互联这些组件的通信总线408。
处理器402可通过通信模块404接收和发送数据以实现网络通信和/或本地通信。
用户接口410包括一个或多个输出设备412，其包括一个或多个扬 声器和/或一个或多个可视化显示器。用户接口410也包括一个或多个输入设备414，其包括诸如，键盘，鼠标，声音命令输入单元或扩音器，触屏显示器，触敏输入板，姿势捕获摄像机或其他输入按钮或控件等。
存储器406可以是高速随机存取存储器，诸如DRAM、SRAM、DDRRAM、或其他随机存取固态存储设备；或者非易失性存储器，诸如一个或多个磁盘存储设备、光盘存储设备、闪存设备，或其他非易失性固态存储设备。
存储器406存储处理器402可执行的指令集，存储器中存储的指令集经配置以由处理器执行，以实现上述本申请中的媒体内容推荐方法中的各步骤，同时还实现本申请的媒体内容推荐装置中各模块的功能。
存储器406包括：
操作系统416，包括用于处理各种基本系统服务和用于执行硬件相关任务的程序；
应用418，包括用于媒体内容推荐的各种应用程序，这种应用程序能够实现上述各实例中的处理流程，比如可以包括媒体内容推荐装置300中的部分或全部单元或者模块。媒体内容推荐装置300中的各单元或模块中的至少一个单元或模块可以存储有机器可执行指令。处理器402通过执行存储器406中各单元或模块中至少一个单元或模块中的机器可执行指令，进而能够实现上述各单元或模块中的至少一个模块的功能。
需要说明的是，上述各流程和各结构图中不是所有的步骤和模块都是必须的，可以根据实际的需要忽略某些步骤或模块。各步骤的执行顺序不是固定的，可以根据需要进行调整。各模块的划分仅仅是为了便于描述采用的功能上的划分，实际实现时，一个模块可以分由多个模块实现，多个模块的功能也可以由同一个模块实现，这些模块可以位于同一 个设备中，也可以位于不同的设备中。
各实施例中的硬件模块可以以硬件方式或硬件平台加软件的方式实现。上述软件包括机器可读指令，存储在非易失性存储介质中。因此，各实施例也可以体现为软件产品。
各例中，硬件可以由专门的硬件或执行机器可读指令的硬件实现。例如，硬件可以为专门设计的永久性电路或逻辑器件(如专用处理器，如FPGA或ASIC)用于完成特定的操作。硬件也可以包括由软件临时配置的可编程逻辑器件或电路(如包括通用处理器或其它可编程处理器)用于执行特定操作。
另外，本申请的每个实例可以通过由数据处理设备如计算机执行的数据处理程序来实现。显然，数据处理程序构成了本申请。此外，通常存储在一个存储介质中的数据处理程序通过直接将程序读取出存储介质或者通过将程序安装或复制到数据处理设备的存储设备(如硬盘和或内存)中执行。因此，这样的存储介质也构成了本申请，本申请还提供了一种非易失性存储介质，其中存储有数据处理程序，这种数据处理程序可用于执行本申请上述方法实例中的任何一种实例。
图4模块对应的机器可读指令可以使计算机上操作的操作系统等来完成这里描述的部分或者全部操作。非易失性计算机可读存储介质可以是插入计算机内的扩展板中所设置的存储器中或者写到与计算机相连接的扩展单元中设置的存储器。安装在扩展板或者扩展单元上的CPU等可以根据指令执行部分和全部实际操作。
以上所述仅为本发明实施例的较佳实施例而已，并不用以限制本发明实施例，凡在本发明实施例的精神和原则之内，所做的任何修改、等同替换、改进等，均应包含在本发明实施例保护的范围之内。

Claims (15)

