JP2012155695A - Program for imparting keyword tag to scene of interest in motion picture contents, terminal, server, and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a program, in continuously broadcasted motion picture contents, capable of imparting the contents to a scene of interest based on a user viewpoint, and to provide a terminal, a server, and a method.SOLUTION: A program according to the present invention functions a computer as comment collection means which collects a comment sentence containing a keyword associated with motion picture contents from a comment server together with a contribution time, contribution count means which divides the motion picture contents into scenes for every predetermined time and counts the number of contributions of comment sentences contributed within a time zone of the relevant scene, scene-of-interest detection means which extracts a plurality of scenes of interest in each of which the number of the contributions becomes equal to or more than a predetermined number, comment keyword extraction means which extracts a comment keyword which is contained in a plurality of comment sentences as many as or more than a predetermined number, and comment tagging means which tags the comment keyword for each scene of interest.

Description

  The present invention relates to a technique for extracting a scene of interest based on a user's viewpoint from video content such as TV broadcast content and video sharing site content.

  In recent years, various video contents have been distributed to a large number of unspecified third parties via television broadcasting or the Internet. The video content is, for example, various media content such as TV broadcast content and video sharing site content.

  For example, in the case of television broadcast content, not only video / audio information as a broadcast program but also subtitle character information displayed in synchronization with the video / audio information is included. The subtitle character information is usually a short sentence summarizing the contents of the broadcast program, and the keyword included in the sentence appropriately represents the video / audio information. The caption character information is given by the content distribution company.

  On the other hand, comment texts from an unspecified number of third parties are sent to sites such as blogs (Web log) and mini blogs (for example, twitter (registered trademark)) via the Internet. It is actively transmitted. Such comment text is characterized by being discussed on one common topic. As such a common topic, there is moving image content distributed to an unspecified number of third parties as described above.

  For example, the user can discuss the broadcast content among a large number of unspecified users via a miniblog or the like while viewing the moving image content. In recent years, for example, a viewing style in which comment texts are transmitted via mini-blogs while watching TV broadcast contents (for example, watching a drama) is becoming widespread. This provides a feeling that one moving image content is shared with many users and viewed.

  In addition, a large number of viewers of the same television broadcast content can extract the keyword that has attracted attention from the comment text written on the blog site and collect the comment text. The keyword corresponds to, for example, a hash tag according to twitter.

  As a conventional technique, there is a technique for detecting a scene of interest (peak) based on the number of posted comment texts (the number of tweets in the case of twitter) and segmenting moving image content (see, for example, Non-Patent Document 1). This technique estimates the contents of each segmented scene from the number of tweets.

  According to the technique described in Non-Patent Document 1, a scene of a debate such as a debate is targeted. For example, it is assumed that the second leader opinion is claimed after the first leader opinion in the moving image content. In this case, the first viewer opinion for the first leader opinion is posted even while the second leader opinion is claimed. Non-Patent Document 1 adjusts a time lag (temporal difference) between a scene of moving image content and the number of posts for the scene.

  From other viewpoints, some viewers cannot view television broadcast content in real time, for example, but often want to extract and watch only the scene of interest. The moving image content is viewed continuously in time, and the summary content obtained by extracting only the attention scene is generally distributed from a content distribution company. On the other hand, there is also a technique for extracting highlights by an approach using image features (see, for example, Non-Patent Document 2). According to this technique, a highlight scene can be extracted by analyzing the characteristics of a moving image of a television program.

D A Shamma, L Kennedy, E F Churchill, `` Tweetthe debates '', Proc WSM '09, 2009 A Hanjalic, “Adaptive Extraction of Highlights From a Sport Video Based on Excitement Modeling Multimedia”, IEEE Transactions on, 2005, [online], [Search January 26, 2011], Internet <URL: http: //www.cse .iitk.ac.in / users / vision / dipen / references / 54_hanjalic-05_highlights-from-sport-video-based-on-excitement-modeling.pdf> “Twitter API Documentation”, [online], [December 26, 2010 search], Internet <URL: http: //apiwiki.twitter.com/w/page/22554679/Twitter-API-Documentation> “Twitter API specification”, [online], [December 26, 2010 search], Internet <URL: http: //watcher.moe-nifty.com/memo/docs/twitterAPI50.txt>

  However, the scene of interest is preferably based on the user viewpoint as much as possible. In addition, giving a comment keyword based on the user viewpoint to the attention scene deepens understanding in user viewing. Furthermore, by searching for a scene of interest using such a comment keyword, a multifaceted scene of interest can be viewed by the user. Note that according to the technique described in Non-Patent Document 2, the amount of calculation of analysis processing is enormous in order to extract image features, and it is impossible to estimate the content of the moving image itself.

