JP5439050B2 - Related content display device and computer program - Google Patents

Description

  The present invention relates to a system for accumulating and viewing a large number of video contents such as broadcast programs in order to search and present related contents in order to support a user to view desired video contents at a desired time. The present invention relates to a content display device and a computer program.

  A conventional search function provided in a search service for web content or the like finds desired content by a user specifying a search keyword. However, when searching for video content such as broadcast programs, entering characters is a cumbersome operation, and while entering characters, it may be necessary to keep an eye on the video content being viewed. It is desirable to be able to perform a search operation without performing the above. For this purpose, it is effective to use a search key as an arbitrary program, such as a program being viewed, to search a program having a high content relation or a part of scenes.

  In digital broadcasting, text describing the outline of program content (hereinafter referred to as “program summary text”) is distributed as an electronic program guide (EPG). Therefore, the degree of relevance between the program outline text of the program serving as the search key and the program outline text of the program to be searched stored in the database is quantified and evaluated, so that the degree of relevance to the program serving as the search key is high. You can search for programs. With this function, users can select featured programs and educational programs that explain the topics covered in more detail by, for example, specifying the news program in use as a search key without explicitly entering a search keyword. You can search.

  On the other hand, Patent Literature 1 discloses a technique for searching program information using a keyword selected by a user from keywords related to a received broadcast program and a related keyword selected by the user from related keywords of the keyword. Is described.

JP 2005-348071 A

  As described above, in order to search for other content related to content such as a broadcast program without using a keyword, a vector space targeted for text describing the outline and details of the content, such as program summary text There is a related content search function by law. Then, a program having a high degree of association is selected as a related program. In this vector space method, the relevance is calculated based on the co-occurrence of a large number of words between the texts accompanying the content. Therefore, the search results are compared with the keyword search that searches for the occurrence of several keywords. A variety of contents are included. For this reason, there is a drawback in that the grasp of the search result by the user is hindered.

  Therefore, in order to improve the quality of the user interface when providing related content searches, it is necessary to display in an easy-to-understand manner the related content detected as a search result is determined to be related. There is. However, there is no effective technique for displaying from what viewpoint the detected related content is related other than displaying the keyword string. Patent Document 1 displays and selects keywords for searching for program information related to a user and related keywords, and does not solve such a problem.

  The present invention has been made in consideration of such circumstances, and its purpose is to present in an easy-to-understand manner information on what the related content is related from when the related content is presented. An object of the present invention is to provide a related content display device and a computer program.

[1] According to one aspect of the present invention, a text information storage unit that stores content identification information and text information that represents the content in association with each other, and an inter-vector relevance analysis unit that calculates the relevance between vectors A first content vector generation unit that generates a key content vector representing a linguistic feature of the text information associated with the specified content identification information, and the text information associated with other content identification information A second content vector generation unit that generates a candidate content vector representing the linguistic features of the key content vector based on the relevance between the key content vector calculated by the inter-vector relevance analysis unit and each of the candidate content vectors The candidate content vector highly related to the key content vector Co-occurs with the related content selection unit selected as the content vector, the text information corresponding to the key content vector, and the text information corresponding to the related content vector based on the key content vector and the related content vector A co-occurrence vector generation unit that generates a word co-occurrence vector that represents the characteristics of the word, and a related content sentence that represents the linguistic characteristics of the sentence for each sentence included in the text information corresponding to the related content vector A first sentence vector generation unit for generating a vector, and the word co-occurrence vector based on the degree of association between the word co-occurrence vector calculated by the inter-vector relevance analysis unit and each of the related content sentence vectors; Select the related content sentence vector that is highly relevant. A related display sentence selection unit that specifies the sentence corresponding to the selected related content sentence vector as a related display sentence of the related content, and a relation that outputs a related display sentence of the related content specified by the related display sentence selection unit A related content display device comprising: a content output unit.
According to the present invention, the related content is selected based on the linguistic similarity between the text indicating the content of the designated content and the text indicating the content of the content that is a candidate for the related content, and the content of the related content is selected. The sentence that best represents the linguistic features common to the sentence indicating the content of the designated content and the sentence selected as the related content is selected and output from the sentences constituting the sentence indicating the content.
As a result, it is possible to present information about what kind of viewpoint the content related to the content viewed by the user using a content display device such as a television or a personal computer is related, and the user is interested. Easier to find related content.

