JP2008210010A - Content delivery method and system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a content delivery method and system, deliverying content while sufficiently reflecting a user's preference as well as the user's potential preference. <P>SOLUTION: A keyword acquisition part 103a includes a preference keyword acquisition part which acquires a keyword of a news article viewed by a user from a news database 101; a non-preference keyword acquisition part which acquires a keyword of a news article not viewed by the user from the news database 101. A clustering part 103b includes a preference clustering part which classifies a content viewed by the user to at least one preference cluster based on the keywords of viewed news articles and a non-preference clustering part which classifies a content not viewed by the user to at least one non-preference cluster based on the keywords of unwatched news articles. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a content distribution method and system, and in particular, content that dynamically generates and updates a cluster based on a user's content viewing history, and performs priority distribution that reflects the user's preference based on the cluster. The present invention relates to a delivery method and system.

  Services that customize information according to the interests of users are being studied. In Patent Literature 1, distribution history information including identification information of news articles distributed to users and user profile information created based on user registration information received from each user prior to service provision are disclosed. A technology that enables news distribution customized for each user is disclosed.

  In Patent Document 2, every time a viewer browses a web page, the words appearing on the page are extracted and classified into genres, and the forgetting concept obtained from the browsing time of each web page is determined as the appearance frequency of words and genres. Apply it to obtain personal preference information, and based on this personal preference information, display a word on the web page that matches the viewer's preference by color, or perform a keyword search with multiple words that match the browser's preference A technique for executing in the background is disclosed.

  However, the technique disclosed in Patent Document 1 is troublesome because the user himself / herself has to input in advance the news genre and keyword that he / she wants to browse. In addition, since keyword selection greatly affects search results in keyword search, if the keyword selection is inappropriate, desired search results may not be obtained.

  In addition, in the technology disclosed in Patent Document 2, preference information is generated based on words appearing on a Web page. Since preference is an abstract concept in the first place, such abstract concepts are classified by word. If the preference information is generated based on the classification result, the preference of the user cannot be sufficiently reflected on the preference information, and the potential preference of the user cannot be reflected.

  In response to such a technical problem, the inventors of the present invention invented a content distribution method and system that sufficiently reflect the user's preference and enables the content distribution reflecting the user's potential preference. The patent application was filed as in Patent Document 3.

In this Patent Document 3, when the content is a news article, paying attention to the fact that the news title expresses the content of the article in a compact manner, the history of viewing the news article by the user is obtained, and obtained from the title of the viewed article. A news article that matches the user's preference is selected by a preference cluster generated by clustering based on the keyword and weighting using articles included in the preference cluster.
JP 2005-242758 A JP 2002-073677 A Japanese Patent Application No. 2006-179756

  In the technique of Patent Document 3, since new content including keywords included in a news article viewed by a user is preferentially distributed, the user has a clear preference limited to a specific area. There was a possibility that the recommendation accuracy would decrease.

  For example, if you watch a lot of professional baseball articles in order to watch a specific baseball player A article, but clustering is performed using “professional baseball” as a keyword, baseball player A appears. Even professional baseball articles that have not been distributed preferentially.

  An object of the present invention is to provide a content distribution method and system capable of solving the above-described problems of the prior art and accurately distributing content that meets user preferences.

  In order to achieve the above-described object, the present invention dynamically generates and updates a preference cluster based on a user's content viewing history, and sets a priority for content distributed to the user based on the preference cluster. In the content distribution method, for each content, a procedure for generating a keyword and content vector representing the content and registering it in the content database, and identifiers of already viewed content that has been viewed by the user and identifiers of unviewed content that has not been viewed , A procedure for extracting a keyword of already viewed content and a keyword of unviewed content from the content database, a procedure of classifying each content into at least one of a plurality of clusters based on the extracted keyword, Are classified into each cluster Based on the content, a procedure for generating a cluster vector representing the characteristics of each cluster, a procedure for setting a weight value for each cluster, and for each new content to be distributed, the content vector and each cluster vector A procedure for calculating the similarity by reflecting the weight value set for each cluster, a procedure for setting the priority for each new content based on the similarity, and providing each new content to the user according to the priority And a procedure for performing.

