JP2007140670A - Topic information providing method, device, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To construct necessary information according to comments such as the story of experience or personal experience at a site where a chat is available while video content is viewed. <P>SOLUTION: This topic information providing method includes acquiring a topic extracted from comments written in each scene, and applying multi-dimensional feature values and relax level to each viewer corresponding to the topic, and calculating the matrix(topic peculiarity) of center and weight in a multi-dimensional space with respect to the multi-dimensional feature values and relax level applied to each viewer corresponding to the topic, and a viewer dictionary is similarity-retrieved by using the topic and the peculiarity of the topic as a retrieval key, and the attributes of the topic offer other site having similarity trends are calculated, and the raw materials of a new topic are collected and output from a topic for relaxation by referring to an event storage utilization rule, based on the attributes and topic of the topic offer other side. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a topic information providing method, apparatus, and program, and particularly to the nature of comments left over a network when a viewer views video content in a place where the user can chat while viewing the video content. The present invention relates to a topic information providing method, apparatus, and program for providing community topic information in video content by making use of effects.

  There are many “Bulletin Boards”, “Chats” and “Blogs” in the cyber space, and viewers can communicate via the network while watching video content. In such an environment, users often write their thoughts in various places according to their own circumstances and preferences and use it as a place to expose themselves. As a result, other users often write comments and opinions on a large amount of documents (or comments) left in cyberspace, or spell similar events or experiences. . Here, talking about yourself is called “self-disclosure”. “Self-disclosure complementation” means that another person responds in the same way by speaking about their own things. For example, in response to the comment “I saw a person wearing a sock of a different color yesterday”, “I have also failed to wear a sock of a different color when I was in the morning” Self-disclosure and self-disclosure complementation will be developed, centering on experiences related to topics such as “wearing socks”. Cyberspace is said to be an environment in which self-disclosure and self-disclosure are easily performed (see, for example, Non-Patent Document 1).

  Self-disclosure is said to play an important role in the development of relationships that advance human relationships and foster trust. Further, it is considered that self-disclosure complements self-disclosure and works to strengthen intimacy and trust (for example, see Non-Patent Document 2).

  It is thought that a “good cycle” that promotes communication will be created by complementing self-disclosure and self-disclosure in the place of communication via many “bulletin board”, “chat” and “blog” networks in cyberspace.

  Since a large amount of documents (or comments) left in the “bulletin board”, “chat” or “blog” in cyberspace covers various topics, the topics are extracted from these large amounts of documents (or comments). There are several techniques to do this. Depending on the characteristics of the object to be processed, the effectiveness of each technology is demonstrated.

  For example, expressions matching topic patterns that specify expressions that can become topics from each document in the sentence group are extracted as topic candidates, and the scores given to each topic pattern and the importance of words in the topic candidates are selected. There is a technique for calculating a topic score representing the likelihood of a topic candidate based on the score, and selecting a topic candidate having the largest calculated topic score as a topic of a document group (for example, see Patent Document 1).

  In addition, an expression that matches a pre-given word string pattern is extracted as a topic candidate for a message that is a sentence group, and includes the number of messages that include the topic among messages that follow the message that includes the topic. There is a technique for obtaining the number of support that is the number of messages. This technology is effective in excluding lingering topics and rough messages, and can express the topic more specifically so that the viewer can easily understand the topic. It is possible to score a topic at a recent excitement level (see, for example, Patent Document 2).

  There are places where viewers can chat via the network while watching video content, in addition to the place of communication via the network such as “bulletin board”, “chat” and “blog”.

  For example, there is a technique for acquiring similar viewers associated with identification information and scene information of video content to be viewed. This technology integrates a system that searches for viewers who have similar interests in units of scenes of video content and a system that provides a place for communication between viewers (see, for example, Patent Document 3).

  In addition, there is a technique for acquiring similar viewers associated with identification information and scene information of video content to be viewed, and recommending video information stored with respect to similar viewer information to viewers (for example, Patent Documents). 4).

  In addition, there is a technique for rearranging comments displayed to the viewer in the order in which the comment target appears in the content (see, for example, Patent Document 5).

  In addition, there is a technique for grasping the viewer's preference with better accuracy and searching for a communication partner by pointing out a scene in which the viewer is interested and inputting a comment such as an evaluation (for example, see Patent Document 6).

  When viewers watch video content, they experience memories of what they have experienced in the past, memories of what they have experienced, such as by the behavior, facial expressions, and dialogue of the characters. It is thought that the memory of things is awakened, and the desire to talk to others or to convey in writing is inspiring. Chatting while viewing video content is thought to leave such comments in places. Comments written based on memories of experiences, memories of experiences, memories of experiences, etc. are referred to herein as experiences and experiences.

  There is a technology that focuses on experiential expression for general “bulletin board”, “chat” and “blog” in cyberspace. This technique is characterized by defining and using features that appear when describing the actual experience of the user as experience expressions (see, for example, Non-Patent Document 2). Chatting while viewing video content is considered to be an important part of creating a mood for the community while watching video content. .

  With regard to comments left over while viewing video content, experiences and experiences are considered a kind of self-disclosure, and communication that reacts to the experiences and experiences is a chat while viewing video content. It is considered a complement to self-disclosure. The chat environment where the viewers chat via the network while watching the video content is more than the chat environment via the network such as “bulletin board”, “chat” and “blog” that exist in cyberspace. It is considered an environment where it is easy to complement self-disclosure and self-disclosure.