[2] One aspect of the present invention is the related content display device described above, wherein for each sentence included in the text information corresponding to the key content vector, a key content sentence representing a linguistic feature of the sentence. A second sentence vector generation unit that generates a vector, and the related display sentence selection unit includes a degree of association between the word co-occurrence vector calculated by the inter-vector relation analysis unit and each of the key content sentence vectors. The key content sentence vector highly relevant to the word co-occurrence vector is selected, the sentence corresponding to the selected key content sentence vector is specified as a related display sentence of the key content, and the related content The output unit outputs a related display sentence of the key content specified by the related display sentence selection unit ; It is characterized by.
According to the present invention, the linguistic feature common to the text indicating the content of the specified content and the text selected as the related content is the most common among the sentences constituting the text indicating the content of the specified content. Select a sentence that is well represented and output it.
As a result, it is possible to present information on what kind of viewpoint the content that the user is currently viewing is related to the related content obtained as a search result. It becomes easy to find.

[3] One aspect of the present invention is the related content display device described above, further including a content storage unit that stores content identification information and related information of the content in association with each other, and the related content output unit includes The related information associated with the same content identification information as the text information corresponding to the related content vector is read from the content storage unit, and the read related information or information generated based on the related information is read from the related display text. It is output together with.
According to the present invention, information related to related content obtained as a result of search and information obtained by processing the information are added to the related display text and output.
As a result, the title, thumbnail, etc. of the related content obtained as a search result can be presented on the content display device, and it becomes easy to find the related content that the user is interested in.

[4] According to one aspect of the present invention, a computer used as a related content display device stores a text information storage unit that stores content identification information and text information that represents the content in association with each other, and a degree of association between vectors An inter-vector relevance analysis unit that calculates a first content vector generation unit that generates a key content vector representing a linguistic feature of the text information associated with the specified content identification information, and other content identification information A second content vector generation unit that generates a candidate content vector representing a linguistic feature of the associated text information, the key content vector calculated by the inter-vector relevance analysis unit, and each of the candidate content vectors Based on the relevance of the key content vector and Corresponding to the text information and the related content vector corresponding to the key content vector based on the related content selection unit that selects the candidate content vector having high relevance as the related content vector, the key content vector, and the related content vector A co-occurrence vector generation unit that generates a word co-occurrence vector representing the characteristics of the words that co-occur with the text information, and for each sentence included in the text information corresponding to the related content vector, the language of the sentence A first sentence vector generation unit that generates a related content sentence vector representing a characteristic, and a degree of association between the word co-occurrence vector calculated by the inter-vector relevance analysis part and each of the related content sentence vectors The relevance of the word co-occurrence vector A related display sentence selection unit that selects the related content sentence vector selected and specifies the sentence corresponding to the selected related content sentence vector as a related display sentence of the related content; A computer program that functions as a related content output unit that outputs a related display sentence of content.
According to the present invention, the related content is selected based on the linguistic similarity between the text indicating the content of the designated content and the text indicating the content of the content that is a candidate for the related content, and the content of the related content is selected. The sentence that best represents the linguistic features common to the sentence indicating the content of the designated content and the sentence selected as the related content is selected and output from the sentences constituting the sentence indicating the content.
As a result, it is possible to present information about what kind of viewpoint the content related to the content viewed by the user using a content display device such as a television or a personal computer is related, and the user is interested. Easier to find related content.

  According to the present invention, the related content of the key content is searched based on the degree of relevance between the texts accompanying the content and describing the content of the content, and the related content obtained as a result of the search is searched. A sentence indicating information as to whether it is determined to be related from such a viewpoint can be extracted from the text accompanying the related content and presented to the user together with the related content of the search result. Therefore, the user can obtain information on from which point of view the related content of the search result is related to the key content, and it is easy to find related content that attracts interest.

It is a functional block diagram of the related content display apparatus by one Embodiment of this invention. It is a figure which shows the example of the data accumulate | stored in the text information storage part by the embodiment. It is a figure which shows the example of the data accumulate | stored in the content storage part by the embodiment. It is a figure which shows the processing flow of the related content display apparatus by the embodiment. It is a figure which shows the example of the search result display screen by the embodiment.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 1 is a functional block diagram showing a configuration of a related content display device 1 according to an embodiment of the present invention, and only functional blocks related to the invention are extracted and shown. In the figure, a related content display device 1 includes a text information storage unit 2, a content storage unit 3, a related content search unit 4, an intervector relevance analysis unit 5, a search result analysis unit 6, and a related content output unit 7. Consists of including.