  According to the present invention, not only the content viewed by the user but also the keywords of the content that has not been viewed are analyzed, and the priority of new content similar to the content that has been viewed is increased, while similar to content that has not been viewed. Since the priority of the new content is lowered, it is possible to deliver the content that accurately reflects the preference to a user who has a clear preference limited to a specific area.

  DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the best embodiment of the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a main part of a content distribution system to which the present invention is applied. Here, online news as content is adaptively distributed to a user's portable terminal 2 via the Internet 3. The news distribution apparatus 1 will be described as an example.

  In the news distribution apparatus 1, in the news database 101, distributed news articles and newly acquired undelivered news articles are managed in a database.

  FIG. 2 is a diagram showing an example of information managed in the news database 101, in which news article text, article vector files, keywords, and the like are managed with article IDs unique to each news article. News article texts are classified into titles, headlines and detailed articles.

  FIG. 3 is a diagram showing an example of the article vector file. For each main term, TF / IDF (Term Frequency / Inverse Document Frequency) is registered. Has been. TF / IDF is the product of TF and IDF, and represents the importance of the term in the document.

  Here, the appearance frequency tfij of the term tj in the article Di is represented by the following expression (1), the appearance frequency idfj of the article Di is represented by the following expression (2) using the number of articles dfj and the total number N of articles in which the term tj appears, Furthermore, if the weight Wtj of the term tj of the article Di is expressed by the following equation (3) and all the different terms in the entire article are n, the article vector of the article Di can be expressed by the equation (4).

  The online news server 102 is a Web server that delivers a news article in response to a request from the mobile terminal 2 that can be connected to the Internet 3, and includes a history acquisition unit 102a, a similarity calculation unit 102b, a priority setting unit 102c, and a priority. A delivery unit 102d is provided.

  The history acquisition unit 102 a acquires a user's viewing history and provides it to the preference extraction management unit 103. In this embodiment, when the user selects an article title and views up to the headline, the viewing history is left for the first time.

  FIG. 4 is a diagram illustrating an example of the viewing history, in which a plurality of pieces of information including at least an article ID unique to a news article viewed by a user and time information at which the news article is accessed are associated with each other. Is remembered. In this embodiment, an article ID unique to a news article that has not been viewed by the user is also registered separately. As the time information, UTC (universal coordinate time) of Coordinated Universal Time is registered in milliseconds on the basis of midnight on January 1, 1970 and the current item.

  As will be described in detail later, the similarity calculation unit 102b compares the content vector of the new content to be newly distributed with the cluster vector that represents the characteristics of the cluster that reflects the user's preference. Calculate similarity. The priority setting unit 102c sets a priority for each new content based on the similarity. The priority delivery unit 102d preferentially selects new content with high priority from the news database 101 and delivers it to the user.

  The preference extraction management unit 103 includes a keyword acquisition unit 103a, a clustering unit 103b, a cluster integration unit 103c, a cluster vector generation unit 103d, and a cluster weighting unit 103e, refers to the news database 101 based on the history information, and will be described later The preference information of each user is generated according to the algorithm.

  The keyword acquiring unit 103a acquires a keyword of a news article viewed by the user from the news database 101, and an unviewed keyword of acquiring a keyword of a news article not viewed by the user from the news database 101. Including an acquisition unit.

  The clustering unit 103b is configured to classify the ID of the already viewed article into at least one preference cluster based on the keyword of the already viewed article, and the ID of the unviewed article based on the keyword of the unviewed article. And a non-preference clustering unit that classifies the data into at least one non-preference cluster.