  1. 一种媒体内容推荐方法，应用于应用服务器，包括：
    获取多个用户对不同媒体内容的评分记录；
    接收应用客户端发送的所述多个用户中第一用户的媒体内容推荐请求；
    根据所述评分记录确定该第一用户对所述多个用户中的各个第二用户的信任度，其中，该第一用户对任一第二用户的所述信任度用于表征该第一用户对于该第二用户对不同媒体内容评分结果的认可程度；
    在确定的该第一用户对各第二用户的信任度中选取达到预定条件的多个信任度；
    从所述选取的多个信任度对应的多个第二用户评分过的媒体内容中确定提供给该第一用户的推荐媒体内容，将所述确定的推荐媒体内容的链接发送给所述应用客户端。
  2. 根据权利要求1所述的方法，其中，确定该第一用户对各个第二用户的信任度包括：
    对于任一第二用户，执行以下操作：
    根据所述评分记录，确定该第二用户与该第一用户的评分值之差在预定范围内的第一媒体内容集合中包含的媒体内容的第一项数；
    根据所述评分记录，确定该第一用户评分过的第二媒体内容集合中包含的媒体内容的第二项数；
    根据所述第一项数与所述第二项数确定该第一用户对该第二用户的支持度；
    根据所述评分记录，确定该第二用户与该第一用户的评分相似度；及
    根据所述支持度及所述评分相似度计算该第一用户对该第二用户的信任度。
  3. 根据权利要求2所述的方法，其中，所述确定该第二用户与该第一用户的评分相似度包括：
    在所述评分记录中查找该第二用户与该第一用户都评分过的第三媒体内容集合，以及该第二用户及该第一用户分别对所述第三媒体内容集合中各媒体内容的评分值；
    根据所述评分记录计算该第一用户对评分过的媒体内容的平均评分值以及该第二用户对评分过的媒体内容的平均评分值；
    根据该第二用户及该第一用户分别对所述第三媒体内容集合中各媒体内容的评分值、该第一用户的所述平均评分值及该第二用户的所述平均评分值，计算该第二用户与该第一用户的评分相似度。
  4. 根据权利要求3所述的方法，其中，通过以下公式(1)计算该第二用户与该第一用户的评分相似度：
    Figure PCTCN2018080785-appb-100001
    其中，Sim(u k，u j)代表用户u k及用户u j之间的评分相似度，I k，j代表用户u k及用户u j共同评分过的第三媒体内容集合，i代表所述第三媒体内容集合中的一项媒体内容，w k，i代表用户u k对媒体内容i的评分值，w j，i代表用户u j对媒体内容i的评分值，
    Figure PCTCN2018080785-appb-100002
    代表用户u k对评分过的媒体内容的平均评分，
    Figure PCTCN2018080785-appb-100003
    代表用户u j对评分过的媒体内容的平均评分。
  5. 根据权利要求2所述的方法，其中，所述确定该第一用户对该第二用户的支持度包括：将所述第一项数与所述第二项数的比值作为该第一用户对该第二用户的支持度。
  6. 根据权利要求2所述的方法，其中，通过以下公式(2)计算该第一用户对该第二用户的信任度：
    Figure PCTCN2018080785-appb-100004
    其中Sup(u k,u j)是用户u k对用户u j的支持度，Sim(u k,u j)是用户u k和用户u j之间的评分相似度，用户u k是第一用户，用户u j为第二用户。
  7. 根据权利要求1所述的方法，其中，所述从所述选取的多个信任度对应的多个第二用户评分过的媒体内容中确定提供给该第一用户的推荐媒体内容包括：
    在所述选取的多个信任度对应的多个第二用户评分过的媒体内容中，选取所述第一用户没有访问过的第四媒体内容集合；
    在所述评分记录中查找所述多个第二用户分别对所述第四媒体内容集合中每一项媒体内容的评分值；
    根据所述多个第二用户对所述第四媒体内容集合中各项媒体内容的评分值，确定各项媒体内容的可推荐度；
    将所述可推荐度达到预定条件的媒体内容作为所述第一用户的推荐媒体内容。
  8. 一种媒体内容推荐装置，包括：处理器，与所述处理器相连接的存储器；所述存储器中存储有机器可读指令单元；所述机器可读指令单元包括：
    评分记录获取单元，用于获取多个用户对不同媒体内容的评分记录；
    推荐请求接收单元，用于接收应用客户端发送的所述多个用户中第一用户的媒体内容推荐请求；