  Therefore, according to the present invention, it is an object to provide a program, a terminal, a server, and a method capable of giving the contents to a scene of interest based on a user's viewpoint in continuously broadcast video contents. To do.

According to the present invention, there is provided a program for causing a computer to function so as to receive moving image content that is disclosed to an unspecified number of third parties and extract a scene of interest,
Comment collecting means for collecting comment text including keywords related to the video content together with the posting time from the comment server;
Post number counting means for dividing the video content into scenes at predetermined time intervals and counting the number of comment text posted in the time zone of the scene;
Attention scene detection means for extracting a plurality of attention scenes for which the number of posts is a predetermined number or more,
Comment keyword extraction means for extracting a comment keyword included in a plurality of comment sentences for each attention scene;
The computer is made to function as comment tagging means for tagging a comment keyword for each scene of interest.

According to another embodiment of the program of the present invention,
The comment keyword extracting means extracts words from a plurality of comment sentences by morphological analysis, and extracts characteristic words as comment keywords by TF-IDF (Term Frequency-Inverse Document Frequency). It is also preferable to make the computer function like this.

According to another embodiment of the program of the present invention,
Comment keyword extraction means
For each scene of interest, the part of speech “person name” is extracted from characteristic words by morphological analysis.
For each target scene, an important word having the highest TF-IDF value is extracted from characteristic words by morphological analysis.
The comment tagging means preferably causes the computer to function so as to tag the part of speech “person name” and the important word as comment keywords for each scene of interest.

According to another embodiment of the program of the present invention,
Attention scene detection means
For each scene, calculate the average value μ and standard deviation σ of the number of posts,
If the number of posts for each scene is equal to or greater than the average value μ + standard deviation σ, it is also preferable to cause the computer to function so as to extract the scene as a target scene.

According to another embodiment of the program of the present invention,
It is also preferable that the computer further function as summary content generation means for generating summary content in moving image content by connecting a plurality of scenes of interest.

According to another embodiment of the program of the present invention,
A number-of-posts graph representing the number of posts according to the passage of time, a number-of-posts graph generating means for generating a number-of-posts graph referring to the attention scene accumulating means and tagged with the comment keyword at the time of attention;
It is also preferable to further cause the computer to function as post number graph transmission means for transmitting the post number graph to the terminal.

According to another embodiment of the program of the present invention,
Further causing the computer to function as a request keyword receiving means for receiving the request keyword from the terminal;
The post number graph generating means preferably causes the computer to function so as to tag only the comment keyword including the request keyword with respect to the post number graph.

According to another embodiment of the program of the present invention,
It is preferable that the posting number graph generating means further causes the computer to function so as to generate a posting number graph representing only the number of postings of comment text including the request keyword.

According to another embodiment of the program of the present invention,
It is also preferable that the computer further function as attention scene transmission means for transmitting only the attention scene tagged with the request keyword to the terminal.

According to another embodiment of the program of the present invention,
It is also preferable to further cause the computer to function as comment text transmission means for transmitting the comment text of the attention scene tagged with the request keyword received from the terminal to the terminal.

According to another embodiment of the program of the present invention,
Video content is TV broadcast content or video sharing site content,
It is also preferable to make the computer function so that the comment server is a mini Web log server.

According to another embodiment of the program of the present invention,
The miniblog server is a twitter (registered trademark) server,
The comment collecting means preferably causes the computer to function so as to collect comment sentences including hash tags based on the moving image content from the twitter server.

According to the present invention, a server that receives video content that is disclosed to an unspecified number of third parties and extracts a scene of interest,
Comment collecting means for collecting comment text including keywords related to the video content together with the posting time from the comment server;
Post number counting means for dividing the video content into scenes at predetermined time intervals and counting the number of comment text posted in the time zone of the scene;
Attention scene detection means for extracting a plurality of attention scenes for which the number of posts is a predetermined number or more,
Comment keyword extraction means for extracting a comment keyword included in a plurality of comment sentences for each attention scene;
Comment tagging means for tagging comment keywords for each scene of interest;
It is characterized by having an attention scene distribution means for distributing the tagged attention scene in response to a request from the terminal.

According to the present invention, a terminal connected to a network,
Video content receiving means for receiving video content to be disclosed to an unspecified number of third parties;
Comment collecting means for collecting comment text including keywords related to the video content together with the posting time from the comment server;
Post number counting means for dividing the video content into scenes at predetermined time intervals and counting the number of comment text posted in the time zone of the scene;
Attention scene detection means for extracting a plurality of attention scenes for which the number of posts is a predetermined number or more,
Comment keyword extraction means for extracting a comment keyword included in a plurality of comment sentences for each attention scene;
It is characterized by having comment tagging means for tagging a comment keyword for each scene of interest.