The cluster integration unit 103c includes a preference cluster integration unit that integrates a plurality of preference clusters with similar contents, and a non-preference cluster integration unit that integrates a plurality of non-preference clusters with similar contents. The cluster vector generation unit 103d generates a cluster vector for each cluster based on the feature of the article classified into each cluster. The cluster weighting unit 103e weights each cluster.
[First embodiment]

  FIG. 5 is a flowchart showing the procedure of the first embodiment of the preference information extraction and update processing in the preference extraction management unit 103. In the present embodiment, a plurality of preference clusters and non-preference clusters are generated as the preference information of each user, and are managed by user IDs as a preference cluster group and a non-preference cluster group, respectively. Further, the preference cluster and the non-preference cluster are continuously updated based on the user's subsequent viewing history.

  In step S1, it is determined whether the viewing history for the news article provided to the user has been received from the history acquisition unit 102a. If it is determined that the viewing history has been received, the process proceeds to step S2. In this viewing history, among the article titles listed on the mobile terminal 2, the ID of the news article selected by the user and viewed up to the headline (viewed article ID), and the article titles listed An ID (unviewed article ID) of a news article that has not been selected by the user and has not been viewed up to the headline is registered. Note that the ID of an article whose title is to be displayed in a list is not registered in the preference history.

  With respect to the already-viewed article ID, the process proceeds to step S3 and subsequent steps, and processing for generating and updating the preference cluster is executed. On the other hand, with respect to the unviewed article ID, the process proceeds to step S15 and subsequent steps, and processing for generating and updating a non-preferred cluster is executed.

  In step S3, the news database 101 is searched using the already viewed article ID as a search key, and the title of the article is extracted. In step S4, keywords are extracted from the article title.

  FIG. 6 is a flowchart showing in detail the keyword extraction process executed by the already viewed keyword acquisition unit of the keyword acquisition unit 103a.

  In step S101, the article title is divided by a space. In step S102, the article title is morphologically analyzed. In the present embodiment, the “chasm (registered trademark)” (Matsumoto Yuji, Kitauchi Kei, Yamashita Tatsuo, Hirano Yoshitaka, Matsuda Hiroshi, Takaoka Kazuma, Asahara Masayuki, 2.3.3 Instructions for use, “Nara Institute of Science and Technology, Graduate School of Information Science, Natural Language Processing, 2003) is used.

  In step S103, the prefix is connected to the subsequent word and extracted as a keyword. In step S104, unknown words are extracted as keywords. In step S105, nouns (excluding non-independence, pronouns, adverbs are possible) are extracted as keywords. In step S106, the noun (suffix) is connected to the previous noun and extracted as a keyword.

  Returning to FIG. 5, in steps S <b> 5 to S <b> 9, as will be described in detail below, the clustering unit 103 b classifies each keyword into at least one of a plurality of preference clusters (clustering).

  In step S5, one of the keywords extracted in step S4 is selected as the current keyword of interest. In step S6, it is determined whether a preference cluster keyword that matches the current attention keyword or a preference cluster keyword including the current attention keyword has already been registered. If the preference cluster keyword has already been registered, the process proceeds to step S8, and the article ID of the keyword of interest is additionally registered in the preference cluster.

  On the other hand, if such a preference cluster keyword is not registered, the process proceeds to step S7, a new preference cluster is generated with the current attention keyword as the preference cluster keyword, and an article ID is newly registered in this preference cluster. The

  In this way, in this embodiment, since preference clusters are generated for all keywords included in articles viewed by the user, not only explicit preferences like clusters in which many already-viewed articles exist, but also articles Potential tastes that do not appear frequently but must be watched can be extracted without omission.

  FIG. 7 is a diagram schematically showing an example of a preference cluster, in which a plurality of preference clusters having each article ID as an element are generated using each keyword as a preference cluster keyword, and these are set as preference cluster groups for each user. Managed. Note that the IDs of articles that include a plurality of preference cluster keywords in the title are registered in duplicate in each preference cluster.

