JP6137960B2 - Content search apparatus, method, and program - Google Patents

Description

  The present invention relates to an apparatus, a method, and a program for searching for recommended content.

  In recent years, enormous contents of various types such as documents, sounds, images, and videos can be browsed on the Internet. In order to efficiently search for these, not only a technique for searching for contents at high speed using keywords or the like, but also a more advanced contents search technique that reflects user preferences is required.

  As an attempt to expand information retrieval using personal preferences, a technology for retrieving information from the perspective of others using social media has also been proposed. For example, in a shopping site on the Internet, a product recommendation method using reviews and purchase histories of other users is used. According to this recommendation method, a function called “collaborative filtering” such as “the user who bought this product also bought such a product” is used to search for, recommend, etc. the product implicitly liked by the user. It can be performed.

  Furthermore, a technique for automatically extracting contents viewed (interested) by a friend on the SNS using a connection between people on an SNS (Social Networking Service) has been proposed (for example, patents). Reference 1). In addition, a technique has been proposed in which content with a high degree of attention is extracted using comment information recorded by an unspecified number of people and the content is automatically recorded (see, for example, Patent Document 2).

JP 2012-256252 A JP 2012-049995 A

  By the way, it is considered that the content includes many items created from various viewpoints with respect to a certain theme. For example, in the case of “programs” such as TV broadcasts, there are some programs that were created from the perspective of “demerits (eg: economic downturn)” against the “consumption tax increase”. Some were created from the perspective of its “merits (eg, enhancement of welfare)”. Therefore, the user often desires to browse information based on these various viewpoints in a balanced manner.

However, when searching for these various programs, if the user's preference information “prefers economic information” is used, many consumption related programs are searched and recommended for consumption tax. It was.
In addition, even with the above-described technology, it is possible to contact content that is not trapped by individual preference by using the action history of others, but it is difficult to obtain information in a balanced manner from various viewpoints.

  An object of this invention is to provide the content search apparatus, method, and program which can extract the content containing the concept which the user has not touched until now.

  A content search apparatus according to the present invention includes a word network storage unit that stores word relationship data describing a relationship between a plurality of words, a content storage unit that stores content information, and a word that extracts a word included in the content information An information acquisition unit, a word map generation unit that generates a word map with the word relation data added to the words extracted by the word information acquisition unit for each content, and a word map for the content selected by the user A word map comparison unit that preferentially extracts content with a small number of common parts of the word map from the other content compared to the word map for the other content.

According to this configuration, the content search device uses a word map generated from information on content viewed by the user and information on other content using word relationship data describing a relationship between words prepared in advance. The generated word maps are compared, and contents with few common parts are preferentially extracted. Therefore, the content including the concept that the user has not touched until now is recommended while being related to the viewed content.
As a result, the user can expand the range of knowledge and thought by comprehensively recommending a content group including many viewpoints while being related to the theme of the viewed content.

  The word map comparison unit includes at least a part of important words determined to be relatively high among words included in the word map for the content selected by the user, and the word maps Content with a long distance may be preferentially extracted.

  According to this configuration, the content search apparatus is highly relevant because important words are common, and can recommend content with different viewpoints preferentially because the distance between word maps is long. As a result, the content search device can recommend appropriate content to the user.

  The word map comparison unit may weight the distance between the word maps based on the type of relation link in the word relation data.

  According to this configuration, the content search device weights the distance between the word maps based on the type of relational link, so it is possible to more accurately determine the conceptual difference between the content and appropriately recommend the content to the user. it can.

  The word map generation unit may generate the word map by excluding words that are less than a predetermined importance from the words extracted by the word information acquisition unit.

  According to this configuration, the content search apparatus excludes words that do not satisfy the predetermined importance when generating the word map. Therefore, the content search apparatus can generate a word map that more accurately represents the content content by eliminating words that cause noise that is not closely related to the content content.

  The importance may be an index based on TF-IDF for the word extracted by the word information acquisition unit.

  According to this configuration, the content search apparatus can easily calculate the importance of words with high accuracy using an index based on TF-IDF.