  Experiences and experiences included in chats while viewing video content and communication that reacts to those experiences and experiences will inherit the complementary nature of self-disclosure and self-disclosure in cyberspace. It is done. In other words, the spoken stories and experiences play an important role in the development of relationships that advance similar relationships and foster trust, and respond to those experiences and experiences Communication is also thought to work to strengthen intimacy and trust. When chatting between viewers via a network while viewing video content, it is considered easier to provide a “good cycle” that promotes communication by complementing self-disclosure and self-disclosure.

  By making it easier for viewers to complement self-disclosure and self-disclosure due to the chat environment over the network while watching video content (in other words, making it easier to perform a “good cycle” in communication) However, it is considered that there is an effect that the viewers can further promote communication in the chatting place through the network while watching the video content.

  However, it is said that users of general “bulletin board”, “chat” and “blog” in cyberspace tend to be more self-referencing than being interested in “others”. (For example, refer nonpatent literature 4). For example, you can leave a lot of comments about what you have experienced or experienced as a result by writing down your experiences that you have experienced or experienced, and writing your own experiences that were remembered by other people's experiences. Often listed and accumulated. The same phenomenon often occurs when viewers chat with each other via a network while viewing video content. Due to the appearance of such a phenomenon, when reading accumulated comments, I feel relaxed and intimate, but because there are many similar topics, I feel that there are many `` stupid '' topics and participate in communication May reduce motivation.

In the case of personal experiences of viewers who have left behind their experiences and experiences, that is, comments such as those that have been left behind are “personal profiles like viewer characteristics (personality and attributes). It is thought to be captured as information. There is a technology that focuses on personal profiles (for example, information such as sex and age) regarding topics extracted from a large amount of documents (or comments) and the degree of interest of users. This technique is considered to aim at providing contents more suitable for the user by accumulating things such as user attributes. In this technology, the user's own profile is analyzed, and contents considered to be of interest to the user are collected and provided (see, for example, Patent Document 7).
"Topic Extraction Device, Topic Extraction Method and Program" Japanese Patent Application Laid-Open No. 2004-258723 “Topic Extraction Device, Topic Extraction Method, Program and Recording Medium” Japanese Patent Application Laid-Open No. 2004-348415 “Communication Support System, Method and Apparatus Between Video Content Viewers, and Communication Support Program and Program Medium” JP2003-87826A “Video Information Recommendation System, Method and Apparatus, and Video Information Recommendation Program and Program Recording Medium” Japanese Patent Application Laid-Open No. 2003-85207 “Content-linked comment display method, comment display system, server device, comment display device, and comment display program” Japanese Patent Application Laid-Open No. 2004-364184 “Video Content Viewer Information Providing System and Method, Viewer Information Providing Apparatus, Program, and Program Recording Medium” JP2003-76699A JP 2003-223406 A "Behavior and psychology on the Internet" by AN Joyson, p142 “Behavior and Psychology on the Internet” by AN Joyson, p147 Ikeda, Kayo et al. "Blog reputation information judging method with experience expression", IEICE, Type 2 study material, W12-2005-28-49, Web intelligent and interaction "Behavior and psychology on the Internet" by AN Joyson, p34)

  When viewers chat with each other over a network while watching video content, unwritten comments (especially unwritten experiences and experiences) are the viewer's personality and attributes (for example, gender, age, etc.) ). Commentary such as experiences and experiences, and information such as “personal profile” such as the personality and attributes of viewers are combined to gain a clearer understanding of viewer characteristics and interests, and a wide variety of experiences By accumulating a lot of comments, such as experiences and experiences, we have developed a complementary cycle of self-disclosure and self-disclosure that has the effect of attracting “others” and provided a “good cycle” that encourages motivation to participate in communication It is considered possible.

  However, it is not easy to link to personal profile information such as the characteristics (personality and attributes) of viewers just by looking at information such as experiences and experiences from a large number of unwritten comments. There is a problem.

  For example, a certain video content is divided into several scenes (scenes) as in “image of input example” shown in FIG. 4 to be described later. "Structure of media", IEICE Pattern Recognition Media Understanding Study Group, Special Lecture PRMU99-42 (1999-07)). For each scene (scene), the comments of the viewers are left behind, the comments that are left behind are read by other viewers, and the comments are left unwritten. Talk and communicate.

  For example, as shown in FIG. 4, the video content is divided into three scenes (scenes), each scene (scene) is a, b, c. Comments are shown in Fig. 5 by communicating with each other and communicating with each other in a manner such as writing comments to other viewers and reading them to other viewers and writing comments. Is acquired.

  The topic extraction means extracts topics as shown in FIG. 6 from the comments left in the three scenes of the video content. Here, WKT (eg, Expo) is a story about experiences and experiences related to viewers. W12, W21, and W32 are topics of comments that are left unwritten for each scene. Here, it is assumed that the topic WKT is extracted in the a scene and the topic WKT is extracted in the b scene. Actually, the following topics are left in each scene.

a scene: WKT, W12
b scene: W21, WKT
c scene: W31
In addition, regarding the WKT topic of the a scene, the topic extraction result of FIG. 6 is extracted in more detail, and the pattern of the topic of (WKT + W1) is extracted, and the pattern of the topic of (WKT + W2) is extracted regarding the WKT topic of the b scene. It was. The topic pattern expresses the topic content as a word string. In the following, WKTP1 and WKTP2 represent topic patterns, and WKT topics for a scene and b scene are represented as follows.

a Scene: WKT → WKTP1 → (WKT + W1)
bScene: WKT → WKTP2 WKTP3 → (WKT + W2)
(W1 and W2 are notation of a word)
It is thought that the topic patterns of the above WKTP1 and WKTP2 are closely related to the characteristics (properties and attributes) of the viewers related to the topic.

  The topic of WKT appears in both scenes a and b of the video content. When the details of the WKT topic are examined, it can be divided into the topic patterns of WKPT1 (WKT + W1) and WKTP2 (WKT + W2). In other words, WKT topics appear in the two scenes, suggesting that there is a difference in topic development.