  Returning to FIG. 5, in step S8, it is determined whether or not the processing has been completed for all the keywords of the viewed article. If not completed, the process returns to step S5, and the above-described procedures are repeated while switching the keywords. . When the processing is completed for all keywords, the process proceeds to step S10, and it is determined whether the processing is completed for all article IDs notified in the viewing history. If not completed, the process returns to step S2, and the above-described procedures are repeated while switching the article ID. When the processing is completed for all article IDs, the process proceeds to step S11, and preference cluster integration processing is executed.

  FIG. 8 is a flowchart showing a procedure of preference cluster integration processing. Preference clusters having a large number of overlapping article IDs are integrated into one preference cluster by the preference cluster combination unit of the cluster combination unit 103c.

  In step S201, one of the preference clusters is selected as the current attention cluster. In step S202, one of the preference clusters other than the target cluster is selected as a comparison control cluster. In step S203, it is determined whether the number of article IDs common to the target cluster and the comparison control cluster exceeds a predetermined reference rate (50% in this embodiment) in any cluster.

  If it exceeds the reference value, the process proceeds to step S205, and the current attention cluster and the comparison control cluster are integrated into one preference cluster. In step S206, the preference cluster keyword is updated. In the present embodiment, if the keyword of one cluster to be integrated includes all the keywords of the other cluster, the keyword of one cluster is adopted as the preference cluster keyword after integration. The keyword becomes a new preference cluster keyword.

  In step S207, it is determined whether or not there are other comparison target clusters. The above-described procedures are repeated until the comparison with all preference clusters is completed, and then the process proceeds to step S208. In step S208, it is determined whether or not integration has been completed for all combinations of preference clusters. If not completed, the process returns to step S201, and the above-described processes are repeated while switching the cluster of interest.

  In the present embodiment, regarding unviewed article IDs, the same processing as in steps S3 to S11 is executed in steps S15 to S23, and a non-preferred cluster is generated / updated.

  Returning to FIG. 5, in step S <b> 12, the cluster vector generation unit 103 d sums the article vectors of all articles included in each cluster for each preference and non-preference cluster, thereby obtaining each preference cluster and non-preference cluster. A cluster vector representative of the features is generated. In step S12, each cluster is weighted by the cluster weighting unit 103e. Here, the weighting method for each cluster will be described using two typical methods as examples.

  The first weighting method is weighting according to the number of articles included in each cluster, and the weight of the cluster is increased as the number of articles included in the cluster increases.

  The second weighting method is weighting based on the concept of forgetting a cluster. The smaller the sum of the time differences between the access time of each article included in the cluster and the current time, the higher the weight of the cluster.

  More specifically, the weight Wi of an article (article vector Di) included in a cluster is defined by the following equation (5) by normalizing the current time τ and the time Ti when the article was viewed by the time T: . Note that λ (0 <λ <1) is a forgetting constant indicating how much forgetting at time T, and the weighting factor W0 = 1.

  The weight value Wc of cluster C using this weight is given by the following equation (6), where m is the number of articles included in the cluster.

  Each preference cluster and the weight value Wc of each preference cluster generated as described above are managed for each user ID by the preference extraction management unit 103 as preference information of each user.

  Next, a method for selecting a news article using the preference information obtained as described above will be described with reference to the flowchart of FIG.

  When the user requests news distribution from his / her mobile terminal 2 and this is detected in step S31, the user ID is notified from the online news server 102 to the preference extraction management unit 103 in step S32. In the preference extraction management unit 103, preference information associated with the notified user ID is provided to the online news server 102. In step S33, in the online news server 102, one of the news articles that is new for the current user is selected on the news database 101 as the current notice article, and the article vector Di is extracted.