  The importance may be an index based on the number of related links in the word relationship data for the word extracted by the word information acquisition unit.

  According to this configuration, the content search apparatus can easily calculate the importance of a word with high accuracy using an index based on the number of related links.

  The content search device according to the present invention may include a history storage unit that stores a content selection history for each user, and the word map comparison unit may preferentially extract content that is not included in the selection history.

  According to this configuration, the content search apparatus preferentially recommends content that is not included in the selection history of the user viewing the content. Priority can be given to content including

  The content search device according to the present invention includes a selection map storage unit that stores, in association with the user, a word that matches a word map for the content selected by the user in the word relation data, and the word map comparison unit May preferentially extract content corresponding to a word map having few common parts with the word group stored in the selection map storage unit.

  According to this configuration, the content search apparatus extracts a word map with few common parts with a word group related to content that the user has viewed and recommends the content. Therefore, the content search apparatus can preferentially recommend content that includes words that the user has not touched on the word network, that is, has a different concept.

  The selection map storage unit stores words that match a word map for the content selected by the user together with the number of selections, and the word map comparison unit stores content corresponding to a word map including words with a low number of selections. You may extract preferentially.

  According to this configuration, the content search apparatus stores the words related to the viewed content together with the number of selections, and selects the word map so as to equalize the number of selections. Therefore, it is possible to efficiently select contents of various concepts.

  The content search apparatus according to the present invention may include a comparison information generation unit that generates comparison information between the word relation data and the word data for each user in the selection map storage unit.

  According to this configuration, the content search apparatus can generate and output comparison information between the word network and the word group or the word map obtained from the content viewed by the user. Therefore, the content search device can make the user easily grasp the areas that the user frequently views, the areas where the viewing frequency is low, and the like.

  The content information may be a user's posted text for the content.

  According to this configuration, since the content search apparatus can adopt the user's posted text as the content information, the amount of information increases and the thought of the user is reflected in the posted text. Is recommended.

  A content search method according to the present invention includes a word network including word relationship data describing a relationship between a plurality of words, and a word information acquisition step in which a computer capable of referring to the content information extracts words included in the content information; For each content, a word map generation step for generating a word map in which the word relation data is added to the word extracted in the word information acquisition step; and a word map for the content selected by the user, And a word map comparison step that preferentially extracts contents with a small number of common parts of the word map from the other contents.

  A content search program according to the present invention includes a word network including word relationship data describing a relationship between a plurality of words, and a word information acquisition step of extracting a word included in the content information to a computer capable of referring to the content information. For each content, a word map generation step for generating a word map in which the word relation data is added to the word extracted in the word information acquisition step; and a word map for the content selected by the user, And a word map comparison step for preferentially extracting contents with a small number of common parts of the word map from the other contents.

  According to the present invention, content including a concept that the user has not touched until now is extracted.

It is a block diagram which shows the structure of the content recommendation system which concerns on embodiment. It is a figure which shows an example of the word network which concerns on embodiment. It is a figure which shows the production | generation process of the word map which concerns on embodiment. It is a figure which shows the example of a comparison process of the word map which concerns on embodiment. It is a figure which shows an example of the program recommendation which concerns on embodiment.

Hereinafter, an example of an embodiment of the present invention will be described.
The content recommendation system 1 of the present embodiment relates to content (for example, a program) viewed by the user with the content viewing device 20 and content from a different viewpoint from the content viewed by the content search device 10 (content search device). ) Searches and recommends to users.

FIG. 1 is a block diagram showing a configuration of a content recommendation system 1 according to the present embodiment.
The content search device 10 constituting the content recommendation system 1 includes a storage unit 11 and a control unit 12. Further, the content search device 10 is connected to the content viewing device 20 via a predetermined network (for example, the Internet).

  The content search apparatus 10 extracts content to be presented to the user anew using content information including the user's content viewing history and post data such as reviews performed on the content.

  The storage unit 11 is a storage area for various programs and various data for causing the hardware group to function as the content search device 10, and may be a ROM, a RAM, a flash memory, a hard disk (HDD), or the like. Specifically, the storage unit 11 stores a content search program that causes the control unit 12 to execute each function of the present embodiment, a word network storage unit 111, a content storage unit 112, a history storage unit 113, A selection map storage unit 114.

