JP5800391B2 - Broadcast program selection method, apparatus and system - Google Patents

Description

The present invention relates to a broadcast program selection method, apparatus and system for selecting a broadcast program suitable for a viewing user from a large number of broadcast programs provided according to a time schedule by TV (television) broadcast, radio broadcast or cable broadcast. In particular, a broadcast program selection method for selecting a topical broadcast program suitable for viewing users using information freely posted by an unspecified number of third parties such as SNS (social networking service) , Apparatus and system .

  In the place of communication, TV news from the previous day is often talked about. At this time, if the news is being watched, communication can proceed smoothly. However, there are many news programs and it is difficult to record and view all of them.

  Patent Document 1 discloses a technique for presenting a TV program according to the preference of the user group from the attributes of one or more users and the TV viewing history. Patent Document 2 discloses a method for estimating an attribute of a contributor from past contents of an SNS contributor.

JP 2011-61668 A Japanese Patent Application No. 2010-210587

  In Patent Document 1, it is possible to view a TV program that suits the taste if attributes of users who are likely to communicate can be input in advance. However, since news program lineups differ from day to day, news that becomes a hot topic in programs that suit your tastes is not always taken up.

  In Patent Document 2, it is possible to determine whether or not the content of the program goes up to a topic by analyzing an SNS post for a TV program. However, since SNS posting is performed simultaneously or delayed with the broadcast of TV programs, all TV programs must be recorded, which is not realistic.

  An object of the present invention is to solve the above-described problems of the prior art and provide a program that provides information that is likely to become a topic in a community to which a user relates from among broadcast programs that have become a topic among contributors such as SNS. Another object of the present invention is to provide a broadcast program selection device that makes it possible to make a reservation for viewing and recording.

In order to achieve the above object, the present invention is characterized in that a broadcast program method, apparatus and system for selecting broadcast programs related to a user's exchange have the following configurations.

  (1) Site information acquisition means for acquiring the description content from a predetermined website, posted information search means for searching for posted information posted by an unspecified number of third parties regarding the description content of the website, Program information acquisition means for acquiring information on stations and time zones that broadcast broadcast programs related to the description contents of the website, a reaction prediction engine that predicts reaction information posted in response to broadcast of the broadcast program, and user interaction An exchange database for managing information related to the user, and a selection means for selecting a viewing target from the predicted broadcast program based on reaction information posted by a poster similar to the user's exchange.

  (2) The screening means compares the response information for the predicted broadcast program with the information about the user's exchange, the number of contributors is larger, the opinion is divided into positive and negative groups, and the attributes of the posters are Based on the policy that the broadcasting programs that are more similar to the attributes of the exchanger are selected more preferentially, the broadcasting programs with higher priority are selected from the predicted broadcasting programs.

  (3) Compare at least one of the time series of the subtitles of the broadcast program recorded in accordance with the recording reservation and the time series of the posted information for the broadcast program and the description content of the site, and extract a recording section whose similarity exceeds a threshold value A similarity calculation means is further provided.

  According to the present invention, the following effects are achieved.

  (1) Among the many information published on the website on the network, among SNS and other broadcast programs that have become a hot topic among contributors such as SNS, SNS contributors that are particularly similar to the user community Since it is possible to select broadcast programs that are hot topics, it is possible to selectively view broadcast programs that provide information that tends to become hot topics in the community to which the user relates.

  (2) When selecting broadcast programs that are talked about among SNS contributors who are similar to the user's exchange layer, broadcast programs with a higher number of contributors are given priority, so more topics with exchange users Only broadcast programs that can share the video can be viewed preferentially.

  (3) When selecting broadcast programs that are talked about among SNS contributors who are similar to the user's social network, broadcast programs in which opinions are divided into positive groups and negative groups are given priority, so they become more topical. Only easy broadcast programs can be viewed preferentially.

  (4) Unnecessary recording parts because a recording section with high similarity to the site feature is extracted from the broadcast program recorded according to the recording reservation, based on the comparison result between the time series of the caption and the site feature. This eliminates the need for fast-forwarding.