According to the present invention, there is provided a method for extracting a scene of interest in an apparatus for receiving moving image content that is disclosed to an unspecified number of third parties,
A first step of collecting comment text including keywords related to the video content together with a posting time from a comment server;
A second step of dividing the video content into scenes at predetermined time intervals, and counting the number of comment sentences posted in the time zone of the scene;
A third step of extracting a plurality of attention scenes in which the number of posts is a predetermined number or more;
A fourth step of extracting a comment keyword included in a plurality of comment sentences in a predetermined number or more for each attention scene;
And a fifth step of tagging a comment keyword for each scene of interest.

  According to the program, the terminal, the server, and the method of the present invention, an attention scene based on a user's viewpoint is extracted based on the number of posted comment sentences (for example, twitter) from continuously broadcast video contents, and the attention scene is extracted. You can tag the comment keyword. The user can deepen understanding by viewing the scene of interest together with a keyword tag that briefly represents the content of the scene. In addition, the user can search for a scene of interest using a comment keyword assigned as a keyword tag, and can cause the user to view a multifaceted scene of interest. Furthermore, the user can view only the scene of interest in the keyword desired by himself / herself in the entire time course of the moving image content.

It is a system configuration diagram in the present invention. It is a sequence diagram in the present invention. It is a graph showing the number of postings with time passage. It is a functional block diagram of the attention scene extraction server in this invention. It is explanatory drawing showing the flow of the text information based on FIG. It is a sequence diagram in other embodiments of the present invention. It is the number-of-posts graph produced | generated by this invention. It is explanatory drawing by which the number-of-posts graph was displayed on the display of the terminal. It is display explanatory drawing of the number-of-posts graph when the display of only a person keyword is requested | required. It is a display explanatory drawing of the number-of-posts graph when display of only an event keyword is demanded. It is a display explanatory view of the number-of-posts graph when display of only a specific person keyword is requested. It is a display explanatory drawing which emphasized the display image of the comment keyword according to the ratio of the number of posting users to the total number of posts. It is a display explanatory drawing which emphasized the display image of the comment keyword according to the ratio of the number of cited posts to the total number of posts. It is a display explanatory drawing of the number-of-posts graph which jumps to the attention scene only of the person keyword specified by the user. It is explanatory drawing in which the comment text was displayed on the display of the terminal. It is a function block diagram of the attention scene extraction server which implement | achieves the sequence of FIG.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a system configuration diagram according to the present invention.

  According to FIG. 1, the content publishing server 2 distributes moving image content to an unspecified number of third parties via the air or the Internet. The moving image content may be, for example, a television broadcast content or a moving image sharing site content. In the case of television broadcast content broadcast via the air, it is distributed by terrestrial digital broadcast, one-segment broadcast, BS (Broadcast Satellite) broadcast, or the like. Further, the moving image sharing site content broadcast via the Internet is distributed by live video streaming based on Ustream (registered trademark), for example.

  The terminal 4 receives and reproduces the moving image content distributed from the content publishing server 2 via the air or the Internet. The user can view the moving image content reproduced on the terminal 4. The terminal 4 may be a television, a personal computer, a mobile phone, a smartphone, or the like. The terminal 4 may receive video content via a set-top box that receives broadcast waves or a router connected to an access network. In the following description, the terminal 4 is assumed to be a mobile terminal such as a smartphone.

  According to FIG. 1, a miniblog server 3 as a comment server is connected to the Internet. The miniblog server 3 is a twitter site server, for example. An unspecified number of third parties can post comment text to the miniblog server 3 while viewing the moving image content using the terminal 4. Further, the user can freely browse other person's comment text posted on the miniblog server 3 using the terminal 4.

  Moreover, according to FIG. 1, the attention scene extraction server 1 which comprises this invention is further connected to the internet. The attention scene extraction server 1 receives the moving image content from the content publishing server 2 and collects the comment text related to the moving image content from the miniblog server 3. The attention scene extraction server 1 extracts the attention scene in the moving image content from the comment text posted to the moving image content. A keyword tag based on the comment text can be assigned to the attention scene. In response to a request from the terminal 4, the attention scene extraction server 1 distributes the attention scene in the moving image content.

  FIG. 2 is a sequence diagram in the present invention.

  The content publishing server 2 distributes moving image content to an unspecified number of third parties. The terminal 4 possessed by an unspecified number of users reproduces and displays the moving image content received via the air or the network.

(S110) A user who is viewing the moving image content can post a comment sentence related to the moving image content to the miniblog server 3 using the terminal 4. The miniblog server 3 accumulates the received comment text and makes it available to an unspecified number of third parties.