  The word network storage unit 111 stores word relationship data describing a relationship between a plurality of words. This word relation data is large-scale data created from an existing corpus, and may be created automatically or manually by an administrator or the like.

FIG. 2 is a diagram illustrating an example of a word network according to the present embodiment.
The word relation data forms a word network in which words are connected by relation links. The relationship link includes a co-occurrence relationship in which words appear in the document at the same time, a causal relationship in which one word represents the cause or result of the other word, a relationship in which one word is a higher or lower concept of the other word, Synonyms, synonyms, synonyms, etc. are set.
Note that the word network used in the present embodiment may be configured from all word relationships or may be configured using only a part thereof.

  The content storage unit 112 stores content that can be provided to the user for viewing and content information associated with the content. The content information is text data indicating the content, and includes, for example, program guide data, program introduction data, subtitles or text broadcast data.

  The history storage unit 113 stores content selection history for each user, that is, history of content viewed by the user. This history includes date and time information of viewing the content.

  The selection map accumulation unit 114 stores, in the word relation data, a word that matches the word map generated for the content viewed (selected) by the user, in association with the viewed user, together with the number of selections.

  The control unit 12 is a part that controls the entire content search apparatus 10, and implements various functions in the present embodiment by appropriately reading and executing various programs stored in the storage unit 11. The control unit 12 may be a CPU (Central Processing Unit).

  The control unit 12 includes a word information acquisition unit 121, a word map generation unit 122, a word map comparison unit 123, a content provision unit 124, and a comparison information generation unit 125.

The word information acquisition unit 121 extracts words included in the content information acquired from the content viewing apparatus 20 or the content storage unit 112.
Here, when the user (U) selects the content (C0) for viewing / browsing, the content viewing apparatus 20 includes the content (C0) together with information that the content (C0) is viewed by the user (U). The content information (CI0) given to is transmitted to the content search apparatus 10.

The word information acquisition unit 121 divides a word into each of the content information (CI0) received from the content viewing apparatus 20 and the content information (CI1 to n) received from the content storage unit 112, and extracts an important word. .
Specifically, the word information acquisition unit 121 performs a morphological analysis process on the text included in the content information and divides it into words. Furthermore, the word information acquisition unit 121 performs importance calculation processing for each of the divided words, and assigns importance values to the words. The word information acquisition unit 121 extracts important words by excluding words that do not satisfy a predetermined importance level.

Here, the importance is, for example, an index (for example, TF-IDF) calculated based on the appearance frequency for each word extracted by word division.
The importance is an index based on the number of relation links in word relation data for words extracted by word division, for example.

  The word map generation unit 122 uses the word network information stored in the word network storage unit 111 to add words related data to the word group extracted by the word information acquisition unit 121 for each content. A map (M0, M1-n) is generated.

FIG. 3 is a diagram showing word map generation processing according to the present embodiment.
The word map generation unit 122 searches whether each word obtained from the word information acquisition unit 121 exists on the word network. If it exists on the word network, the word map generator 122 extracts these words. Further, when there are related links between these words on the word network, the word map generating unit 122 also extracts these related links.

For example, when the word group obtained from the word information acquisition unit 121 is A, E, F, G, and J, the word map generation unit 122 extracts the relationship link existing between these words and words from the word network. The extracted word map is M0.
Further, since the importance level is assigned to each word group obtained from the word information acquisition unit 121, the word map generation unit 122 performs filtering using this importance level, and the words A and E having high importance levels. , F may be used to generate a word map M0 ′.

The word map comparison unit 123 compares the word map for the content selected by the user with the word map for the other content stored in the content storage unit 112, and from these contents, there are few common parts of the word map. Extract content preferentially.
Specifically, the word map comparison unit 123 shares at least some of the important words determined to have a relatively high importance among the words included in the word map for the content selected by the user, and Content with a long distance between word maps is preferentially extracted.