  R11, R12, and R13 participate in the WKT topic in the a scene, and R21, R22, and R23 participate in the WKT topic in the b scene. However, the following viewers are involved in WKT topics between viewers while watching video content.

  There is a problem that the WKT topic patterns (WKPT1, WKTP2) cannot easily be read as to what other viewers are interested in. In other words, just by looking at the comments such as unwritten experience stories and experience stories, the information of those topic patterns is linked to "personal profile" information such as the nature and attributes of any other viewer, It is not easy to grasp the characteristics and interests of viewers and develop topics (so that a lot of comments such as various experiences and experiences can be accumulated). If the topic can be further developed, self-disclosure that interests “others” and complement of self-disclosure can be developed, and a “good cycle” that promotes willingness to participate in communication can be provided.

  When viewers watch video content, the behavior, display, and dialogue of the characters will evoke memories that the viewer has experienced in the past, memories that have been tasted, memories that have been experienced, etc. There are many comments left in places where people chat or chat via the network while watching video content. However, the nature and effectiveness of these comments (experiences and experiences) often end without being utilized.

  The present invention has been made in view of the above points, and in a place where video content can be viewed while chatting, necessary information (characteristics and attributes of viewers) is obtained from comments such as experiences and experiences. It is an object of the present invention to provide a topic information providing method, apparatus, and program capable of constructing "personal profile information" or "community profile information such as community characteristics".

  FIG. 1 is a diagram for explaining the principle of the present invention.

The present invention (Claim 1) is a topic information providing method for supporting activation of communication between viewers when chatting via a network while viewing video content,
In the storage means,
Multi-dimensional feature (feature vector) for each viewer corresponding to a topic that relaxes communication between viewers such as experiences and experiences that have a significant impact on the mood of the corresponding community. And a viewer feature amount assigning procedure for giving a degree of relaxation (score) of the corresponding viewer,
A community feature extraction procedure to calculate a matrix of center points and weights (topic buzz) in the multidimensional space of the topic in question,
Event memory utilization rules for collecting new topic material from the attributes of topic providers with similar tendencies and topics to relax for the calculated topical traps,
And store
When a topic for relaxing communication extracted from a comment written for each scene of video content is acquired, the topic for relaxing the viewer feature amount stored in the storage means is referred to. Viewer feature amount giving step (step 1) for giving a multidimensional feature amount and a degree of relaxation for each corresponding viewer;
Refer to the community feature extraction procedure stored in the storage means for the multidimensional feature value and the degree of relaxation given to each viewer corresponding to the topic to be relaxed, and the center point in the multidimensional space A community feature calculation step (step 2) for calculating a weight matrix (topic buzz);
A topic provider search step (step 3) that searches the viewer dictionary in which viewer information is stored using the topic for relaxation and the topic trap as a search key, and obtains attributes of the topic provider having a similar tendency. When,
Based on the attributes of the topic-providing partner and the topic to be relaxed, refer to the event memory utilization rules stored in the storage means, and collect the material for the ware fee that collects new topic material from the topic to be relaxed Step (step 4);
An output step (step 5) for outputting the collected topical material is performed.

Further, the present invention (Claim 2) provides a viewer feature amount providing procedure stored in the storage means.
For viewers corresponding to the topic to be relaxed, many viewers corresponding to the topic to be relaxed are used by using the viewer's degree of view to the video content and the instantaneous audience rating of each scene in the video community. Using a dimension feature and the pictograph used by the viewer for the comment,
In the viewer feature amount assigning step,
With reference to the viewer feature amount assigning procedure, the number of pictograms used by the viewer in the comment is obtained as the degree of relaxation of the viewer.

  FIG. 2 is a principle configuration diagram of the present invention.

The present invention (Claim 3) is a topic information providing apparatus for supporting activation of communication between viewers when chatting via a network while viewing video content,
Multi-dimensional feature (feature vector) for each viewer corresponding to a topic that relaxes communication between viewers such as experiences and experiences that have a significant impact on the mood of the corresponding community. And a viewer feature amount assigning procedure storage unit 331 for storing a viewer feature amount assigning procedure for assigning the degree of relaxation (score) of the corresponding viewer,
Community feature extraction procedure storage means 332 storing a community feature extraction procedure for calculating a matrix of the center point and weight (topic buzz) in a multidimensional space of topics for relaxation;
Event storage utilization rule storage means 352 storing event storage utilization rules for collecting new topic material from the attributes of the topic providing partner having a similar tendency and the topics for relaxation with respect to the calculated topic trap. And having.

The present invention (Claim 4) is a topic information providing device for supporting activation of communication between viewers when chatting via a network while viewing video content,
Multi-dimensional feature (feature vector) for each viewer corresponding to a topic that relaxes communication between viewers such as experiences and experiences that have a significant impact on the mood of the corresponding community. And a viewer feature amount assigning procedure storage unit 331 for storing a viewer feature amount assigning procedure for assigning the degree of relaxation (score) of the corresponding viewer,
Community feature extraction procedure storage means 332 storing a community feature extraction procedure for calculating a matrix of center points and weights (topic buzz) in a multidimensional space of the relevant topic;
Event storage utilization rule storage means 352 storing event storage utilization rules for collecting new topic material from the attributes of the topic providing partner having a similar tendency and the topics for relaxation with respect to the calculated topic trap. When,
A viewer dictionary 351 storing viewer feature values and attributes corresponding to video content scenes;
When a topic for relaxing the communication extracted from the comments written for each scene of the video content is acquired, the viewer corresponding to the topic for relaxing is referred to the viewer feature amount assigning procedure storage unit 331. Viewer feature amount giving means 311 for giving a multidimensional feature amount and a degree of relaxation for each,
With reference to the community feature extraction procedure storage means 332 for the multidimensional feature value and the degree of relaxation given to each viewer corresponding to the topic to be relaxed, a matrix of center points and weights in the multidimensional space ( Community feature calculating means 312 for calculating
Topic provider search means 321 that searches the viewer dictionary 351 in a similar manner using the topic's trap as a search key, and obtains attributes of the topic provider having a similar tendency;
Based on the attribute of the topic providing partner and the topic for relaxing, the event storage utilization rule storage unit 352 is referred to, and the topic providing material collecting unit 322 that collects a new topic material from the topic for relaxing,
Output means 4 for outputting collected topical material.