(S111) The attention scene extraction server 1 collects comment text related to the moving image content from the miniblog server 3. Comment sentences including “keywords” related to the moving image content are collected from the comment server together with the posting time. For example, according to the twitter site server, it is possible to search for a comment sentence (tweet) including a specified keyword. For example, when searching for tweets containing the keyword “tigers”, the following URL (Uniform Resource Locator) is specified.
http://search.twitter.com/search.atom?phrase=tigers

Furthermore, according to twitter, in order to collect comment text related to video content, it is also possible to collect only comment text to which a hash tag (# and an English character string) related to the video content is added. In the case of twitter, by using hash tags, topics can be shared, and comment sentences with the same hash tag added can be grouped. For example, when searching for a tweet to which the hash tag “#tigers” is added, the following URL is specified.
http://search.twitter.com/search.atom?q=%23tigers
You can also search for tweets for up to one week by specifying the date. These API specifications are described in Non-Patent Documents 3 and 3, for example.

(S112) Next, the moving image content is divided into scenes at predetermined time intervals, and the number of comment sentences posted in the time zone of the scene is counted. For example, the number of posts every 2 minutes is counted.

  FIG. 3 is a graph showing the number of posts over time. According to FIG. 3, the number of posts is shown every two minutes.

(S113) Next, a plurality of attention scenes in which the number of posts is a predetermined number or more are extracted. Here, it is also preferable to detect a swelling scene that is not limited to a predetermined number but has a sharp increase in the number of posts. In order to detect such a scene of interest, it is also preferable to use a Newton method such as a gradient method. The “Newton-Raphson method” refers to a method of obtaining an approximate value of a point (x intercept) where the function f (x) intersects the x axis by iterative calculation.

  As a specific embodiment, it is also preferable to use the average value μ and the standard deviation σ of the number of posts for each scene (for example, 2 minutes). The average value μ is an average for each scene (for example, 2 minutes) in a predetermined time frame (for example, 10 minutes) (smoothing). The standard deviation σ is the square root of variance in the number of posts for each scene. Variance refers to “the square of the difference between the actual number of posts and the average number of posts” and indicates how much the actual number of posts varies. If the number of posts is “average value μ + standard deviation σ or more” for each scene, the scene is extracted as an exciting scene based on the number of posts.

  According to FIG. 3, “average value μ + standard deviation σ” is represented by a broken line. The portion of the number of posts that exceeds the broken line is represented by Δ, and represents an exciting scene of interest.

(S114) Next, for each scene of interest, a comment keyword included in a plurality of comment sentences is included in a predetermined number or more. Here, a plurality of words are extracted from a plurality of comment sentences by morphological analysis. “Morphological analysis” refers to a technique in which sentences are divided into meaningful words and the part of speech and contents are discriminated using a dictionary. The “morpheme” means the smallest unit having meaning among the elements of the sentence.

  Here, for example, an open source morphological analysis engine “MeCab” can be used for the morphological analysis. This engine has a hierarchical part-of-speech system and can also analyze morpheme part-of-speech. This has a “soft wakachi writing” function that can output all possible morpheme sequences for an input sentence with marginal probabilities. Therefore, the part of speech such as “noun”, “proprietary noun”, “organization”, “region”, “general”, etc. is output for each morpheme.

  Here, according to the present invention, it is also preferable to extract “part of speech“ person name ”among characteristic words by morphological analysis for each scene of interest. In the scene, “person name” is the most characteristic word.

  Then, a characteristic word is extracted as a comment keyword by TF-IDF (Term Frequency-Inverse Document Frequency). TF-IDF is a technique that weights each word, expresses a document from a query in a vector space, and ranks the documents based on the similarity between the query and the query. The higher the ranked value, the more important keywords are recognized. That is, the content of the moving image content is appropriately represented by the appearance frequency of the keyword appearing in the text information.

  For each scene of interest, an important word having the highest TF-IDF value is extracted from characteristic words by morphological analysis. Here, it is preferable to extract (1) a characteristic word of the part of speech “person name” based on morphological analysis and (2) other characteristic words.

(S115) Then, for each target scene, the part of speech “person name” and other characteristic words are tagged as comment keywords. As a result, tags of “main person” and “important word” can be assigned to each scene of interest, and the user can easily understand the exciting scene.

(S116) The target scene tagged with the comment keyword is stored in the target scene storage unit. The comment keyword tagged to the scene of interest is used not only for the user's understanding but also as a search key in the scene of interest storage.

(S117) Optionally, a summary content in the moving image content may be generated by connecting a plurality of scenes of interest. The moving image content based on the graph of FIG. 3 is 260 minutes (4 hours and 20 minutes), while 36 minutes of summary content is generated by connecting only 18 attention scenes.

  The attention scene extraction server 1 transmits a attention scene in a specific moving image content in response to a request from the terminal 4 based on a user operation. As a result, the user can view only the desired scene of interest using the terminal 4. The attention scene extraction server 1 receives the request keyword from the terminal 4, searches using the attention scene storage unit, and transmits the corresponding attention scene to the terminal 4.