  In step S34, one cluster is selected as the current attention cluster from the preference cluster and the non-preference cluster corresponding to the user ID. In step S35, the similarity calculation unit 102b determines the similarity sim (Qc, Di) between the cluster vector Qc of the cluster of interest and the article vector Di, the weight value Wc assigned to this cluster, and the preference cluster vector. The product of Qc and the cosine distance of the article vector Di of the article of interest is obtained by the following equation (7).

  In step S36, it is determined whether or not the similarity calculation with all preference / non-preference clusters has been completed for the current article of interest. If not completed, the process returns to step S34, and the above-described processes are repeated while switching the cluster of interest. When the similarity calculation between the current article of interest and all the preference / non-preference clusters is completed, the process proceeds to step S37, and it is determined whether the cluster of interest having the maximum similarity is a preference cluster or a non-preference cluster. .

  If it is a preference cluster, it will progress to step S38 and the maximum similarity will be registered as a priority of the attention article of this time by the priority setting part 102c. If it is a non-preferred cluster, it will progress to step S39 and the negative value which converted the maximum similarity into negative will be registered as a priority of the attention article of this time. In step S40, it is determined whether or not there are other new articles. The process returns to step S33 until the similarity calculation with all new news articles is completed, and the above-described processing is repeated while switching the article of interest.

  As described above, when the priorities of all new articles are obtained, the articles are sorted in descending order of priority in step S41. At this time, since the priority with the non-preference cluster is a negative value, the priority of the article selected based on the non-preference cluster does not exceed the priority of the article selected based on the preference cluster.

In step S42, the sorted articles are distributed from the priority distribution unit 102d to the user. On the user's mobile terminal 2, articles with high priority are displayed on the initial screen, and articles with low priority can be viewed by scrolling the screen.
[Second Embodiment]

  Next, a second embodiment of preference information extraction and update procedures in the preference extraction management unit 103 will be described.

  In the first embodiment described above, the preference cluster generated / updated based on the already-viewed article ID and the non-preference cluster generated / updated based on the unviewed article ID are independent. Then, the number of clusters will increase. Therefore, in the second embodiment of the present invention, an integrated cluster (group) in which the above-described preference cluster and non-preference cluster are integrated is constructed, and the already-viewed article ID and the unviewed article ID are mixed in each integrated cluster. To be registered.

  FIG. 10 shows an independent management method in which preference clusters and non-preference clusters are managed independently as in the first embodiment, and preference clusters and non-preference clusters are integrated and managed as in the second embodiment. It is the figure which expressed typically the integrated management method.

  In the independent management method shown in (a) of the figure, the already viewed article ID (○) and the unviewed article ID (&times;) are managed in different cluster groups, whereas in the figure (b) In the integrated management method shown, the already-viewed article ID (◯) and the unviewed article ID (&times;) are mixedly registered in each cluster of the integrated cluster group.

  FIG. 11 is a flowchart showing a procedure for extracting and updating preference information in the second embodiment.

  In this embodiment, a plurality of integrated clusters are generated as preference information for each user, and are managed for each user ID as an integrated cluster group. Furthermore, this integrated cluster is continuously updated based on the user's subsequent viewing history.

  The viewing history is notified from the history acquisition unit 102a to the preference extraction management unit 103. When this is detected in step S51, the news database 101 is searched in step S52 using one of the article IDs notified in the viewing history as a search key. And the title of the article is extracted. In step S53, keywords are extracted from the article title. In steps S54 to S58, the keywords are clustered into at least one of a plurality of preference clusters by the clustering unit 103b.

  In step S54, one of the keywords extracted in step S53 is selected as the current keyword of interest. In step S55, it is determined whether or not an integrated cluster keyword matching the current attention keyword or an integrated cluster keyword including the current attention keyword has already been registered. If the integrated cluster keyword is already registered, the process proceeds to step S57, and the article ID of the keyword of interest is additionally registered in the integrated cluster. At this time, an already-viewed / unviewed flag indicating whether the article ID is an already-viewed article ID or an unviewed article ID is also registered.