  At this time, the word map comparison unit 123 weights the distance between the word maps based on the type of relation link in the word relation data. For example, compared to synonyms or synonyms, it is determined that the distance between words in the synonym relationship is far.

  In addition, the word map comparison unit 123 preferentially extracts content that is not included in the viewing history of the history storage unit 113. Furthermore, the word map comparison unit 123 preferentially extracts content corresponding to a word map having few common parts with the word group stored in the selection map accumulation unit 114. When the selection map storage unit 114 also stores the number of selections, the word map comparison unit 123 preferentially extracts content corresponding to a word map including a word with a small number of selections.

  Then, the word map comparison unit 123 compares the map with each of the word maps M1 to Mn for the other content stored in the content storage unit 112 with respect to the word map M0 for the content viewed by the user. Ranking is performed in order from the farthest, and the ranking data is passed to the content providing unit 124.

FIG. 4 is a diagram illustrating an example of word map comparison processing according to the present embodiment.
First, the word map comparison unit 123 selects one or more important words included in the word map M0 generated from the content information CI0. This important word is, for example, a word that has been determined to have the highest importance by the word information acquisition unit 121, a word that has been given a certain degree of importance, or a word related to content that has been viewed by many users. .

  In this example, the word A is selected as the important word. The word map comparison unit 123 extracts a word map including the word A from the word maps M1 to Mn generated from the content information CI1 to CIn of the content storage unit 112. As a result, word maps M1 and M2 were extracted.

  Here, M0 and M1 share four words (A, E, F, G) and three related links (a, g, h), and M0 and M2 share two words (A , D) and one related link (b). Therefore, the content information CI0 and CI1 are close in content because there are many common words and relationships, and the content information CI0 and CI2 are determined as being far from each other because the number of common words and relationships is small. That is, the content C1 is similar to the content C0 that has already been viewed, and the content C2 is determined to be different.

  Since the important word A is included in common, C1 and C2 are content related to a common topic, but C2 is likely to be content including many viewpoints different from C0. In other words, by recommending C2 to C1 instead of C1 or together with C1, the user can be provided with content that includes a different viewpoint from the viewed content C0, and the user's viewpoint can be expanded. .

  The content providing unit 124 selects a word map Mx that is far from the word map M0 based on the ranking data acquired from the word map comparison unit 123, and provides or recommends the content Cx corresponding to the word map to the user. To do.

  In addition, the word map comparison unit 123 refers to the history storage unit 113 or the selection map storage unit 114 and preferentially extracts content for efficiently filling a word region that has not yet been touched by the user. Thereby, the content search apparatus 10 can make a user view the content based on various viewpoints with sufficient balance.

  The comparison information generation unit 125 generates comparison information between the word relationship data and the word data for each user in the selection map storage unit 114. By outputting this comparison information, the viewing range and frequency for each user in the word network can be grasped.

  In addition, the content search device 10 adds weights to words on the word network each time the user views the content, and stores the weights in the selection map storage unit 114, so that the region viewed by the user, and An unknown area becomes clear. The comparison information generation unit 125 can present such information to the user.

FIG. 5 is a diagram showing an example in which the content recommendation system 1 according to the present embodiment is applied to program recommendation.
The content search device 10 is realized, for example, in the form of a website having a VOD (Video On Demand) function and a program review function.

  The user searches and views the program C0 that suits his / her preference. At this time, the user posts a review regarding the viewed program C0. Each program is given program-related information indicating an outline of the program. The content search apparatus 10 accumulates user viewing history, post data, and program related information.

  The content search apparatus 10 stores a word network generated from a large-scale language resource (corpus). When the program (C0 to Cn) to be provided to the user is registered, the content search device 10 collates words extracted from program-related information or posted data of each program with a word network, and a word map ( M0 to Mn).

  When the user views the program C0, the content search device 10 searches for a word map of another program including the same important word K using the important word K of the word map M0 corresponding to the program C0 as a key. Then, the content search apparatus 10 compares M0 with the word map obtained as a result of the search, and ranks the word maps in descending order of the number of common words and related links.