  (5) Unnecessary because a recording section with a high similarity to the site feature amount is extracted from the broadcast program recorded according to the recording reservation based on the comparison result between the time series of post information for the broadcast and the site feature amount. Eliminates the need to fast-forward through the recorded part.

It is the figure which expressed the outline | summary of this invention typically. It is a functional block diagram of a viewing reservation apparatus according to the first embodiment of the present invention. It is the figure which showed an example of the teacher data made to learn with a prediction engine. It is the flowchart which showed the operation | movement of 1st Embodiment of this invention. It is a functional block diagram of the viewing-and-listening apparatus which concerns on 2nd Embodiment of this invention. It is a functional block diagram of the recording cutout part in 2nd Embodiment. It is the figure which showed typically the recording cutout method in 2nd Embodiment.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a diagram schematically showing a utilization scene of a broadcast program selection apparatus according to the present invention. Here, a description will be given by taking an example of selection of a television program.

  Attention users belong to many communities such as “company colleagues”, “friends of the same generation” or “friends through hobbies” in the real world, and each community has a topic that is unique or characteristic to the community. There are many conversations.

  One such topic is a broadcast program broadcast on TV or radio, and if it is a colleague of the company, a broadcast program related to its product becomes a topic, and if it is a friend of the same generation, the generation concerned A broadcast program with a content that interests the user is a hot topic, and if it is a friend through a hobby, a broadcast program related to the hobby is often a hot topic. Therefore, if a broadcast program related to the community is missed, it may be difficult to enter a conversation, and there is a desire to know in advance the broadcast program related to the community.

  In the present invention, for example, a feature amount of the website is extracted based on a news article or the like posted on a website that provides topical information such as a news website. Then, by applying this site feature amount to a learned program prediction engine (not shown) to be described later, a station (channel) and a time zone for broadcasting topical information on TV are predicted.

  Furthermore, in the present invention, by applying the predicted TV broadcast attribute information to a learned reaction prediction engine, which will be described later, a broadcast program that is more likely to be a topic with a community exchange person related to the user. Sort out. Then, these are presented to the user of interest as viewing targets or recording targets, and further viewing reservations or recording reservations are made as necessary.

  As will be described in detail later, the program prediction engine applies meta information (site information) about the description content of the website in the past, and the news article to Twitter (registered trademark), mixi (registered trademark) or Facebook ( Information posted by each SNS user (registered trademark) on the SNS community (hereinafter referred to as “SNS posted information”) And a large number of relationships with stations and time zones where the news articles are actually broadcasted are previously constructed as teacher data.

  As will be described in detail later, the response prediction engine determines the number of contributors, the position of the broadcast content (positive or negative), and the attributes of the contributors (age, selection, occupation, etc.) as the response of the SNS user to the predicted broadcast program. , Hobbies, etc.).

  Then, based on reaction information posted by a poster who is similar to the user's exchange layer, a broadcast program suitable as a topic in the community is selected from the predicted broadcast programs and presented to the user. .

[First embodiment]
FIG. 2 is a functional block diagram showing the configuration of the first embodiment of the television viewing reservation apparatus to which the broadcast program selection apparatus of the present invention is applied. Here, the configuration unnecessary for the description of the present invention is omitted. ing.

  In the news database 4, a large number of Web sites (news sites) for posting news articles are registered. In the viewing reservation apparatus 1, the site information acquisition unit 101 acquires information on each Web site from the news database 4. In this embodiment, an API of a news site can be used.

  The site feature quantity extraction unit 102 analyzes the description content of the acquired website, the title of the news article posted on the website, the classification of each news article, the keyword, and the uniqueness set in the website Extracts meta information such as tag information (hereinafter also referred to as site feature) and URL.

  In the present embodiment, TF-IDF (Term Frequency-Inverse Document Frequency: word appearance frequency-reverse appearance frequency) can be employed for the keyword extraction. The extracted site feature amount and URL are described in the search request as a search keyword and notified to the SNS post information server 103 such as a Twitter server, for example.