    信任度确定单元，用于根据所述评分记录确定该第一用户对所述多个用户中的各个第二用户的信任度，其中，该第一用户对任一第二用户的所述信任度用于表征该第一用户对于该第二用户对不同媒体内容评分结果的认可程度；
    信任度选取单元，用于在确定的该第一用户对各第二用户的信任度中选取达到预定条件的多个信任度；
    推荐媒体内容确定单元，用于从所述选取的多个信任度对应的多个第二用户评分过的媒体内容中确定提供给该第一用户的推荐媒体内容，将所述确定的推荐媒体内容的链接发送给所述应用客户端。
  9. 根据权利要求8所述的装置，其中，所述信任度确定单元，用于：
    对于任一第二用户，执行以下操作：
    根据所述评分记录，确定该第二用户与该第一用户的评分值之差在预定范围内的第一媒体内容集合中包含的媒体内容的第一项数；
    根据所述评分记录，确定该第一用户评分过的第二媒体内容集合中包含的媒体内容的第二项数；
    根据所述第一项数与所述第二项数确定该第一用户对该第二用户的支持度；
    根据所述评分记录，确定该第二用户与该第一用户的评分相似度；及
    根据所述支持度及所述评分相似度计算该第一用户对该第二用户的信任度。
  10. 根据权利要求9所述的装置，其中，所述信任度确定单元，还用于：
    在所述评分记录中查找该第二用户与该第一用户都评分过的第三 媒体内容集合，以及该第二用户及该第一用户分别对所述第三媒体内容集合中各媒体内容的评分值；
    根据所述评分记录计算该第一用户对评分过的媒体内容的平均评分值以及该第二用户对评分过的媒体内容的平均评分值；
    根据该第二用户及该第一用户分别对所述第三媒体内容集合中各媒体内容的评分值、该第一用户的所述平均评分值及该第二用户的所述平均评分值，计算该第二用户与该第一用户的评分相似度。
  11. 根据权利要求10所述的装置，其中，所述信任度确定单元通过以下公式(1)计算该第二用户与该第一用户的评分相似度：
    Figure PCTCN2018080785-appb-100005
    其中，Sim(u k，u j)代表用户u k及用户u j之间的评分相似度，I k，j代表用户u k及用户u j共同评分过的第三媒体内容集合，i代表所述第三媒体内容集合中的一项媒体内容，w k，i代表用户u k对媒体内容i的评分值，w j，i代表用户u j对媒体内容i的评分值，
    Figure PCTCN2018080785-appb-100006
    代表用户u k对评分过的媒体内容的平均评分，
    Figure PCTCN2018080785-appb-100007
    代表用户u j对评分过的媒体内容的平均评分。
  12. 根据权利要求9所述的装置，其中，所述信任度确定单元，还用于：
    将所述第一项数与所述第二项数的比值作为该第一用户对该第二用户的支持度。
  13. 根据权利要求9所述的装置，其中，所述信任度确定单元通过以下公式(2)计算该第一用户对该第二用户的信任度：
    Figure PCTCN2018080785-appb-100008
    其中Sup(u k,u j)是用户u k对用户u j的支持度，Sim(u k,u j)是用户u k和用户u j之间的评分相似度，用户u k是第一用户，用户u j为第二用户。
  14. 根据权利要求8所述的装置，其中，所述推荐媒体内容确定单元，用于：
    在所述选取的多个信任度对应的多个第二用户评分过的媒体内容中，选取所述第一用户没有访问过的第四媒体内容集合；
    在所述评分记录中查找所述多个第二用户分别对所述第四媒体内容集合中每一项媒体内容的评分值；
    根据所述多个第二用户对所述第四媒体内容集合中各项媒体内容的评分值，确定各项媒体内容的可推荐度；
    将所述可推荐度达到预定条件的媒体内容作为所述第一用户的推荐媒体内容。
  15. 一种非易失性计算机可读存储介质，其中所述存储介质中存储有机器可读指令，所述机器可读指令可以由处理器执行以完成权利要求1-7任一项所述的方法。
PCT/CN2018/080785 2017-03-28 2018-03-28 媒体内容推荐方法、装置及存储介质 WO2018177303A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US15/929,122 US11182418B2 (en) 2017-03-28 2019-04-10 Media content recommendation method and apparatus and storage medium