Further, according to the present invention (Claim 5), in the viewer feature amount assigning procedure storage unit 331,
For viewers corresponding to the topic to be relaxed, many viewers corresponding to the topic to be relaxed are used by using the viewer's degree of view to the video content and the instantaneous audience rating of each scene in the video community. Using a dimension feature and the pictograph used by the viewer for the comment,
The viewer feature amount giving means 311
With reference to the viewer feature amount assigning procedure storage unit 331, the number of pictographs used by the viewer in the comment is obtained as the degree of relaxation of the viewer.

The present invention (Claim 6) is provided for each viewer corresponding to a topic for relaxing communication between viewers such as experiences and experiences that greatly affect the mood of the corresponding community. Viewer feature value giving procedure storage means for storing a viewer feature value giving procedure for giving a multidimensional feature value (feature vector) and the degree of relaxation (score) of the corresponding viewer;
A community feature extraction procedure storage means storing a community feature extraction procedure for calculating a center point and a weight matrix (topic buzz) in a multidimensional space of the relevant topic;
Event memory utilization rule storage means for storing event storage utilization rules for collecting new topic material from the attributes of topic providers having similar tendencies and topics for relaxing with respect to the calculated topic trap ,
A viewer dictionary storing viewer features and attributes corresponding to video content scenes;
On a computer with
When the topic for relaxing the communication extracted from the comment written for each scene of the video content is acquired, each viewer corresponding to the topic for relaxing is referred to the viewer feature amount adding procedure storage means. A viewer feature amount adding step for providing a multidimensional feature amount and a degree of relaxation to
Refer to the community feature extraction procedure storage means for the multidimensional feature and relaxation degree given to each viewer corresponding to the topic to be relaxed, and the matrix of center points and weights in the multidimensional space (topic A community feature calculating step for calculating
Using a topic trap as a search key, a similar search is performed in the viewer dictionary, and a topic provider search step for obtaining attributes of a topic provider having a similar tendency;
Based on the attribute of the topic providing partner and the topic for relaxing, the event memory utilization rule storage means is referred to, and the wear fee providing material collecting step for collecting new topic material from the topic for relaxing,
And a topic information providing program for executing an output step of outputting collected topic material.

Further, the present invention (Claim 7) provides a viewer feature amount providing step,
For viewers corresponding to the topic to be relaxed, many viewers corresponding to the topic to be relaxed are used by using the viewer's degree of view to the video content and the instantaneous audience rating of each scene in the video community. Using the dimension feature and the pictograph used by the viewer for the comment, referring to the viewer feature amount assigning procedure storage means including the procedure for giving the viewer a degree of relaxation, the viewer used the comment The number of pictographs is obtained as the degree of relaxation of the viewer.

  As described above, according to the present invention, communication can be relaxed among viewers by using the properties and effects of comments left over when viewers chat via a network while viewing video content. In this place, it becomes possible to obtain information on video content and promote viewing of the video content.

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

  The present invention presupposes a topic extraction technique for relaxing communication between viewers such as experiences and experiences. The technology detects a community viewing average scene using the number of viewing start accesses to a scene of a video content of a viewer, and extracts a corresponding topic based on a distribution of topics appearing in the detected community viewing average scene. Is.

  FIG. 3 shows an apparatus configuration according to an embodiment of the present invention.

  The topic information providing apparatus shown in FIG. 1 includes an input unit 1, a topic processing unit 2, a topic providing support unit 3, a topic extraction dictionary 211, a topic extraction result file 221, a viewer feature amount assignment procedure storage unit 331, and a community feature calculation procedure. It is composed of a storage unit 332, a viewer feature amount assignment result file 341, a community feature calculation result file 342, a viewer dictionary 351, an event storage utilization rule storage unit 352, a topic provision partner result file 361, and a topic provision material collection result file 362. The

  The input unit 1, the topic extraction dictionary 211, the topic extraction result file 221, and the output unit 4 are connected to the topic processing unit 2.

  The topic processing unit 2 includes a topic extracting unit 21 and a topic providing unit 22, which are connected to the topic providing support unit 3. The topic processing unit 2 is a component corresponding to the topic extraction technique described above.

  The topic provision support unit 3 includes a community mood detection unit 31 and an event storage utilization unit 32. The community mood detection unit 31 includes a viewer feature amount giving unit 311 and a community feature calculation unit 312.

  The viewer feature amount assignment procedure storage unit 331 stores a viewer feature amount provision procedure, the community feature calculation procedure storage unit 332 stores a community feature calculation procedure, and the event storage utilization rule storage unit 352 stores Stores event memory utilization rules.