FIG. 4 is a functional configuration diagram of the attention scene extraction server in the present invention.
FIG. 5 is an explanatory diagram showing the flow of text information based on FIG.

  According to FIG. 4, the attention scene extraction server 1 includes a communication interface unit 10 connected to the Internet, a comment collection unit 111, a posted number counting unit 112, a attention scene detection unit 113, and a comment keyword extraction unit 114. , A comment tagging unit 115, an attention scene storage unit 116, and a summary content generation unit 117. These functional components are realized by executing a program that causes a computer mounted on the server to function. Of course, it may be realized by executing a program for causing a computer mounted on the terminal to function.

  The comment collection unit 111 collects comment text including keywords related to the moving image content from the comment server (for example, twitter site server) together with the posting time (similar to S111 in FIG. 2 described above). According to FIG. 5, as comment sentences, for example, “Nakazawa's rise is in good condition ...” “If you strike out from the first time, then you will hit after that…” “... Naka! Nakazawa is in good condition from the first time. " The collected comment text is output to the posted number counting unit 112.

  The posting number counting unit 112 divides the moving image content into scenes at predetermined time intervals, and counts the number of comment sentences posted in the time zone of the scene (similar to S112 in FIG. 2 described above). As in the graph of FIG. 3, for example, the number of posts for each scene (for example, every two minutes) is counted. The counted number of posts is output to the attention scene detection unit 113 for each scene.

  The attention scene detection unit 113 extracts a plurality of attention scenes in which the number of posts is a predetermined number or more (similar to S113 in FIG. 2 described above). As shown in the graph of FIG. 3, a scene in which the number of posts is equal to or greater than the average value μ + standard deviation σ is detected as a scene of interest. The extracted attention scene is output to the comment keyword extraction unit 114.

  The comment keyword extraction unit 114 extracts a comment keyword included in a plurality of comment texts in a plurality of comment sentences for each attention scene (similar to S114 in FIG. 2 described above). According to FIG. 5, the count number (or TF-IDF value) for each word based on morphological analysis is clearly shown for one attention scene. Here, the comment keyword extraction unit 114 extracts the word with the part of speech “person name” and the word other than the person name having the highest TF-IDF value. According to FIG. 5, the number of times that the part of speech “person name” and a word other than the person name appear at the same time (the number of times they co-occur) is clearly shown. The comment keyword extraction unit 114 outputs the set of words having the largest number of co-occurrence to the comment tagging unit 115 in this way.

  The comment tagging unit 115 tags a comment keyword for each scene of interest (similar to S115 in FIG. 2 described above). The comment tagging unit 115 tags the part of speech “person name” and the important word as a comment keyword for each scene of interest.

  The attention scene storage unit 116 stores attention scenes tagged with comment keywords. The comment keyword is also used as a search key for searching for a scene of interest.

  The summary content generation unit 117 generates summary content in the moving image content by connecting a plurality of attention scenes. The generated summary content is stored in the attention scene storage unit 116.

  FIG. 6 is a sequence diagram in another embodiment of the present invention.

(S121) The attention scene extraction server 1 generates a posting number graph representing the number of postings according to the passage of time. The number-of-posts graph refers to the attention scene accumulation unit 116 and is tagged with the comment keyword at the attention time.

  FIG. 7 is a post number graph generated by the present invention.

  In the graph of FIG. 7, the number of posts is represented on the vertical axis, and the passage of time is represented on the horizontal axis. A tagged comment keyword is displayed for each point of interest. This post number graph is represented as image information and can be understood by the user as it is. The user can understand the change in the number of posts over time even with relatively long moving image content by simply looking at the number-of-posts graph. In addition, the main comment keywords in the scene of interest can be known.

(S122) The attention scene extraction server 1 transmits the generated post number graph to the terminal 4. The post number graph is displayed on the display of the terminal 4.

  FIG. 8 is an explanatory diagram in which a post number graph is displayed on the display of the terminal.

  According to FIG. 8, a moving image display area 400, a posted number graph display area 401, a noticed person selection button 402, and a noticed event selection button 403 are displayed on the display 40 of the smartphone. In the moving image display area 400, the moving image content received from the content publishing server 2 is reproduced. In the post number graph display area 401, the post number graph received from the attention scene extraction server 1 is displayed. Also, the attention person selection button 402 and the attention event selection button 403 are associated with a relatively large number of keywords among the comment keywords added to the posted number graph as buttons that can be selected by the user. When the user touches a button of a desired keyword, the user can reproduce a scene of interest including the keyword and can view the number of posts based only on the keyword.

(S123) It is assumed that the user touches either the attention person selection button 402 or the attention event selection button 403 displayed on the display of the terminal 4. The terminal 4 transmits the keyword corresponding to the button as a request keyword to the attention scene extraction server 1. The attention scene extraction server 1 searches the attention scene storage unit 116 for attention scenes based on the request keyword.