  The text information storage unit 2 and the content storage unit 3 are realized by a hard disk device, a semiconductor memory, or the like. The text information storage unit 2 stores a content ID, which is identification information for specifying the content, and text information data indicating text describing information related to the contents of the content in association with each other. The content storage unit 3 stores a content ID and content data that is digital data of the content body in association with each other.

  The inter-vector relevance analysis unit 5 receives a set of vectors and calculates the relevance between these vectors.

The related content search unit 4 includes a first content vector generation unit 41, a second content vector generation unit 42, and a related content selection unit 43.
When the first content vector generation unit 41 acquires the text information data specified by the content ID serving as a search key from the text information storage unit 2, the first content vector generation unit 41 applies to each word appearing in the sentence indicated by the acquired text information data. A weight corresponding to the importance is assigned, and a content vector expressing the linguistic feature of the sentence as a vector is generated based on the numerical value of the weight. Hereinafter, the content specified by the content ID serving as a search key is generated from “key content”, the text information data specified by the content ID is generated from “key content text information data”, and the text indicated by the key content text information data. The content vector is referred to as “key content vector”.

  When the second content vector generation unit 42 acquires the text information data of the content to be searched as a candidate for the related content of the key content from the text information storage unit 2, the second content vector generation unit 42 acquires the acquired text by the same method as the generation of the key content vector. A content vector is generated from each sentence indicated by the information data. Hereinafter, the content that is a candidate for the related content is “candidate content”, the text information data of the candidate content is “candidate content text information data”, and the content vector generated from the sentence indicated by the candidate content text information data is “candidate content vector” ".

  The related content selection unit 43 acquires the relevance between the key content vector generated by the first content vector generation unit 41 and each candidate content vector generated by the second content vector generation unit 42 from the inter-vector relevance analysis unit 5. When the related content is selected from the candidate content based on the acquired relevance level, the search result is obtained by using the candidate content text information data and the candidate content vector of the selected candidate content as the related content text information data and the related content vector, respectively. Output to the analysis unit 6.

The search result analysis unit 6 includes a co-occurrence vector generation unit 61, a first sentence vector generation unit 62, a second sentence vector generation unit 63, and a related display sentence selection unit 64.
The co-occurrence vector generation unit 61 uses the key content vector and the candidate content vector to generate word co-occurrence representing the characteristics of the words included in common in the text indicated by the key content text information data and the text indicated by the related content text information data. Generate a vector.

  The first sentence vector generation unit 62 represents the linguistic features of each sentence as a vector for each sentence in the sentence indicated by the related content text information data by the same method as the generation of the key content vector and the candidate content vector. Generated sentence vectors. Hereinafter, a sentence vector generated from a sentence in a sentence indicated by related content text information data is referred to as a “related content sentence vector”.

  The second sentence vector generation unit 63 expresses the linguistic features of each sentence as a vector for each sentence in the sentence indicated by the key content text information data by the same method as the generation of the key content vector and the candidate content vector. Generated sentence vectors. Hereinafter, the sentence vector generated from the sentence in the sentence indicated by the key content text information data is referred to as “key content sentence vector”.

  The related display sentence selection unit 64 acquires the degree of association between the word co-occurrence vector, each related content sentence vector, and each key content sentence vector from the inter-vector relevance analysis unit 5, and based on the obtained degree of relation Then, a related display sentence that is the sentence that best indicates the relevance between the key content and the related content is selected from the sentences indicated by the related content text information data and the key content text information data.

  When the related content output unit 7 acquires the title and content data of the related content as the related information of the content from the content storage unit 3, image data such as the title, a thumbnail generated based on the content data, and the search result The related display text selected by the analysis unit 6 is output and displayed on the display of a content display device such as a television receiver or a personal computer where the user is viewing the content.