  In the SNS post information server 103, a lot of information posted by SNS users (SNS post information) for news articles introduced on the website is registered, and in response to the search request, SNS posting information having a high degree of similarity with the site feature amount and URL is extracted and provided to the SNS feature amount extraction unit 104.

  The SNS feature amount extraction unit 104 includes a profile identification unit 104a for identifying a profile (gender, age, occupation, etc.) of an SNS user who has posted the extracted SNS post information, and positive or negative for each news article of the SNS user. A position identification unit 104b for identifying a negative position is provided, and the profile, the position with respect to the news article, the number of posts for each elapsed time from the announcement time of the news article, and the like are sent to the program prediction engine 105 as the SNS feature amount. Notice.

  As for the method of profile identification by the profile identification unit 104a, for example, a paper by the inventors et al. “A Twitter Profiler Estimation Method for Market Analysis” (Kazufumi Ikeda, Motoshi Hattori, Kazunori Matsumoto, Tomohiro Ono, Teruo Higashino “Market” Twitter poster profile estimation method for analysis ", Multimedia, Distributed, Collaboration and Mobile (DICOMO) Symposium, 7E-1, July 6, 2011).

  The program prediction engine 105 includes a support vector machine (hereinafter abbreviated as “SVM”), a Bayesian network, and a C4.5 classifier as a knowledge base, and examples of many news articles observed in the past. Each news article (example) is represented as a multidimensional feature vector having a number of features (features). Then, for each news article, the separation plane is learned from the training data classified by the station (channel) that broadcasted it on TV and its time zone, and how the description content of the newly requested Web site depends on the separation plane. On the basis of whether the program is classified as a viewing reservation candidate, a station and a time zone of a suitable broadcast program are predicted.

  In this embodiment, the site feature value extracted by the site feature value extraction unit 102 and the SNS feature value extracted by the SNS feature value extraction unit 104 are used as the features of the cases. As features related to the site feature amount, meta information such as a website domain, a title, a genre, a posting position on a Web screen, a keyword, and a posting time is given. The features related to the SNS feature amount include the number of posts for each elapsed time from the announcement time of the news article, the poster's position (whether the post content is positive or negative), and the poster's Profile information is given.

  FIG. 3 is a diagram showing an example of input of the learning data. For each case (news article), the feature related to the site feature amount (Web site domain, title, genre, placement position, keyword, publication time,...) , Features related to the post information feature amount (number of posts for each elapsed time, post content is positive, post content is negative ...) and features related to the prediction target (whether broadcast station, time zone, news article was broadcast) Is registered.

  If information about the station and time zone (program information) that broadcasts the broadcast program related to the description contents of the website can be obtained separately from outside, the program information is obtained instead of the program prediction engine 105. A program information acquisition unit may be provided.

  The reaction prediction engine 106 also includes an identifier such as SVM, Bayesian network or C4.5 as a knowledge base, and the number of posts from SNS users, the contents of each post (negative positive), posts about broadcast programs broadcast in the past. When the attribute information (age, gender, hobby, occupation, etc.) of each SNS user is learned and given meta information, station, time zone, etc. of the predicted broadcast program, the number of posts for each broadcast program, position ( Positive negative) and the attributes of each contributor.

  In the exchange relation database 109, attribute information such as acquaintances, friends or related persons who have exchanges with the user in various communities is stored in advance. The program selection unit 107 compares the predicted reaction information for each broadcast program with the exchange relationship, and has a larger number of contributors, opinions are divided into positive groups and negative groups, and the contributors are their own exchangers. Based on a policy that a broadcast program that is more similar to a layer is selected with higher priority, a broadcast program with a higher priority is selected from the predicted broadcast programs.

  The viewing reservation unit 108 makes a viewing reservation with respect to the TV 3 for the selected N broadcast programs, or makes a recording reservation with respect to the STB 2.