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710191230.4 2017-03-28
CN201710191230.4A CN108664484A (zh) 2017-03-28 2017-03-28 媒体内容推荐方法及装置

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US15/929,122 Continuation US11182418B2 (en) 2017-03-28 2019-04-10 Media content recommendation method and apparatus and storage medium

Publications (1)

Publication Number Publication Date
WO2018177303A1 true WO2018177303A1 (zh) 2018-10-04

Family

ID=63674245

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/080785 WO2018177303A1 (zh) 2017-03-28 2018-03-28 媒体内容推荐方法、装置及存储介质

Country Status (3)

Country Link
US (1) US11182418B2 (zh)
CN (1) CN108664484A (zh)
WO (1) WO2018177303A1 (zh)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111368202B (zh) * 2020-03-06 2023-09-19 咪咕文化科技有限公司 搜索推荐方法、装置、电子设备及存储介质
CN112559888A (zh) * 2020-12-25 2021-03-26 北京明略软件系统有限公司 一种推荐内容追溯方法、系统、电子设备及可读存储介质
CN113051085B (zh) * 2020-12-28 2024-04-30 北京达佳互联信息技术有限公司 服务调用方法、装置、服务器及存储介质

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104156436A (zh) * 2014-08-13 2014-11-19 福州大学 一种社交云媒体协同过滤推荐方法
CN105340286A (zh) * 2013-06-10 2016-02-17 汤姆逊许可公司 用于向用户推荐内容的方法和系统
US9392314B1 (en) * 2014-04-07 2016-07-12 Google Inc. Recommending a composite channel
CN106126586A (zh) * 2016-06-21 2016-11-16 安徽师范大学 一种基于综合评价信任的社交网络推荐模型构建方法