  The user may browse the post number graph displayed on the terminal 4 and input a desired attention time. As a result, the terminal 4 transmits the attention time to the attention scene extraction server 1. The attention scene extraction server 1 searches the attention scene storage unit 116 for attention scenes based on the attention time.

  Here, S121 and S122 may be executed again. The attention scene extraction server 1 can also tag only the comment keyword including the request keyword on the post number graph. The attention scene extraction server 1 can also generate a posting number graph that represents only the number of postings of comment text including the requested keyword.

(S124) The attention scene extraction server 1 transmits only the attention scene tagged with the request keyword to the terminal 4.

(S125) Also, the attention scene extraction server 1 transmits the comment text of the attention scene tagged with the request keyword received from the terminal to the terminal.

  FIG. 9 is a display explanatory diagram of the number-of-posts graph when the display of only the person keyword is requested.

The post number graph of FIG. 9 is displayed on the display of the terminal 4 when the user operates as follows.
"Attention person selection button" = "Display all"
"Featured event selection button" = "Hide"
Thus, the user can understand at a glance which attention scene has a large number of comment sentences including “person” of interest as a keyword.
In addition, a playback bar is also displayed, and the playback bar moves as time passes for the moving image content being played back.

  FIG. 10 is a display explanatory diagram of the number-of-posts graph when the display of only the event keyword is requested.

The post number graph in FIG. 10 is displayed on the display of the terminal 4 when the user operates as follows.
"Attention person selection button" = "Hide"
"Featured event selection button" = "Display all"
Thus, the user can understand at a glance which attention scene has a large number of comment sentences including “event” as his / her attention as a keyword.

  FIG. 11 is a display explanatory diagram of a post count graph when display of only a specific person keyword is requested.

The post number graph of FIG. 11 is displayed on the display of the terminal 4 when the user operates as follows.
"Featured person selection button" = "Nakazawa"
"Featured event selection button" = "Display all"
Here, as for the number of posts, the number of comment sentences including “Nakazawa” as a keyword is displayed for each time.
Thus, the user can understand at a glance which attention scene has a large number of comment sentences including “Nakazawa” as his keyword.

  FIG. 12 is a display explanatory diagram in which the display image of the comment keyword is emphasized according to the ratio of the number of posting users to the total number of postings.

  According to the graph of FIG. 12, the comment keyword tagged to the target scene of the post number graph is visually emphasized and displayed, and the degree of display emphasis differs for each comment keyword. Here, the ratio of the number of posting users to the total number of postings in the target scene (the number of posting users / the total number of postings) is calculated for the attention scene in which the comment keyword is tagged in the posting number graph. Then, the comment keywords tagged to the attention scene having a higher ratio are visually emphasized and displayed. For example, a display that makes the label color stand out is displayed. The comment keyword of the attention scene with a high ratio of many people posting is displayed more emphasized.

  For example, when the number of posting users is 50 for the number of postings (the number of Twitter tweets) for the first notable scene, the ratio is as follows: 50/50 = 1. Also, when the number of posting users is 20 for 50 postings for the second scene of interest (when each posting user posts 2.5 on average), the ratio is 20/50 = 0.4. In this case, the comment keyword of the first target scene is displayed with emphasis over the second target scene.

  FIG. 13 is a display explanatory diagram in which the display image of the comment keyword is emphasized in accordance with the ratio of the number of quoted posts to the total number of posts.

  According to the graph of FIG. 13, the comment keywords tagged to the target scene of the post number graph are visually emphasized and displayed, and the degree of display emphasis differs for each comment keyword. Here, the ratio of the number of quoted posts to the number of all posts in the target scene (the number of quoted posts / the number of all posts) is calculated for the target scene in which the comment keyword is tagged in the post number graph. Then, the comment keywords tagged to the attention scene having a higher ratio are visually emphasized and displayed. For example, a display that makes the label color stand out is displayed.

  “Quote posting” means, for example, ReTweet in twitter. Retweet means quoting someone else's posted text (Tweet) and including it in your posted text. The comment keywords of the scene of interest with many citations are highlighted.

  For example, when the number of quoted posts is 20 for 50 posts (Twitter Tweets) for the first scene of interest, the ratio is 20/50 = 0.4. Further, when the number of quoted posts is 10 for 50 posts for the second target scene, the ratio is 10/50 = 0.2. In this case, the comment keyword of the first target scene is displayed with emphasis over the second target scene.

  FIG. 14 is a display explanatory diagram of a post number graph for jumping to the attention scene of only the person keyword designated by the user.

The post number graph of FIG. 14 is displayed on the display of the terminal 4 when the user operates as follows.
"Attention person selection button" = "Display all"
"Featured event selection button" = "Display all"
Here, it is assumed that the user designates one of the displayed keywords. This designation may be a click operation or a long press operation with a pointing device. As a result, the playback bar can be jumped to the scene of interest including the keyword.
The user can immediately watch the scene of interest including the keyword specified by the user.