  The viewer feature amount adding procedure is for viewers who respond to topics that relax the communication between viewers such as experiences and experiences that have a great influence on the mood of the corresponding community. This is a rule for assigning a multidimensional feature quantity (feature quantity vector) and the degree of relaxation (score) of the corresponding viewer. Specifically, for viewers who deal with topics such as experiences and experiences that relax communication between viewers, viewers' views on video content and each scene in the video community This is a procedure for assigning a feature amount using the instantaneous audience rating. In addition, the viewer feature amount assigning procedure is performed by using the multidimensional feature amount and pictograms used in the comments written by the corresponding viewer for the viewers corresponding to the topic to be relaxed. It also includes a procedure for providing a degree of relaxation.

  The community feature calculation procedure refers to the multi-dimensional feature value given to viewers who respond to topics for relaxing communication between viewers such as experience stories and experience stories, and the degree of relaxation of the corresponding viewers. This is a procedure for calculating a matrix of center points and weights in a multidimensional space in the relevant topic.

  The event memory utilization rule is a rule for collecting new topic material from the attribute of the topic provider who has a similar tendency and the topic to be relaxed against the calculated trap of the topic to be relaxed. is there. Specifically, a similarity search is performed on the viewer dictionary 351 in which viewer information is stored in advance using the calculated topical heel for relaxation as a search key, and attributes (age, gender) of the topic provider Etc.) and rules for collecting new topical material from topics for relaxing.

  FIG. 4 shows an input example input to the input unit 1 of the topic information providing apparatus, and FIG. 5 shows an input example of a comment for each partial scene to be input.

  Hereinafter, for the topic extraction result example shown in FIG. 6 based on the input shown in FIG. 5, when chatting via the network while viewing the video content, the topic information provision of the community in the video content will be described. A form is demonstrated.

  FIG. 7 is a diagram for explaining the overall operation in one embodiment of the present invention.

  FIG. 8 is a flowchart of the operation of the topic processing unit in one embodiment of the present invention.

  Step 101) When the comment shown in FIG. 5 is input to the input unit 1, it is sent to the topic processing unit 2. The topic extraction unit 21 of the topic processing unit 2 refers to the topic extraction dictionary 211 and has a great influence on the mood of the corresponding community in the comment for each scene for the input comment (comment in FIG. 5). A topic for relaxing communication between viewers such as experiences and experiences is extracted, a topic extraction result file 221 as shown in FIG. 6 is created, and the extracted topics are transferred to the topic providing unit 22. To do. In order to perform such processing, it is assumed that the topic extraction dictionary 211 has registered therein contents capable of extracting a topic for relaxing communication between viewers.

  Step 102) The topic providing unit 22 transfers the data of the topic extraction result file 221 to the community mood detecting unit 31 of the topic providing support unit 3. Thereby, the community mood detecting unit 31 acquires the topic in the viewer feature amount providing unit 311.

  Step 103) Further, the topic providing unit 22 outputs the data of the topic extraction result file 221 to the output unit 4.

  For example, the topic processing unit 2 refers to the topic extraction dictionary 211 with respect to the comment input example shown in FIG. 5 in the topic extraction unit 21, from the comments written in the scene a, the scene b, and the scene c. 6 is extracted (topics for relaxation) such as topic extraction result examples (notation expressing topics for each scene, topic patterns, sentence numbers, viewer names), and the results are extracted in the topic extraction result file 221. It is stored and transferred to the viewer feature amount assigning unit 311 of the communication mood detecting unit 31 together with the number of scene viewing start accesses shown in FIG.

  Next, the topic providing support unit 3 will be described.

  First, operation | movement of the communication mood detection part 31 of the topic provision assistance part 3 is demonstrated.

  FIG. 10 is a flowchart of the operation of the communication mood detection unit in one embodiment of the present invention.

  Step 201) The viewer feature amount giving unit 311 of the communication mood detecting unit 31 targets the viewer start feature amount for each scene viewing start access number shown in FIG. 9 transferred from the topic providing unit 22 of the topic processing unit 2. With reference to the viewer feature value adding procedure in the assigning procedure storage unit 331, the multi-dimensional feature value and the degree of relaxation of the corresponding viewer are given to the viewers corresponding to the topic for relaxing, and the result is viewed. Stored in the person feature value assignment result file 341. Further, the viewer feature amount assigning unit 311 transfers the topic extraction result acquired from the topic providing unit 22 and the data stored in the viewer feature amount giving result file 341 to the community feature calculating unit 312.

  Step 202) The community feature calculation unit 312 uses the topic extraction result transferred from the viewer feature amount assigning unit 311 and the data of the viewer feature amount assignment result as a target for the community feature calculation procedure of the community feature calculation procedure storage unit 332. To the multi-dimensional features assigned to the viewers that correspond to topics for relaxing communication between viewers such as experiences and experiences, and the degree of relaxation of the viewers The matrix of the center point and the weight in the multidimensional space is calculated, and the result is stored in the community feature calculation result file 342.

  Step 203) Further, the community feature calculation unit 312 transfers the data of the community feature calculation result file 342 to the topic offer partner search unit 321 of the event storage utilization unit 32 together with the topic extraction result.

  Next, the operation of the event storage utilization unit 32 will be described.

  FIG. 11 is a flowchart of the operation of the event storage utilization unit in one embodiment of the present invention.

  Step 301) The topic offer partner search unit 321 of the event storage utilization unit 32 uses the topic feature calculation result transferred from the community mood detection unit 31 and the topic extraction result, and a topic trap for relaxing (community feature calculation). The center point, weight matrix and topic (word string notation: “WKT + W1”, for example) calculated by the section 312 are calculated, and viewer information is stored using the topic trap as a search key. A similar search is performed on the dictionary 351 for selecting a topic providing partner, and the attributes of the topic providing partner having a similar tendency (for example, age, sex) and a corresponding relaxing topic are extracted. The result is stored in the topic providing partner result file 361 and the data of the topic partner result file 361 is transferred to the topic providing material collection unit 322. Here, the similarity search is to perform the distance calculation by inputting the center point and the weight matrix into a calculation formula described later.