  In the following description, the text information storage unit 2, the content storage unit 3, the related content search unit 4, the inter-vector relevance analysis unit 5, the search are connected to one or a plurality of servers connected to the content display device via the network. The case where the result analysis part 6 and the related content output part 7 are provided is demonstrated. However, the text display storage unit 2, the content storage unit 3, the related content search unit 4, the inter-vector relevance analysis unit 5, the search result analysis unit 6, and the related content output unit 7 may all be provided in the content display device. In addition, any part of the functional units may be provided in the content display device. For example, the text information storage unit 2 and the content storage unit 3 may be provided in the server, other functional units may be provided in the content display device, only the content storage unit 3 may be provided in the server, and other functional units may be provided in the content display device. Also good.

  FIG. 2 is a diagram illustrating an example of data stored in the text information storage unit 2. As shown in the figure, the text information storage unit 2 stores a content ID and text information data L in association with each other. Here, as text information data L, program summary text explaining the program content of a broadcast program such as an electronic program guide (EPG) is stored in a file format. The program summary text consists of a plurality of sentences.

  FIG. 3 is a diagram illustrating an example of data stored in the content storage unit 3. As shown in the figure, the content storage unit 3 stores a content ID, content data, and data indicating the title of the content in association with each other. The content data is, for example, a moving image data or audio data file that constitutes a program video of a broadcast program.

FIG. 4 is a diagram illustrating a processing flow of the related content display device 1.
In the figure, the related content search unit 4 of the related content display device 1 receives the search key in which the content ID of the key content is set from the content display device (step S105). For example, the content display device displays the content ID of a broadcast program selected for viewing by the user, the content ID input by the user, or the content ID of the broadcast program displayed when a search instruction is input by the user. Send. Here, it is assumed that the content ID of the key content k is set in the search key.

The first content vector generation unit 41 of the related content search unit 4 acquires the content ID from the search key received as the search request in step S105, and the key content text information data that is the text information data L specified by the content ID. L _k is read from the text information storage unit 2. Subsequently, the second content vector generation unit 42 is candidate content text information data L _j (1 ≦ j ≦ m), which is content text information data L other than the key content text information data L _k read in step S110, and j ≠ k) is read from the text information storage unit 2 (step S110). Here, m is the number of text information data L stored in the text information storage unit 2.

First content vector generation unit 41 generates a content vector from the program summary text indicated by the key content text information data L _k (step S115). Further, the second content vector generation unit 42 generates a content vector from each program summary text indicated by each candidate content text information data L _j (step S120).

A content vector w _p of a sentence indicated by the text information data L _p of a certain content p is expressed as (Equation 1) below.

Content vector w _p = (w _p1 ,..., W _pi ,..., W _pN ) (Expression 1)

The element w _pi is the importance of the word i (1 ≦ i ≦ N) in the sentence indicated by the text information data L _p of the content p. When the word i does not appear in the sentence indicated by the text information data L _p, w _pi = 0, and when it appears, w _pi > 0. N is the number of different words that appear in the sentence indicated by the entire text information data L. The content vector can be a vector expression using TF / IDF (word importance evaluation method), for example. For this vector expression, see, for example, Takenobu Tokunaga, “Information Retrieval and Language Processing”, University of Tokyo Press, Chapter 2, p. 32-33, ISBN: 4130654055.

In TF / IDF, when there is a document group composed of a plurality (DN) of documents, a TF (i) value indicating the number of occurrences of a word i (keyword) in a document and the word in the document group. Using the DF (i) value indicating the number of documents including i, the TF-IDF value of the word i in the document is calculated by TF (i) × log (DN / DF (i)). The vector expression representing the characteristics of each document is expressed by using the TF-IDF value of each word as an element. That is, the element w _pi of the content vector w _p is the TF-IDF value of the word i.

Specifically, the first content vector generation unit 41 performs morphological analysis on the key content text information data L _k read in step S110, and the second content vector generation unit 42 reads the candidate content text information data read in step S110. Each program summary text indicated by L _j is morphologically analyzed. The first content vector generation unit 41 determines a word as a keyword by extracting a specific part of speech such as a noun from the results of these morphological analyses, and determines a correspondence between each word and a vector element. Subsequently, the TF (i) value of each word i (i = 1 to N) as a keyword is acquired from the program summary text indicated by the key content text information data L _k and each candidate content text information data L _j , The DF (i) value for the word i is acquired. Then, reading the number of content text information data L stored in the text information storage section 2 as a DN, the first content vector generator 41, the key content text information about the data L _k TF-IDF value of each word i And the second content vector generation unit 42 calculates the TF-IDF value of each word i for each program summary text indicated by each candidate content text information data L _j .