  FIG. 15 is an explanatory diagram in which comment text is displayed on the display of the terminal.

  According to FIG. 15, the comment text is displayed in the moving image display area of FIG. For example, Twitter's Tweet is displayed. As a result, the user can view the posted content.

  FIG. 16 is a functional configuration diagram of the attention scene extraction server that realizes the sequence of FIG.

  According to FIG. 16, in addition to the functional configuration of FIG. 4, the attention scene extraction server 1 includes a posting number graph generation unit 121, a posting number graph transmission unit 122, a request keyword reception unit 123, and a attention scene transmission unit 124. And a comment text transmission unit 125. These functional components are also realized by executing a program that causes a computer mounted on the server to function.

  The number-of-posts graph generation unit 121 is a number-of-posts graph representing the number of posts according to the passage of time, and refers to the attention scene storage unit 116 to generate a number-of-posts graph tagged with the comment keyword at the attention time (See S121 in FIG. 6 described above). Here, the posting number graph generation unit 121 may tag only the comment keyword including the request keyword with respect to the posting number graph. Moreover, the posting number graph generation unit 121 may generate a posting number graph representing only the number of postings of comment text including the request keyword. The generated posting number graph is output to the posting number graph transmission unit 122.

  The post number graph transmission unit 122 transmits the post number graph to the terminal 4 (see S122 in FIG. 6 described above).

  The request keyword receiving unit 123 receives a request keyword from the terminal 4 (see S123 in FIG. 6 described above). The request keyword is output to the posted number graph generation unit 121 and the attention scene transmission unit 124. The request keyword receiving unit 123 may receive the time of interest from the terminal 4. This is the time of the scene of interest at which the user requested the terminal 4 to play back. The attention time requested to be reproduced is output to the attention scene transmission unit 124.

  The attention scene transmission unit 124 refers to the attention scene storage unit 116, and transmits only the attention scene tagged with the request keyword to the terminal (see S124 of FIG. 6 described above). Moreover, the attention scene corresponding to the attention time is transmitted to the terminal.

  The comment text transmission unit 125 refers to the attention scene storage unit 116 and transmits the comment text of the attention scene tagged with the request keyword received from the terminal 4 (see S125 of FIG. 6 described above).

  As described above in detail, according to the program, the terminal, the server, and the method of the present invention, the number of comments (for example, twitter) posted (for example, with a hash tag) in the continuously broadcast video content It is possible to extract an attention scene (swelling scene) based on the user's viewpoint based on the number of tweets, and tag the attention scene with the comment keyword. The comment keyword includes a person name and an important word. The user can deepen understanding by viewing the scene of interest together with a keyword tag that briefly represents the content of the scene. The user can search for a scene of interest using a comment keyword assigned as a keyword tag, and can view a multifaceted scene of interest.

  In addition, the user can know the time of the scene of interest by browsing the posting number graph over the entire time of the video content, and can also view only the scene of interest of the keyword he / she desires.

  Various changes, modifications, and omissions of the above-described various embodiments of the present invention can be easily made by those skilled in the art. The above description is merely an example, and is not intended to be restrictive. The invention is limited only as defined in the following claims and the equivalents thereto.

DESCRIPTION OF SYMBOLS 1 Attention scene extraction server 10 Communication interface part 111 Comment collection part 112 Post number count part 113 Attention scene detection part 114 Comment keyword extraction part 115 Comment tagging part 116 Attention scene storage part 117 Summary content generation part 121 Post number graph generation part 122 Number-of-posts graph transmission unit 123 Request keyword reception unit 124 Scene of interest transmission unit 125 Comment text transmission unit 2 Content disclosure server 3 Mini-blog server 4 Terminal

Claims (17)