  FIG. 4 is a flowchart showing a viewing reservation method in the viewing reservation apparatus 1. In step S1, the site information acquisition unit 101 acquires information about a news website from the news database 4. In step S2, the site feature amount extraction unit 102 analyzes the contents of the news articles posted on the acquired websites, and extracts site feature amounts of the websites. In step S3, the SNS post information server 103 searches the SNS post information related to the site feature amount of each Web site.

  In step S4, the SNS feature amount extraction unit 104 extracts an SNS feature amount from the search result of the SNS post information. In step S5, the site feature amount and the SNS feature amount are applied to the program prediction engine 105, and a station and a time zone in which the description content of the Web site is broadcast on TV are predicted.

  In step S6, in the reaction prediction engine 106, based on the number of posts from the SNS user for the broadcast program broadcasted in the past, the posted content, the attribute information of the poster (age, gender, hobby, occupation, etc.), etc. A response to the predicted broadcast program is predicted.

  That is, in step S6a, the number m of posters for each broadcast program is predicted. In step S6b, the poster's position for each broadcast program is predicted as a positive rate r1 and a negative rate r2. In step S6c, the similarity c of the attribute between the contributor for each broadcast program and his / her exchange layer is predicted.

  In step S7, the program selection unit 107 compares the predicted response to each broadcast program with the data related to the user's exchange, and finally selects the broadcast program to be presented to the viewing user.

  That is, in step S7a, the function f (m) increases in value as the number of contributors m increases, and the function g (r1, r2m) increases in value as both the positive rate r1 and the negative rate r2 of the opinion increase or antagonize. , And the sum [f (m) + g (r1, r2m) + h (c)] of the function h (c), the value of which increases as the cos similarity c between the poster and his / her exchange group increases Further, the priority Pi of each broadcast program i is calculated as a function of this sum.

  In step S7b, the priority Pi of each broadcast program i is compared with a predetermined threshold value Pref. If Pi ≧ Pref, the process proceeds to step S7c and the broadcast program is selected. On the other hand, if Pi <Pref, the process proceeds to step S7d and the broadcast program is unselected.

  In step S8, the viewing reservation unit 108 generates a viewing reservation or recording reservation for the selected broadcast program, and notifies the television 3 or STB 2 in step S9, respectively.

  According to the present embodiment, among a lot of information published on the website on the network, among the broadcast programs that have become a hot topic among contributors such as SNS, especially similar to the user's exchange layer Since it is possible to select a broadcast program that has become a topic among the SNS contributors who have made it, it becomes possible to selectively view a broadcast program that provides information that is likely to be a topic in the community to which the user relates.

  Also, when selecting broadcast programs, priority is given to broadcast programs with a larger number of contributors, so that only broadcast programs that can share a topic with an exchanger can be preferentially viewed.

  Further, when selecting broadcast programs, priority is given to broadcast programs whose opinions are divided into positive groups and negative groups, so that only broadcast programs that are more likely to be talked about can be preferentially viewed.

[Second Embodiment]
FIG. 5 is a functional block diagram of the television viewing reservation apparatus according to the second embodiment of the present invention, and the same reference numerals as those described above represent the same or equivalent parts, and thus the description thereof is omitted. The present embodiment is characterized in that a cutout unit 110 is provided that deletes unnecessary portions or cuts out only essential portions from the TV broadcast recorded in the STB 2 based on the recording reservation request.

  FIG. 6 is a functional block diagram of the cutout unit 110, and the subtitle-based similarity calculation unit 110a compares a time series of subtitles related to TV broadcast recorded based on the recording reservation request and the site feature amount. Then, a recording section whose similarity exceeds a threshold is identified. The SNS-based similarity calculation unit 110b compares a time series of SNS posting information for a TV broadcast recorded based on the recording reservation with the site feature amount, and identifies a recording section in which the similarity exceeds a threshold value. In the present embodiment, the Cos similarity is used for similarity calculation in each of the similarity output units 110a and 110b, and a recording section having a high Cos similarity is identified.