  Step 302) The topic providing material collection unit 322 uses the event storage utilization rule storage unit 352 for the topic providing partner result, and from the topic for relaxing the topic providing partner attribute (for example, age, gender, etc.). A new topic material is collected, and the result is stored in the topic providing material collection result file 362 and transferred to the output unit 4.

  Hereinafter, a specific example of the above operation will be described with reference to FIGS.

First, the viewer feature amount assigning unit 311 of the community mood detecting unit 31 receives the multi-dimensional feature amount and the degree of relaxation of the corresponding viewer from the input shown in FIG. 9 to the viewers corresponding to the topic for relaxing. Is given (step 201). At this time, using the viewer feature value assigning procedure in the viewer feature value assigning procedure storage unit 331 shown in FIG.
(1) Targeting the data of the topic extraction result file 221 acquired from the topic providing unit 22, the audience shown to the viewer's video content from the data shown in FIG. Calculate the instantaneous audience rating.

Here, the viewership of the video content of a certain viewer is
The audience rating is obtained by the number of access start scenes of a viewer / the number of scenes of video content, and the instantaneous audience rating of each scene in the video community is
Instantaneous viewing rate = number of viewing start accesses of a scene in a certain time zone / total number of viewing start accesses of video content in a certain time zone.

  (2) The audience rating and the instantaneous audience rating calculated above are assigned as feature quantities for each viewer related to the relevant relaxation topic pattern, and stored in the audience feature quantity assignment result file 341 as shown in FIG. .

  Further, as shown in FIG. 12 (3), the number of pictograms used in the relevant relaxation topic can be given as the degree of relaxation (score) of the relevant viewer. The initial score is 1.

Next, the community feature calculation unit 312 of the community mood detection unit 31 stores the data of the viewer feature amount assignment result file 341 shown in FIG. 13 in the community feature calculation procedure storage unit 332 as shown in FIG. Using the community feature calculation procedure, for example, the feature amount of one viewer R11 of the “WKT + W1” topic is 0.133,..., 0.273, 0.292, 0.500,. Similarly, the feature values of the other viewers R12 and R13 are read from the viewer feature value assignment result file 341, and are used to determine the center point q and multidimensional in the multidimensional space in the “WKT + W1” topic. A matrix of weights in the space is calculated, and the result (for example, the center point q1 and the weight C _jk 1) is stored in the community feature calculation result file 342 as shown in FIG. 15, and the topic extraction result as shown in FIG. At the same time, it is transferred to the topic provider search unit 321 of the event storage utilization unit 32 (steps 202 and 203).

The center point q in the above multi-dimensional space, the weight C _ij in multidimensional space is calculated as follows (for details, reference "Yoshiharu Ishikawa, Ravishankar subramanya, Christos Faloutsos:" MindReader "" Database query method based on example ”: Information Processing Society of Japan Research Report, Vol.1998, No.57, 98-DBS-116-39, pp.93-100, Information Processing Society of Japan, July 1998”.

イベント記憶活用部３２では、コミュニティ特徴算出部３１２から転送された図１５に示すコミュニティ特徴算出結果ファイル３４２のデータと、図６に示すような話題抽出結果を対象に、話題提供相手検索部３２１において、算出された「WKT+W1」話題と、当該話題における多次元空間での中心点ｑ、多次元空間での重みＣ_ｉｊの行列（リラックスさせるための話題の癖）を検索キーとして、図１６に示すような視聴者辞書３５１を類似検索する。ここで、視聴者辞書３５１は、シーン番号毎に、当該シーンに関わる視聴者、視聴度、瞬間視聴率等からなる特徴量と、性別や年齢等の属性から構成される。

In the event storage utilization unit 32, the topic provider search unit 321 uses the data of the community feature calculation result file 342 shown in FIG. 15 transferred from the community feature calculation unit 312 and the topic extraction result shown in FIG. 16 using the calculated “WKT + W1” topic, the center point q of the topic in the multidimensional space, and the matrix of the weights C _{ij in} the multidimensional space (the topic trap for relaxing) as search keys. A similar search is performed on the viewer dictionary 351 as shown in FIG. Here, the viewer dictionary 351 includes, for each scene number, a feature amount including a viewer, audience rating, instantaneous audience rating, and the like related to the scene, and attributes such as sex and age.

  The topic provider search unit 321 extracts, for example, information on a topic provider (viewer RR7) having a similar tendency that is ranked higher, the attribute of which is “male”, and “age 20”. The result (topic = “WKT + W1”, topic providing partner = RR7, gender = male, age = 20's) is stored in the topic providing partner result file 361 and transferred to the topic providing material collecting unit 322 as shown in FIG. To do.

  Here, as a method of similarity search, “Keiji Ishikawa, Ravishankar Subramanya, Chistos Faloutsos:“ MindReader ”:“ Database Query Method Based on Examples ”, Information Processing Society of Japan Research Report, Vol. 1998, No. 57, 98-DBS- 116-39, pp.93-100, Information Processing Society of Japan, July 1998 ”. In this reference, the following calculation algorithm is used to obtain the distance calculation D for the similarity search.