Hereinafter, a content vector generated based on the key content text information data L _k is referred to as a key content vector w _k , and a content vector generated from the candidate content text information data L _{j is referred} to as a candidate content vector w _j .

Subsequently, when the related content selection unit 43 selects one of the candidate content vectors w _j generated in step S120, the related content analysis is performed on the key content vector w _k and the selected candidate content vector w _j. Output to unit 5. The inter-vector relevance analysis unit 5 substitutes the received key content vector w _{k and} candidate content vector w _j for w _a and w _b in (Equation 2) below, and the relevance R (w _k), which is a cosine measure. , W _j ). The more similar the two vectors, the narrower the angle formed between these vectors, and the greater the value of relevance.

Relevance R (w _a , w _b ) = (w _a · w _b ) / (| w _a || w _b |) (Expression 2)
However, w _a · w _b is the inner product of the vector w _a and the vector w _b .

The inter-vector relevance analysis unit 5 outputs the calculated relevance R (w _k , w _j ) to the related content selection unit 43.
If there is a candidate content vector for which the degree of association with the key content vector w _k has not yet been calculated, the related content selection unit 43 selects one of the candidate content vectors and repeats the above processing, and for each of all candidate content vectors w _j The degree of association R (w _k , w _j ) between the key content vector w _k and the candidate content vector w _j is acquired (step S125).

The related content selection unit 43 selects the highest relevance level or a predetermined number of relevance levels from the relevance level R (w _k , w _j acquired in step S125 (step S130). Here, it is assumed that one relevance level R (w _k , w _j ) has been selected, and the related content selection unit 43 calculates the selected relevance level R (w _k , w _j ). When the candidate content vector w _j is specified, the candidate content text information data L _j that is the generation source of the candidate content vector w _j and the content ID associated with the candidate content text information data L _j are specified. . Related content selection unit 43, the identified content ID, a particular candidate content text information data L _j, identified candidate con Ntsubekutoru w _j, respectively, the content ID of the related content r, related content text information data L _r, and outputs the search result analysis unit 6 as the related content a vector w _r, the content ID of the key content k, key content vector w _k and and it outputs the key content text information data _{L k} to the search result analysis unit 6 (step S135).

The co-occurrence vector generation unit 61 of the search result analysis unit 6 receives the received key content vector w _k = (w _k1 ,..., W _ri ,..., W _kN ) and the related content vector w _r = (w _r1,. The word co-occurrence vector w _kr is generated from the following (formula 3) from w _{ri ,.}

Word co-occurrence vector w _kr = (w _k1 · w _r1 ,..., W _ki · w _ri ,..., W _kN · w _rN ) (Equation 3)

The word co-occurrence vector w _kr represents the characteristics of a word included in common in the program summary text indicated by the key content text information data L _k and the program summary text indicated by the related content text information data L _r . In the word co-occurrence vector w _kr , the element of the word that does not co-occur is 0, and the element of the word that co-occurs has a positive value. The larger the value of the co-occurring word element, the higher the importance of both program summary texts indicated by the key content text information data L _k and the related content text information data L _r .

Next, the first sentence vector generation unit 62 determines the importance of each word appearing in each sentence from each sentence s constituting the program summary text indicated by the related content text information data L _r by the same method as the content vector. The following related content sentence vector w _rs based on the degree is generated (step S145). The related content sentence vector w _rs is expressed as the following (Formula 4).

Related content sentence vector w _rs = (w _rs1 ,..., W _rsi ,..., W _rsN ) (Expression 4)

The element w _rsi is the importance of the word i (1 ≦ i ≦ N) in the sentence s in the program summary text indicated by the related content text information data L _r , and when the word i does not appear in the sentence s, w _rsi = 0, and when it appears, w _rsi > 0.
When TF / IDF is used for the related content sentence vector, the first sentence vector generation unit 62 uses the word i for each sentence s based on the result of morphological analysis of the program summary text indicated by the related content text information data L _r. TF (i) value is obtained, and a related content sentence vector is generated. The DF (i) value and the document number DN of each word i may be received from the related content search unit 4 or may be obtained by analyzing the text information data L stored in the text information storage unit 2. .