  The search result of the SNS post information server 103 can be used as the SNS post information for the recorded TV broadcast. As the site feature amount, the site feature amount extracted by the site feature amount extraction unit 102 can be used.

  The partial deletion / playback execution unit 110c erases other sections of the TV broadcast recorded in the STB 2 while leaving the identified recording sections, or selectively selects only the identified recording sections. Or play back. At this time, only the logical product section of the recording sections identified by the similarity output units 110a and 110b may be left or reproduced, or only the logical sum section may be left or reproduced. You may make it do.

  FIG. 7 is a diagram schematically showing the state of the recording cutout, in which recording sections having both high caption base similarity and SNS base similarity are cut out.

  Subtitles relating to the recorded TV broadcast may be acquired by specifying the station and time zone from a subtitle DB (not shown) installed outside, or the subtitle DB ( (Not shown) may be provided to accumulate all TV broadcast subtitles without omission and extract subtitles corresponding to the station and time zone from the subtitle DB.

  According to the present embodiment, since the broadcast section that the user wants to watch can be further carefully selected, an operation such as fast-forwarding an unnecessary part becomes unnecessary, and only a desired broadcast can be viewed.

  In each of the above embodiments, the site information acquisition unit 101 has been described as acquiring a news article from a news site. However, the present invention is not limited to this, for example, “company colleague”. If you want to acquire topics in the community in particular, make sure that websites searched using keywords such as “Company Name”, “Industry”, “Handling Products” are acquired as keywords that characterize this community. Also good. Or you may make it acquire the web site searched using the interest and topic which are common with the exchange person as a keyword.

  DESCRIPTION OF SYMBOLS 1 ... Viewing reservation apparatus, 2 ... STB, 3 ... Television, 4 ... News database, 101 ... Site information acquisition part, 102 ... Site feature-value extraction part, 103 ... SNS contribution information server, 104 ... SNS feature-value extraction part, 105 ... Program prediction engine, 106 ... Reaction prediction engine, 107 ... Program selection unit, 108 ... Viewing reservation unit, 109 ... AC relation database, 110 ... Cut-out unit

Claims (14)