  On the other hand, if such an integrated cluster keyword is not registered, the process proceeds to step S56 to generate a new preference cluster using the current attention keyword as the preference cluster keyword, and the article ID is already viewed / Registered with the unviewed flag. It should be noted that the ID of an article including a plurality of integrated cluster keywords in the title is also redundantly registered in each integrated cluster in this embodiment.

  In step S58, it is determined whether or not the processing has been completed for all the extracted keywords. If the processing has not been completed, the process returns to step S54, and the above procedures are repeated while switching the keywords. When the process is completed for all keywords, the process proceeds to step S59, and it is determined whether or not the process is completed for all article IDs notified in the viewing history. If not completed, the process returns to step S52, and the above-described procedures are repeated while switching the article ID. When the processing is completed for all article IDs, the process proceeds to step S60, and the preference cluster integration processing is executed in the same manner as described above.

  In step S61, the cluster vector generation unit 103d generates an integrated cluster vector representing the characteristics of each integrated cluster. In step S62, each cluster is weighted by the cluster weighting unit 103e.

  In the present embodiment, the weight Wc is calculated based on the following equation (8) with the browsing rate of articles included in each integrated cluster and the elapsed time from browsing as parameters. Here, m is the total number of article IDs included in the integrated cluster, n is the total number of already viewed article IDs included in the integrated cluster, and Pi is a variable representing the elapsed time since each article IDi was viewed. Therefore, in this embodiment, the weight value Wc increases as the number of already viewed articles having a short elapsed time from the viewing time increases and the number of unviewed articles decreases.

  Next, a method for selecting a news article using the preference information obtained as described above will be described with reference to the flowchart of FIG.

  When a user requests news distribution from his / her mobile terminal 2 and this is detected in step S71, the user ID is notified from the online news server 102 to the preference extraction management unit 103 in step S72. In the preference extraction management unit 103, preference information associated with the notified user ID is provided to the online news server 102. In step S73, in the online news server 102, one of the news articles that is new for the current user is selected on the news database 101 as the current notice article, and the article vector Di is extracted.

  In step S74, one integrated cluster is selected as the current attention cluster from the preference information corresponding to the user ID. In step S75, in the similarity calculation unit 102b, the similarity sim (Qc, Di) between the vector Qc of the cluster of interest and the article vector Di is set to the weight value Wc assigned to each integrated cluster, and the integrated cluster. The product of the vector Qc and the cosine distance of the article vector Di of the article of interest is obtained by the above equation (7).

  In step S76, it is determined whether or not the similarity calculation with all integrated clusters has been completed for the current article of interest. If not completed, the process returns to step S74, and the above-described processes are repeated while switching the cluster of interest. When the similarity calculation between the current noticed article and all the integrated clusters is completed, the process proceeds to step S77, and the priority setting unit 102c registers the maximum similarity as the priority of the current noticed article.

  In step S78, it is determined whether or not there are other new news articles, and the process returns to step S73 until the similarity calculation with all new news articles is completed, and the above-described processing is repeated while switching the article of interest.

As described above, when the priorities of all new news articles are obtained, in step S79, the news articles are sorted in descending order of priority. In step S80, the sorted news articles are distributed from the priority distribution unit 102d to the user. On the user's mobile terminal 2, articles with high priority are displayed on the initial screen, and articles with low priority can be viewed by scrolling the screen.
[Third embodiment]

  Next, a third embodiment of the present invention will be described. In the integrated management method of the second embodiment described above, when the unviewed ratio is large, there is a high possibility that the variation between a cluster including many unviewed article IDs and a small cluster is large. As a result, in a cluster including many unviewed article IDs, the weight value is lowered by weighting based on the audience rating. Therefore, the integrated management method is effective when the ratio of unviewed article IDs is relatively small.

  On the other hand, the independent management method works effectively even when the unviewed article ID is large, but the number of clusters increases about twice as much as the case of the integrated management method, and the similarity calculation processing that is most burdensome increases accordingly. Therefore, it is necessary to suppress the use as much as possible.