次に、話題提供素材収集部３２２は、図１８に示すようなイベント記憶活用規則記憶部３５２に格納されているイベント記憶活用規則を用いて、図１７のような話題提供相手結果ファイル３６１のデータに対して、話題提供相手の属性（例えば、年齢、性別等）とリラックスさせるための話題に基づいて、新たな話題の素材（のアドレス）をネットワークを介して収集し（ステップ３０２）、収集した結果の件数と共に、図１９に示すような話題提供素材収集結果ファイル３６２に格納すると共に、出力部４に転送する（ステップ３０３）。図１９の例では、視聴者のRR7とR31は共に、図５では、『万博』(WKT)という話題に参加していないが、視聴者辞書３５１の検索により、類似するため、それぞれの話題の素材が収集されている。

Next, the topic provision material collection unit 322 uses the event storage utilization rule stored in the event storage utilization rule storage unit 352 as shown in FIG. On the other hand, new topic materials (addresses) were collected via the network (step 302) based on the attributes of the topic provider (eg, age, gender, etc.) and the topic to be relaxed. Together with the number of results, it is stored in a topic providing material collection result file 362 as shown in FIG. 19 and transferred to the output unit 4 (step 303). In the example of FIG. 19, the viewers RR7 and R31 are not participating in the topic “Expo” (WKT) in FIG. Material is collected.

  In this way, by managing the characteristics of the relevant viewers and their viewers, such as experiences and experiences that have a significant impact on the mood of the community, the topics related to relaxing communication between viewers are applicable. Understand the trend of viewers related to the topic, link the information of the topic pattern with other "personal profile" information such as personality and attributes of any viewer, grasp the characteristics and interests of the viewer, develop the topic (To make it possible to accumulate a wide variety of experiences and comments such as experiences), to develop self-disclosure and self-disclosure complementary to interest in "others", and willingness to participate in communication Provide a “good cycle” that is promoted.

  The output unit 4 displays the topic providing material collection result of FIG. 19 on the screen.

  In addition, the operation of the topic information providing apparatus in the above embodiment is constructed as a program, and has at least a viewer feature adding procedure storage unit, a community feature calculation procedure storage unit, a viewer estate, and an event storage utilization rule storage unit. It can be installed and executed on a computer used as a topic information providing apparatus, or distributed via a network.

  In addition, the constructed program can be installed or distributed on a hard disk drive or a portable storage medium such as a flexible disk / CD-ROM.

  The present invention is not limited to the above-described embodiment, and various modifications and applications can be made within the scope of the claims.

  The present invention is applicable to a communication system via a network.

It is a figure for demonstrating the principle of this invention. It is a principle block diagram of this invention. It is an apparatus block diagram in one embodiment of this invention. It is an image (scene and comment) of the input example in one embodiment of this invention. It is an example of comment input of each scene in one embodiment of the present invention. It is an example of a topic extraction result in one embodiment of the present invention. It is a figure which shows the flow of the data in one embodiment of this invention. It is a flowchart of operation | movement of the topic process part in one embodiment of this invention. It is an example of each scene viewing start access number input in one embodiment of the present invention. It is a flowchart of operation | movement of the communication mood detection part in one embodiment of this invention. It is a flowchart of operation | movement of the event memory utilization part in one embodiment of this invention. It is an example of description of the viewer feature-value provision procedure in one embodiment of this invention. It is an example of description of the viewer feature-value provision result file in one embodiment of this invention. It is an example of the community feature calculation procedure in one embodiment of the present invention. It is an example of the community characteristic calculation result file in one embodiment of this invention. It is an example of the viewer dictionary in one embodiment of this invention. It is an example of the topic provision partner result file in one embodiment of this invention. It is an example of the event memory utilization rule in one embodiment of this invention. It is an example of a topic provision material collection result in one embodiment of the present invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Input part 2 Topic process part 3 Topic provision assistance part 4 Output part 21 Topic extraction part 22 Topic provision part 31 Community mood detection part 32 Event memory utilization part 211 Topic extraction dictionary 221 Topic extraction result file 311 Viewer feature-value provision part 312 Community feature calculation unit 321 Topic offer partner search unit 322 Topic offer material collection unit 331 Viewer feature amount assignment procedure storage unit 332 Community feature calculation procedure storage unit 341 Viewer feature amount assignment result file 342 Community feature calculation result file 351 Viewer dictionary 352 Event rule utilization rule storage unit 361 Topic provision partner result file 362 Topic provision material collection result file

Claims (7)