Subsequently, when the related display sentence selection unit 64 selects one of the related content sentence vectors w _rs generated in step S145, the related co-occurrence vector w _kr and the selected related content sentence vector w _rs are set between vectors. Output to the relevance analysis unit 5. The inter-vector relevance analysis unit 5 substitutes the received word co-occurrence vector w _{kr and} related content sentence vector w _rs for w _a and w _b in (Equation 2) described above to obtain the relevance R (w _kr , w _{When rs} ) is calculated, it is output to the related display sentence selection unit 64.

If there is a related content sentence vector w _rs for which the degree of relevance with the word co-occurrence vector w _kr has not yet been calculated, the related display sentence selection unit 64 selects one of them and repeats the above processing, and all the related content for each sentence vector _{w rs,} word cooccurrence vector _{w kr} and related content sentence vector _{w rs} and of relevance _{R (w kr,} w _rs) acquires (step S150).

The related display text selection unit 64 selects the highest relevance degree from the relevance degrees R (w _kr , w _rs ) acquired in step S150. The related display sentence selection unit 64 selects the sentence of the program summary text that is the generation source of the related content sentence vector w _rs for which the selected degree of relevance has been calculated as the related display sentence of the related content r (step S155). . As a result, the sentence that best represents the characteristic of the word that is commonly included in the program summary text indicated by the key content text information data L _k and the program summary text indicated by the related content text information data L _r is the related content. It is selected from the program summary text indicated by the text information data L _r.

Subsequently, the second sentence vector generation unit 63 appears in each sentence from the sentences s constituting the program summary text indicated by the key content text information data L _k by the same processing as the related content sentence vector w _rs. A key content sentence vector w _ks = (w _ks1 ,..., W _ksi ,..., W _ksN ) based on the importance of each word is generated (step S160). The element w _ksi is the importance of the word i (1 ≦ i ≦ N) in the sentence s in the program summary text indicated by the key content text information data L _k . When the word i does not appear in the sentence s, w _ksi = 0, w _ksi > 0 when it appears.

Subsequently, when the related display sentence selection unit 64 selects one of the key content sentence vectors w _ks generated in step S160, the word co-occurrence vector w _kr and the selected key content sentence vector w _ks are set between vectors. Output to the relevance analysis unit 5. The inter-vector relevance analysis unit 5 substitutes the received word co-occurrence vector w _kr and key content sentence vector w _ks for w _a and w _b in (Equation 2) described above to obtain the relevance R (w _kr , w _{When ks} ) is calculated, the calculation result is output to the related display text selection unit 64.

If there is a key content sentence vector w _ks for which the degree of association with the word co-occurrence vector w _kr has not yet been calculated, the related display sentence selection unit 64 selects one of them and repeats the above process, and all the key contents for each sentence vector _{w ks,} word cooccurrence vector _{w kr} and relevance _{R (w kr,} w _ks) of the key content sentence vector _{w ks} acquires (step S165)

The related display text selection unit 64 selects the highest relevance degree from the relevance degrees R (w _kr , w _ks ) acquired in step S165. The related display text selection unit 64 selects, as the related display text of the key content k, the text of the program summary text that is the generation source of the key content text vector w _ks for which the selected degree of relevance has been calculated (step S170). . As a result, the sentence that best represents the feature of the word that is commonly included in the program summary text indicated by the key content text information data L _k and the program summary text indicated by the related content text information data L _r is the key content. It is selected from the program summary text indicated by the text information data L _k.

  The related display text selection unit 64 relates the content ID of the key content k, the content ID of the related content r, the related display text of the related content r selected in step S160, and the key content k selected in step S170. The display sentence is output to the related content output unit 7 (step S175).

  The related content output unit 7 reads the title corresponding to the content ID of the key content k from the content storage unit 3, reads the content data and the title corresponding to the content ID of the related content r, and reads the content data of the related content r that has been read out A still image such as a thumbnail is generated based on (Step S180). The related content output unit 7 generates screen data of a search result display screen that displays the title of the key content k and the related display text, the title of the related content r, the related display text, and the thumbnail, and outputs the screen data to the content display device (step). S185). The content display device displays the received screen data on the display. As a result, the user grasps information regarding the related content r of the key content k currently being viewed and the viewpoint from which the related content r is determined to be similar.

  In step S130, when the related content selection unit 43 selects a plurality of related contents r, the processes of steps S140 to S180 are performed for each related content r. In step S185, screen data of a search result display screen that displays related content r in descending order of relevance is generated.