A program that allows a computer to function to receive video content that is disclosed to an unspecified number of third parties and extract a scene of interest,
Comment collection means for collecting comment text including keywords related to the video content, together with a posting time, from a comment server;
A number-of-posts counting unit that divides the video content into scenes every predetermined time, and counts the number of comments posted in the time zone of the scenes;
Attention scene detection means for extracting a plurality of attention scenes in which the number of posts is a predetermined number or more;
Comment keyword extraction means for extracting a comment keyword included in a plurality of comment sentences in a plurality of comment sentences for each of the attention scenes;
A program that causes a computer to function as comment tagging means for tagging the comment keyword for each scene of interest. The comment keyword extraction means extracts a word from a plurality of comment sentences by morphological analysis, and extracts a characteristic word from the comment keyword by TF-IDF (Term Frequency-Inverse Document Frequency). The program according to claim 1, wherein the computer functions so as to be extracted as a program. The comment keyword extracting means is
For each noted scene, extract the part of speech “person name” in the characteristic word by the morphological analysis,
For each of the scenes of interest, an important word having the highest TF-IDF value is extracted from the characteristic words by the morphological analysis.
The program according to claim 2, wherein the comment tagging unit causes a computer to function so as to tag the part of speech "person name" and the important word as the comment keyword for each of the attention scenes. The attention scene detection means includes
For each scene, calculate the average value μ and standard deviation σ of the number of posts,
4. The computer according to claim 1, wherein if the number of posts for each scene is equal to or greater than an average value μ + standard deviation σ, the computer is caused to function as a scene of interest. The program described in the section. The program according to any one of claims 1 to 4, further comprising a computer that functions as summary content generation means for generating summary content in the moving image content by connecting a plurality of the scenes of interest. A number-of-posts graph representing the number of posts according to the passage of time, the number-of-posts graph generating means for generating a number-of-posts graph referring to the attention scene accumulating means and tagged with the comment keyword at the time of interest;
6. The program according to claim 1, further causing a computer to function as post number graph transmission means for transmitting the post number graph to the terminal. 6. Further causing the computer to function as a request keyword receiving means for receiving a request keyword from the terminal;
The program according to claim 6, wherein the posting number graph generating unit causes the computer to function such that only the comment keyword including the request keyword is tagged to the posting number graph. The program according to claim 7, wherein the posting number graph generating unit further causes the computer to generate a posting number graph representing only the number of postings of comment text including the request keyword. The program according to any one of claims 6 to 8, further causing a computer to function as attention scene transmission means for transmitting only the attention scene tagged with the request keyword to the terminal. 10. The computer according to claim 6, further causing a computer to function as comment text transmission means for transmitting the comment text of the scene of interest tagged with the request keyword received from the terminal to the terminal. 11. The program described in. The posting number graph generating means includes:
For the attention scene tagged with the comment keyword in the posting number graph, the ratio of the number of posting users to the total number of postings in the attention scene (the number of posting users / the total number of postings) is calculated.
The program according to any one of claims 6 to 10, wherein the computer is caused to function such that a comment keyword tagged to the attention scene having a higher ratio is visually highlighted. The posting number graph generating means includes:
For the attention scene tagged with the comment keyword in the post number graph, the ratio of the number of quoted posts to the total number of posts in the attention scene (quotation posts / total number of posts) is calculated.
The program according to any one of claims 6 to 10, wherein the computer is caused to function such that a comment keyword tagged to the attention scene having a higher ratio is visually highlighted. The video content is a television broadcast content or a video sharing site content,
The program according to any one of claims 1 to 12, wherein the comment server causes a computer to function as a mini web log server. The miniblog server is a twitter (registered trademark) server,
The program according to claim 13, wherein the comment collection unit causes a computer to collect comment text including a hash tag based on the moving image content from a twitter server. A server that receives video content that is disclosed to an unspecified number of third parties and extracts a scene of interest,
Comment collection means for collecting comment text including keywords related to the video content, together with a posting time, from a comment server;
A number-of-posts counting unit that divides the video content into scenes every predetermined time, and counts the number of comments posted in the time zone of the scenes;
Attention scene detection means for extracting a plurality of attention scenes in which the number of posts is a predetermined number or more;
Comment keyword extraction means for extracting a comment keyword included in a plurality of comment sentences in a plurality of comment sentences for each of the attention scenes;
Comment tagging means for tagging the comment keyword for each noted scene;
A server comprising: a featured scene distribution means for delivering the tagged scene of interest in response to a request from a terminal. A terminal connected to the network,
Video content receiving means for receiving video content to be disclosed to an unspecified number of third parties;
Comment collection means for collecting comment text including keywords related to the video content, together with a posting time, from a comment server;
A number-of-posts counting unit that divides the video content into scenes every predetermined time, and counts the number of comments posted in the time zone of the scenes;
Attention scene detection means for extracting a plurality of attention scenes in which the number of posts is a predetermined number or more;
Comment keyword extraction means for extracting a comment keyword included in a plurality of comment sentences in a plurality of comment sentences,
A terminal having comment tagging means for tagging the comment keyword for each scene of interest. A method of extracting a scene of interest in an apparatus for receiving video content that is disclosed to an unspecified number of third parties,
A first step of collecting comment text including a keyword related to the video content from a comment server together with a posting time;
A second step of dividing the video content into scenes at predetermined time intervals and counting the number of posts of the comment text posted in the time zone of the scene;
A third step of extracting a plurality of attention scenes in which the number of posts is a predetermined number or more;
A fourth step of extracting a comment keyword included in a plurality of comment sentences in a predetermined number or more for each of the attention scenes;
And a fifth step of tagging the comment keyword for each of the target scenes.

Images