A topic information providing method for supporting activation of communication between viewers when chatting via a network while watching video content,
In the storage means,
Multi-dimensional feature (feature vector) for each viewer corresponding to a topic that relaxes communication between viewers such as experiences and experiences that have a significant impact on the mood of the corresponding community. And a viewer feature amount assigning procedure for giving a degree of relaxation (score) of the corresponding viewer,
A community feature extraction procedure to calculate a matrix of center points and weights (topic buzz) in the multidimensional space of the topic in question,
Event memory utilization rules for collecting new topic material from the attributes of topic providers with similar tendencies and topics to relax for the calculated topical traps,
And store
When a topic for relaxing communication extracted from comments written for each scene of video content is obtained, the viewer feature amount adding procedure stored in the storage means is referred to Viewer feature amount giving step for giving a multidimensional feature amount and a degree of relaxation for each viewer corresponding to the topic,
With reference to the community feature extraction procedure stored in the storage means for the multidimensional feature value and the degree of relaxation given to each viewer corresponding to the topic to be relaxed, a multidimensional space A community feature calculation step of calculating a matrix of center points and weights at
A topic providing partner search step for performing a similar search on a viewer dictionary in which viewer information is stored using the topic for relaxing and the topic 癖 as a search key, and obtaining an attribute of the topic providing partner having a similar tendency;
Based on the attribute of the topic providing partner and the topic to be relaxed, referring to the event memory utilization rules stored in the storage means, and a wear fee for collecting new topic material from the topic to be relaxed Provide material collection step,
An output step for outputting the collected material of the topic;
A topic information providing method characterized by: The viewer feature amount giving procedure stored in the storage means is:
For the viewer corresponding to the topic for relaxation, the viewer corresponding to the topic for relaxation using the viewer's degree of view to the video content and the instantaneous audience rating of each scene in the video community. Using a multi-dimensional feature amount and pictographs used by the viewer for comments, including a step of giving the viewer a degree of relaxation;
In the viewer feature amount giving step,
With reference to the viewer feature amount assignment procedure, the number of pictograms used by the viewer in the comment is obtained as the degree of relaxation of the viewer.
The topic information providing method according to claim 1. A topic information providing device for supporting the activation of communication between viewers when chatting via a network while watching video content,
Multi-dimensional feature (feature vector) for each viewer corresponding to a topic that relaxes communication between viewers such as experiences and experiences that have a significant impact on the mood of the corresponding community. Viewer feature value giving procedure storage means for storing a viewer feature value giving procedure for giving the degree of relaxation (score) of the corresponding viewer;
Community feature extraction procedure storage means for storing a community feature extraction procedure for calculating a matrix of center points and weights (topic traps) in a multidimensional space of topics for relaxation;
An event that stores the rules of event storage to collect new topic material from the attributes of the topic provider with similar tendencies and topics to relax, for the topic of relaxation and the calculated topic trap Memory utilization rule storage means;
A topic information providing device characterized by comprising: A topic information providing device for supporting the activation of communication between viewers when chatting via a network while watching video content,
Multi-dimensional feature (feature vector) for each viewer corresponding to a topic that relaxes communication between viewers such as experiences and experiences that have a significant impact on the mood of the corresponding community. Viewer feature value giving procedure storage means for storing a viewer feature value giving procedure for giving the degree of relaxation (score) of the corresponding viewer;
Community feature extraction procedure storage means for storing a community feature extraction procedure for calculating a matrix of center points and weights (topic traps) in a multidimensional space of topics for relaxation;
An event that stores the rules of event storage to collect new topic material from the attributes of the topic provider with similar tendencies and topics to relax, for the topic of relaxation and the calculated topic trap Memory utilization rule storage means;
A viewer dictionary storing viewer features and attributes corresponding to video content scenes;
When a topic for relaxing the communication extracted from the comment written for each scene of the video content is acquired, the viewer corresponding to the topic for relaxing is referred to the viewer feature amount assigning procedure storage means Viewer feature amount giving means for giving a multidimensional feature amount and a degree of relaxation for each,
Refer to the community feature extraction procedure storage means for the multidimensional feature value and the degree of relaxation given to each viewer corresponding to the topic to be relaxed, and the center point and weight in the multidimensional space Community feature calculation means for calculating the matrix of
The topic providing partner search means for performing a similar search in the viewer dictionary using the topic for relaxing and the topic trap as a search key, and obtaining an attribute of the topic providing partner having a similar tendency,
Based on the topic providing partner attribute and the topic for relaxing, referring to the event storage utilization rule storage means, topic providing material collecting means for collecting new topic material from the topic for relaxing,
Output means for outputting the collected material of the topic;
A topic information providing device characterized by comprising: The viewer feature amount assigning procedure storage means includes:
For the viewer corresponding to the topic for relaxation, the viewer corresponding to the topic for relaxation using the viewer's degree of view to the video content and the instantaneous audience rating of each scene in the video community. Using a multi-dimensional feature amount and pictographs used by the viewer for comments, including a step of giving the viewer a degree of relaxation;
The viewer feature amount giving means includes:
With reference to the viewer feature amount assigning procedure storage means, the number of pictograms used by the viewer in the comment is obtained as the degree of relaxation of the viewer.
The topic information providing apparatus according to claim 4. Multi-dimensional feature (feature vector) for each viewer corresponding to a topic that relaxes communication between viewers such as experiences and experiences that have a significant impact on the mood of the corresponding community. Viewer feature value giving procedure storage means for storing a viewer feature value giving procedure for giving the degree of relaxation (score) of the corresponding viewer;
A community feature extraction procedure storage means storing a community feature extraction procedure for calculating a center point and a weight matrix (topic buzz) in a multidimensional space of the relevant topic;
Event memory utilization rule storage means for storing event storage utilization rules for collecting new topic material from the attributes of topic providers having similar tendencies and topics for relaxing with respect to the calculated topic trap ,
A viewer dictionary storing viewer features and attributes corresponding to video content scenes;
On a computer with
When a topic for relaxing communication extracted from a comment written for each scene of video content is acquired, a viewer corresponding to the topic for relaxing is referred to the viewer feature amount adding procedure storage means Viewer feature amount assigning step for giving a multidimensional feature amount and a degree of relaxation for each,
Refer to the community feature extraction procedure storage means for the multidimensional feature value and the degree of relaxation given to each viewer corresponding to the topic to be relaxed, and the center point and weight in the multidimensional space Community feature calculation step for calculating the matrix of
A topic providing partner search step for performing a similar search in the viewer dictionary using the relaxation topic and the topic trap as a search key, and obtaining attributes of the topic providing partner having a similar tendency;
Based on the attribute of the topic providing partner and the topic to be relaxed, referring to the event storage utilization rule storage means, and collecting a material for providing a new topic from the topic for relaxing ,
An output step for outputting the collected material of the topic;
Topic information providing program characterized by running In the viewer feature amount giving step,
For the viewer corresponding to the topic for relaxation, the viewer corresponding to the topic for relaxation using the viewer's degree of view to the video content and the instantaneous audience rating of each scene in the video community. Using the multi-dimensional feature value and the pictograph used by the viewer for the comment, referring to the viewer feature value adding procedure storage means including the procedure for giving the viewer a degree of relaxation, the viewer uses the comment The number of pictographs made is calculated as the degree of relaxation of the viewer.
The topic information providing program according to claim 6.

Images