  FIG. 5 is a diagram illustrating an example of a search result display screen displayed on the content display device. In the figure, the title g1 of the key content k and the related display sentences g2 and g3 of the key content k are displayed on the search result display screen. The related display sentences g2 and g3 of the key content k are displayed in the order corresponding to the related content r having a high degree of relevance. In the figure, the related display sentence g2 of the key content having the highest degree of relevance is highlighted, the title g4 of the related content r having the highest degree of relevance, the related display sentence g5 of the related content r, and the related content. A thumbnail g6 of r is displayed.

  When the user clicks the related display sentence g3 of the key content with a mouse or the like, a display instruction for the related content having the second highest degree of relevance is transmitted from the content display device to the related content output unit 7. In this case, the related content output unit 7 highlights the related display sentence g3 of the key content k, and displays the title g1 of the key content k, the related display sentences g2 and g3 of the key content k, and the second most related content. The screen data of the search result display screen that displays the title g4 of r, the related display text g5 of the related content r, and the thumbnail g6 of the related content r are generated and returned to the content display device.

  Instead of acquiring a search key from the user's content display device, the content ID currently received from the external content distribution device that distributes the content to the user's content display device or the content ID received from the content display device May be received as a search key. In this case, the related content output unit 7 may output the search result display screen to the content distribution device, and transmit the search result display screen from the content distribution device to the content display device. The content distribution device may include a part or all of the functional units of the related content display device 1.

  In the above description, a broadcast program has been described as an example of content. However, it may be a moving image distributed over the Internet or the like, or may be a still image, text, audio data, or a combination thereof.

In the above, when generating a content vector, a word as a keyword is extracted from the result of morphological analysis. However, word information as a keyword and information on vector elements corresponding to each word i are previously stored as text information. It may be stored in the storage unit 2.
In addition, the content information generated based on the program summary text indicated by the text information data and the content ID of the text information data are stored in advance in the text information storage unit 2 and received in step S105. The content vector stored in association with the content ID may be read as the key content vector w _k , and the content vector stored in association with the other content ID may be read out as the candidate content vector w _j .
Similarly, a text vector generated based on each sentence of the program summary text indicated by the text information data and the content ID of the text information data are stored in the text information storage unit 2 in association with each other. The sentence vector stored in association with the content ID of the content r is read out as the related content sentence vector w _rs , and the sentence vector stored in association with the content ID of the key content _k is read out as the key content sentence vector w _ks. May be.

  In the above, TF / IDF is used to generate the content vector, but any method for generating a vector expression indicating a linguistic feature based on the importance of each word appearing in the sentence is used. Can do. For example, an arbitrary sentence group such as a newspaper article is analyzed in advance to determine each word to be a keyword and its weight, and when the word appears in text information data, the weight corresponding to the word is set as a content vector or content. It may be used as an element of a sentence vector.

  According to the present embodiment, other content that is associated with the content and is related based on the degree of relevance between the texts in which the content of the content is described is searched, and how is the other content obtained as a result of the search? It is possible to extract a sentence indicating information that is determined to be related from various viewpoints from the text accompanying the other content, and present it to the user together with the content as a result of the search. Therefore, the user can grasp from what point of view the content of the search result is related.

  Conventionally, when searching for related content, only keywords such as “President XX” and “Prime Minister” are presented to the user. On the other hand, in this embodiment, “President XX prepared for a meeting with the Prime Minister △”, “President XX did not meet with the Prime Minister △,” and so on. , A sentence including many characteristic keywords included in both the key content and the search result content is presented together. In particular, a negative expression such as “not done” is often not searchable by displaying only keywords, but in this embodiment, by indicating the relationship between contents by sentences, the user can simply Information that cannot be obtained by presenting keyword strings can be obtained, and it becomes easier to find interesting content.

  The related content display device 1 described above has a computer system therein. The operation processes of the related content search unit 4, the inter-vector relevance analysis unit 5, the search result analysis unit 6, and the related content output unit 7 of the related content display device 1 are recorded in a computer-readable form in the form of a program. The above-described processing is performed when the computer system reads and executes the program stored in the medium. The computer system here includes a CPU, various memories, an OS, and hardware such as peripheral devices.