  Therefore, in this embodiment, the management method of each user's preference information is adaptively switched to either the independent management method of the first embodiment or the integrated management method of the second embodiment according to the audience rating. There is a feature in the point.

  FIG. 13 is a flowchart showing the operation of the present embodiment. In step S91, the user's unviewed rate Pn is compared with the reference unviewed rate Pref, and if Pn> Pref, the process proceeds to step S92 and the independent management method is used. Applied. On the other hand, if Pn ≦ Pref, the process advances to step S93 to apply the integrated management method.

  FIG. 14 is a flowchart showing a procedure for setting the reference non-viewing rate Pref. This Pref setting process is periodically executed independently of the management method switching process described with reference to FIG. The set reference non-viewing rate Pref is reflected in step S91 of the flowchart of FIG.

  In step S301, the initial value Pref0 is set to the recommended value Pref_i of the reference non-viewing rate Pref. In step S302, a group of users whose current unviewed rate Pn is equal to or greater than a predetermined selection threshold value Pref1 is extracted as a monitoring target user, and the independent management method is forcibly applied to these monitored users. In step S303, the non-viewing rate Pn is continuously measured only for the monitoring users whose unviewing rate Pn has fallen among the monitoring users. In step S304, the management method of the monitoring user whose unviewed rate Pn is lower than the recommended value Pref_i is switched to the integrated management method.

  In step S305, it is determined whether or not the ratio of monitoring users whose unviewing rate Pn has increased among the monitoring users to which the integrated management method is applied is equal to or greater than the threshold value Nref. If it is Nref or more, the process proceeds to step S307, and the management method of the monitoring user whose unviewing rate Pn has started to rise is returned to the independent management method. In step S308, the recommended value Pref_i is decreased by a predetermined unit amount δ, and then the process returns to step S303. If it is determined in step S305 that the ratio of monitored users whose unviewed rate Pn has started to increase is below the threshold Nref, the process proceeds to step S309, and the current recommended value Pref_i is updated to the reference unviewed history Pref. be registered.

  In each of the above-described embodiments, the case where the content is online news has been described as an example. However, the present invention is not limited to this, and can be similarly applied even when the content is a mail magazine.

  Further, in each of the above-described embodiments, all news articles are distributed regardless of the priority level, and only the display order and the display position are described according to the priority levels. However, the present invention is limited to this. It is not limited, and only articles with higher priority may be selectively distributed, and other articles may not be distributed.

It is a block diagram of a content distribution system to which the present invention is applied. It is the figure which showed an example of the information managed by the news database. It is the figure which showed an example of the article vector file. It is the figure which showed an example of viewing history. It is the flowchart which showed the procedure of 1st Embodiment of the extraction / update process of preference information. It is the flowchart which showed the procedure of the keyword extraction process. It is the figure which expressed typically an example of a taste cluster. It is the flowchart which showed the procedure of preference cluster integration processing. It is the flowchart which showed the procedure of 1st Embodiment which selects a news article using preference information. It is the figure which expressed typically the difference with the independent management method and the integrated management method of a cluster. It is the flowchart which showed the procedure of 2nd Embodiment of the extraction / update process of preference information. It is the flowchart which showed the procedure of 2nd Embodiment which selects a news article using preference information. It is the flowchart which showed the procedure of 3rd Embodiment. It is the flowchart which showed the procedure which sets a reference | standard non-viewing rate in 3rd Embodiment.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... News distribution apparatus, 2 ... Portable terminal, 3 ... Internet, 101 ... News database, 102 ... Online news server, 103 ... Preference extraction management part

Claims (9)