In a broadcast program sorting device for sorting broadcast programs related to a user's exchange,
Site information acquisition means for acquiring the description content from a predetermined website,
Program information acquisition means for acquiring information on a station and a time zone that broadcast a broadcast program related to the description content of the acquired website,
A reaction prediction engine that predicts reaction information posted in response to broadcast of a broadcast program corresponding to the acquired station and time zone information and the attribute of the poster;
An exchange database that manages the user's exchange attributes,
Refers to the predicted reaction information and attributes, and responds to information on the acquired station and time zone based on reaction information predicted to be posted by a poster whose attributes are similar to those of the exchange managed by the exchange database A broadcast program sorting device comprising: sorting means for sorting a viewing target from the broadcast program. In a broadcast program sorting device for sorting broadcast programs related to a user's exchange,
Site information acquisition means for acquiring the description content from a predetermined website,
Post information search means for searching for post information posted by an unspecified number of third parties regarding the description content of the acquired website;
Program prediction for predicting a station and a time zone for broadcasting a broadcast program related to the description content of the website based on the site feature value extracted from the description content of the acquired website and the feature value extracted from the searched posting information Engine,
A reaction prediction engine that predicts reaction information posted in response to broadcast of a broadcast program corresponding to the predicted station and time zone information and the attribute of the poster;
An exchange database that manages the user's exchange attributes,
Refers to the predicted reaction information and attributes, and responds to information on the acquired station and time zone based on reaction information predicted to be posted by a poster whose attributes are similar to those of the exchange managed by the exchange database A broadcast program sorting device comprising: sorting means for sorting a viewing target from the broadcast program.   The program prediction engine uses the description content of the past website as an example, the site feature value extracted from the website, the feature value extracted from the posted information for the content of the website, and the description content of the website The broadcast program selection apparatus according to claim 2, wherein a knowledge base constructed on the basis of teacher data whose target is a station that broadcasts the TV and its time zone is used. The reaction prediction engine further predicts the number of contributors who post the reaction information and the position of each contributor to the broadcast content,
The screening means that the number of contributors is greater, the broadcast program in which the position is divided into the positive group and the negative group, and the attribute of the poster is more similar to the attribute of the exchanger, is selected more preferentially. 4. The broadcast program selection apparatus according to claim 1, wherein a broadcast program having a high priority is selected from the broadcast programs based on a policy.   5. The broadcast program selection device according to claim 1, further comprising means for making a reservation for viewing and / or recording for the broadcast program selected as the viewing target.   Compare at least one of the subtitle time series of the broadcast program recorded according to the recording reservation and the time series of post information for the broadcast program with the site feature amount extracted from the description content of the website, and the similarity is a threshold value. 6. The broadcast program selection apparatus according to claim 5, further comprising similarity calculation means for extracting a recording section exceeding the recording section.   The similarity calculation means includes a recording section extracted by comparing a time series of captions of the recorded broadcast program and the site feature amount, a time series of post information for the broadcast program, and the site feature amount, 7. The broadcast program selection apparatus according to claim 6, wherein only a section of a logical product with a recording section extracted by comparing the two is extracted.   The similarity calculation means includes a recording section extracted by comparing a time series of captions of the recorded broadcast program and the site feature amount, a time series of post information for the broadcast program, and the site feature amount, 7. The broadcast program selection apparatus according to claim 6, wherein only a section of a logical sum with a recording section extracted by comparing the two is extracted.   9. The broadcast program selection unit according to claim 6, further comprising recording cutout means for erasing other sections of the recorded TV broadcast while leaving the extracted recording sections. apparatus.   9. The broadcast program selection device according to claim 6, further comprising a recording cutout unit that selectively reproduces the extracted recording section of the recorded TV broadcast.   The broadcast program selection device according to any one of claims 1 to 10, wherein the site information acquisition unit acquires the description content from a Web site related to a community to which the user and the exchange partner belong.   The broadcast according to any one of claims 1 to 10, wherein the site information acquisition means acquires the description content from a Web site related to interests and / or topics common to the user and their exchanges. Program sorting device. In a broadcast program selection method for selecting broadcast programs related to a user's exchange,
A procedure in which a computer acquires description contents of a predetermined website from a news database in which a large number of websites for posting news articles are registered ;
The computer, an unspecified number of third parties from registered post information server Submissions post information, searches the post information about the description contents of the acquired Web site against the description contents of the Web site Procedure and
Based on the site feature amount extracted from the acquired description content of the website and the feature amount extracted from the retrieved posting information , the computer can determine a station and a time zone for broadcasting a broadcast program related to the description content of the website. The steps to predict,
A step of the computer, to predict the response information and attributes of its author posted in response to the broadcast of the broadcast program corresponding to the predicted station and time zone information,
The computer compares the predicted reaction information and attribute with the attribute of the user's exchange person, and based on the reaction information posted by a poster who has similar attributes to the user's exchange person, the acquired station and time zone a step of selecting a viewing target from the broadcast program corresponding to the information,
A broadcast program selection method comprising: In a broadcast program selection system for selecting broadcast programs related to user exchangers,
A news database with a large number of websites that post news articles,
Site information acquisition means for acquiring the description content from the Web site;
Post information search means for searching for post information posted by an unspecified number of third parties regarding the description content of the acquired website;
Program prediction for predicting a station and a time zone for broadcasting a broadcast program related to the description content of the website based on the site feature value extracted from the description content of the acquired website and the feature value extracted from the searched posting information Engine,
A reaction prediction engine that predicts reaction information posted in response to broadcast of a broadcast program corresponding to the predicted station and time zone information and the attribute of the poster;
An exchange database that manages the user's exchange attributes,
Refers to the predicted reaction information and attributes, and responds to information on the acquired station and time zone based on reaction information predicted to be posted by a poster whose attributes are similar to those of the exchange managed by the exchange database A broadcast program selection system comprising: screening means for selecting a viewing target from the broadcast program.