Further, the “computer system” includes a homepage providing environment (or display environment) if a WWW system is used.
The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Furthermore, the “computer-readable recording medium” dynamically holds a program for a short time like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, a volatile memory in a computer system serving as a server or a client in that case, and a program that holds a program for a certain period of time are also included. The program may be a program for realizing a part of the functions described above, and may be a program capable of realizing the functions described above in combination with a program already recorded in a computer system.

DESCRIPTION OF SYMBOLS 1 ... Related content display apparatus 2 ... Text information storage part 3 ... Content storage part 4 ... Related content search part 41 ... 1st content vector generation part 42 ... 2nd content vector generation part 43 ... Related content selection part 5 ... Inter-vector relation Degree analysis unit 6 ... Search result analysis unit 61 ... Co-occurrence vector generation unit 62 ... First sentence vector generation unit 63 ... Second sentence vector generation unit 64 ... Related display sentence selection unit 7 ... Related content output unit

Claims (4)

A text information storage unit that stores content identification information and text information that represents the content in association with each other;
An inter-vector relevance analysis unit for calculating the relevance between vectors;
A first content vector generation unit that generates a key content vector representing the linguistic characteristics of the text information associated with the specified content identification information;
A second content vector generation unit that generates a candidate content vector representing the linguistic features of the text information associated with other content identification information;
The candidate content vector having high relevance to the key content vector is selected as a related content vector based on the relevance between the key content vector calculated by the inter-vector relevance analysis unit and each of the candidate content vectors. A related content selection section;
Based on the key content vector and the related content vector, a word co-occurrence vector representing the characteristics of the words that co-occur in the text information corresponding to the key content vector and the text information corresponding to the related content vector A co-occurrence vector generation unit to generate;
A first sentence vector generation unit that generates a related content sentence vector representing a linguistic feature of the sentence for each sentence included in the text information corresponding to the related content vector;
Based on the relevance between the word co-occurrence vector calculated by the inter-vector relevance analysis unit and each of the related content sentence vectors, the related content sentence vector highly relevant to the word co-occurrence vector is selected. A related display sentence selection unit that identifies the sentence corresponding to the selected related content sentence vector as a related display sentence of the related content;
A related content output unit that outputs a related display sentence of the related content specified by the related display sentence selection unit;
A related content display device comprising: For each sentence included in the text information corresponding to the key content vector, further comprising a second sentence vector generation unit that generates a key content sentence vector representing a linguistic feature of the sentence,
The related display sentence selection unit is highly related to the word co-occurrence vector based on the degree of association between the word co-occurrence vector calculated by the inter-vector relevance analysis unit and each of the key content sentence vectors. Selecting the key content sentence vector, specifying the sentence corresponding to the selected key content sentence vector as a related display sentence of the key content;
The related content output unit outputs a related display text of the key content specified by the related display text selection unit ;
The related content display device according to claim 1. A content storage unit that stores the content identification information and the related information of the content in association with each other;
The related content output unit reads the related information associated with the same content identification information as the text information corresponding to the related content vector from the content storage unit, and generates based on the read related information or the related information Outputting information together with the related display sentence;
The related content display device according to claim 1 or 2, wherein A computer used as a related content display device,
A text information storage unit that stores content identification information and text information that represents the content in association with each other;
A relevance analysis unit for calculating relevance between vectors,
A first content vector generation unit that generates a key content vector representing a linguistic feature of the text information associated with the specified content identification information;
A second content vector generation unit that generates a candidate content vector representing a linguistic feature of the text information associated with other content identification information;
The candidate content vector having high relevance to the key content vector is selected as a related content vector based on the relevance between the key content vector calculated by the inter-vector relevance analysis unit and each of the candidate content vectors. Related content selection section,
Based on the key content vector and the related content vector, a word co-occurrence vector representing the characteristics of the words that co-occur in the text information corresponding to the key content vector and the text information corresponding to the related content vector A co-occurrence vector generation unit to generate,
A first sentence vector generation unit that generates a related content sentence vector representing a linguistic feature of the sentence for each sentence included in the text information corresponding to the related content vector;
Based on the relevance between the word co-occurrence vector calculated by the inter-vector relevance analysis unit and each of the related content sentence vectors, the related content sentence vector highly relevant to the word co-occurrence vector is selected. A related display sentence selection unit that identifies the sentence corresponding to the selected related content sentence vector as a related display sentence of the related content;
A related content output unit that outputs a related display sentence of the related content specified by the related display sentence selection unit;
A computer program that functions as a computer program.

Images