In a content distribution method for dynamically generating and updating a cluster reflecting user preferences based on a user's content viewing history, and setting a priority for content distributed to the user based on the cluster,
For each content, a procedure for generating a keyword and content vector representing the content and registering it in the content database;
A procedure for receiving identifiers of already viewed content viewed by a user and identifiers of unviewed content not viewed;
A procedure for extracting already viewed content keywords and unviewed content keywords from the content database;
A procedure for classifying each content into at least one of a plurality of clusters based on the extracted keywords;
Generating a cluster vector representing the characteristics of each cluster based on the content classified into each cluster;
A procedure for setting a weight value for each cluster;
For each new content to be distributed, a procedure for calculating the similarity between the content vector and each cluster vector reflecting the weight value set for each cluster;
A procedure for setting a priority for each new content based on the similarity;
And providing each new content to the user according to the priority. The classification procedure is
A procedure for classifying already viewed content into at least one of a plurality of preference clusters based on the keyword;
Classifying unviewed content into at least one of a plurality of non-preferred clusters based on the keyword,
In the procedure for setting the priority, a new content having a higher similarity to a preference cluster is set to a higher priority, and a new content having a higher similarity to a non-preference cluster is set to a lower priority. The content distribution method according to claim 1. The classification procedure is
A procedure for classifying previously viewed content into at least one of a plurality of integrated clusters based on the keywords;
Classifying unviewed content into at least one of the plurality of integrated clusters based on the keywords,
The procedure for setting the weight value is as follows:
The content distribution method according to claim 1, wherein a higher weight value is set for a cluster having a higher proportion of already viewed content in the total number of registered contents. A first content delivery procedure according to claim 2;
A second content delivery procedure according to claim 3;
And a procedure for calculating the content non-viewing rate for each user,
The content is distributed according to the first content distribution procedure to a user whose ratio of the unviewed content is equal to or greater than a predetermined reference value, and to the user whose ratio of the unviewed content is less than a predetermined reference value, A content distribution method characterized by distributing content according to the content distribution procedure.   The content distribution method according to claim 1, wherein the content is an online news article. In a content distribution system that dynamically generates and updates a cluster based on a user's content viewing history, and sets a priority for content distributed to the user based on the cluster,
For each content, a content database in which keywords and content vectors representing the content are registered,
History acquisition means for acquiring content viewing history for each user;
Keyword acquisition means for referring to a content database based on the viewing history and acquiring keywords of already viewed content viewed by the user and keywords of unviewed content not viewed;
Clustering means for classifying each content into at least one cluster based on the acquired keyword;
Cluster vector generation means for generating a cluster vector representing the characteristics of each cluster based on the content classified into each cluster;
Weighting means for setting a weight value for each cluster;
For each new content to be distributed, similarity calculation means for calculating the similarity between the content vector and each cluster vector reflecting the weight value;
Priority setting means for setting a priority for each new content based on the similarity;
A content distribution system comprising: content distribution means for distributing each new content according to the priority. The clustering means is
Means for classifying already viewed content into at least one of a plurality of preference clusters based on the keyword;
Classifying unviewed content into at least one of a plurality of non-preferred clusters based on the keyword,
The priority setting unit sets a higher priority for new content having a higher similarity to a preference cluster, and sets a lower priority for a new content having a higher similarity to a non-preference cluster. 6. The content distribution system according to 6. The clustering means is
Means for classifying already viewed content into at least one of a plurality of integrated clusters based on the keywords;
Classifying unviewed content into at least one of the plurality of integrated clusters based on the keywords;
The weighting means is
7. The content distribution system according to claim 6, wherein a higher weight value is set for a cluster having a higher proportion of already viewed content in the total number of registered contents. A first content distribution system according to claim 7;
A second content distribution system according to claim 8;
Means for determining the unviewed rate of each user,
The content is distributed by the first content distribution system to a user whose ratio of the unviewed content is equal to or greater than a predetermined reference value, and the second content is transmitted to the user whose ratio of the unviewed content is less than a predetermined reference value. A content distribution system, wherein content is distributed by the content distribution system.

Images