Family Cites Families (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6112186A (en) * 1995-06-30 2000-08-29 Microsoft Corporation Distributed system for facilitating exchange of user information and opinion using automated collaborative filtering
US7617127B2 (en) * 2000-04-28 2009-11-10 Netflix, Inc. Approach for estimating user ratings of items
DE10247928A1 (de) * 2001-10-31 2003-05-28 Ibm Auslegen von Empfehlungssystemen, so dass sie allgemeine Eigenschaften im Empfehlungsprozess behandeln
WO2005013597A2 (en) * 2003-07-25 2005-02-10 Keepmedia, Inc. Personalized content management and presentation systems
US8086605B2 (en) * 2005-06-28 2011-12-27 Yahoo! Inc. Search engine with augmented relevance ranking by community participation
US8738467B2 (en) * 2006-03-16 2014-05-27 Microsoft Corporation Cluster-based scalable collaborative filtering
US8001132B2 (en) * 2007-09-26 2011-08-16 At&T Intellectual Property I, L.P. Methods and apparatus for improved neighborhood based analysis in ratings estimation
US8010536B2 (en) * 2007-11-20 2011-08-30 Samsung Electronics Co., Ltd. Combination of collaborative filtering and cliprank for personalized media content recommendation
US8732104B2 (en) * 2008-10-03 2014-05-20 Sift, Llc Method, system, and apparatus for determining a predicted rating
US8204878B2 (en) * 2010-01-15 2012-06-19 Yahoo! Inc. System and method for finding unexpected, but relevant content in an information retrieval system
US20130097056A1 (en) * 2011-10-13 2013-04-18 Xerox Corporation Methods and systems for recommending services based on an electronic social media trust model
CN103345503B (zh) * 2013-07-01 2016-04-13 杭州万事利丝绸科技有限公司 一种基于小波网络的丝绸产品个性化推荐方法
CN104462093B (zh) * 2013-09-13 2019-12-10 Sap欧洲公司 个人推荐方案
CN103995823A (zh) * 2014-03-25 2014-08-20 南京邮电大学 一种基于社交网络的信息推荐方法
US10102559B1 (en) * 2014-09-30 2018-10-16 Amazon Technologies, Inc. Diversification of recommendations
US9558244B2 (en) * 2014-10-22 2017-01-31 Conversable, Inc. Systems and methods for social recommendations
US9767102B2 (en) * 2014-12-01 2017-09-19 Comcast Cable Communications, Llc Content recommendation system
US10380609B2 (en) * 2015-02-10 2019-08-13 EverString Innovation Technology Web crawling for use in providing leads generation and engagement recommendations
CN106484876A (zh) * 2016-10-13 2017-03-08 中山大学 一种基于典型度和信任网络的协同过滤推荐方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105340286A (zh) * 2013-06-10 2016-02-17 汤姆逊许可公司 用于向用户推荐内容的方法和系统
US9392314B1 (en) * 2014-04-07 2016-07-12 Google Inc. Recommending a composite channel
CN104156436A (zh) * 2014-08-13 2014-11-19 福州大学 一种社交云媒体协同过滤推荐方法
CN106126586A (zh) * 2016-06-21 2016-11-16 安徽师范大学 一种基于综合评价信任的社交网络推荐模型构建方法

Also Published As

Publication number Publication date
CN108664484A (zh) 2018-10-16
US20190236091A1 (en) 2019-08-01
US11182418B2 (en) 2021-11-23

Similar Documents

Publication Publication Date Title
US11830031B2 (en) Methods and apparatus for detection of spam publication
US10133710B2 (en) Generating preview data for online content
US9442903B2 (en) Generating preview data for online content
US9595053B1 (en) Product recommendation using sentiment and semantic analysis
US9536005B2 (en) Social distance based search result order adjustment
US10409874B2 (en) Search based on combining user relationship datauser relationship data
US10789304B2 (en) Method and system for measuring user engagement with content items
US11487769B2 (en) Arranging stories on newsfeeds based on expected value scoring on a social networking system
US20140172877A1 (en) Boosting ranks of stories by a needy user on a social networking system
US20110218946A1 (en) Presenting content items using topical relevance and trending popularity
CN108536786B (zh) 一种信息推荐方法、装置、服务器及存储介质
US10552291B2 (en) Method and system for measuring user engagement with content items
US9514197B2 (en) System and method of selecting events or locations based on content
US9710563B2 (en) Search engine analytics and optimization for media content in social networks
US20110321129A1 (en) Disambiguating online identities
US20180121470A1 (en) Object Annotation in Media Items
WO2018177303A1 (zh) 媒体内容推荐方法、装置及存储介质
US20120203865A1 (en) Apparatus and methods for providing behavioral retargeting of content from partner websites
US20130262463A1 (en) Method and system to provide smart tagging of search input
WO2016168961A1 (zh) 智能终端主题更换方法及装置
US9336553B2 (en) Diversity enforcement on a social networking system newsfeed
US20150310530A1 (en) Navigation and recommendation on payment checkout in a professional social network
CN108319622B (zh) 一种媒体内容推荐方法及装置
JP2017076376A (ja) 算出装置、算出方法および算出プログラム
US20140095976A1 (en) Social Context for Content from Systems External to a Social Networking System

Legal Events

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

Ref document number: 18776253

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18776253

Country of ref document: EP

Kind code of ref document: A1