  When the word map Mx having the smallest number of common words and related links is extracted as a result of ranking, the content search apparatus 10 recommends a program Cx corresponding to the word map Mx to the user. For example, the content search apparatus 10 may provide another aspect (for example, improvement of welfare) to a user who has watched a program dealing with a subject aspect (for example, labor cost reduction, purchase refusal) regarding a certain theme “tax increase”. Can be recommended, and the range of knowledge of the user can be expanded.

  Further, the content search apparatus 10 can generate a word map using not only program related information but also posted sentences such as reviews or comments posted by a plurality of users on the program C0. Thereby, it is possible to provide a program with a new viewpoint that is not included in the user's idea.

When the user's posted text is targeted, the content storage unit 112 stores text posted by the user for each content. Post data used for word extraction and word map generation may be for a user other than the user who is viewing the content.
At this time, in the word map related to the content viewed by many users, the weight between words may be set large. Thereby, since the priority of the area in which many users are interested is adjusted, the content search apparatus 10 can control the recommendation order of the content in accordance with a request such that it matches or does not match other users' preferences.

As described above, according to the present embodiment, the content recommendation system 1 compares the word map generated from the information of the content viewed by the user with the word map generated from the information of the other content, and the common part Extract content with less priority. Therefore, the content including the concept that the user has not touched until now is recommended while being related to the viewed content.
At this time, the content recommendation system 1 can preferentially recommend content having different viewpoints because the important words are common and highly relevant and the distance between the word maps is long.

  Moreover, since the content recommendation system 1 weights the distance between the word maps based on the type of the related link, the content recommendation system 1 can more accurately determine the conceptual difference between the contents and appropriately recommend the content to the user.

Further, the content recommendation system 1 excludes words that do not satisfy a predetermined importance when generating a word map. Therefore, the content recommendation system 1 can generate a word map that more accurately represents the content content by eliminating words that are noises that are not closely related to the content content.
At this time, the content recommendation system 1 can calculate the importance level easily and accurately using, for example, an index based on the number of TF-IDFs or related links.

  Further, the content recommendation system 1 stores a selection history in which the user has viewed the content, and preferentially recommends content not included in the selection history. Therefore, the content recommendation system 1 can preferentially recommend content including a concept that has not been viewed by the user, that is, the user has not touched until now.

Also, the content recommendation system 1 stores words that match the word map for the content viewed by the user, extracts a word map that has few common parts with these word groups, and recommends the content. Therefore, the content recommendation system 1 can preferentially recommend content that includes words that the user has not touched on the word network, that is, has a different concept.
Furthermore, since the content recommendation system 1 stores the words related to the viewed content together with the number of selections and selects the word map so as to equalize the number of selections, it is possible to efficiently select contents of various concepts.

  Further, the content recommendation system 1 can generate and output comparison information between a word network and a word group or a word map obtained from content viewed by the user. Therefore, the content recommendation system 1 can allow the user to easily grasp the area that the user frequently views, the area where the viewing frequency is low, and the like.

  In addition, the content recommendation system 1 can employ a user's posted text as the content information. As a result, the amount of information increases and the thought of the user is reflected in the posted text, so that suitable content is recommended for each user.

  As mentioned above, although embodiment of this invention was described, this invention is not restricted to embodiment mentioned above. Further, the effects described in the present embodiment are merely a list of the most preferable effects resulting from the present invention, and the effects of the present invention are not limited to those described in the present embodiment.

The selection map storage unit 114 uses the previous or predetermined number of viewing histories, the viewing history of a specific period (for example, the past one week), or the accumulated viewing history of the entire period, etc. May be generated and stored.
As described above, in the process of viewing the content, the words related to the content viewed in the predetermined period are recorded, so that the comparison information generating unit 125 allows the user to perform the comparison on the word network during the predetermined period. An area related to content that has already been viewed and an area that has not been viewed can be output.

In the present embodiment, the content information used when generating the word map may be program related information, posted text related to the program, or a combination thereof.
Further, in both cases of using the user's viewing history and using the user's posted sentence, the content recommendation system 1 has a word map that is close to a word map related to content that the user has viewed in the past and a distant word map, respectively. You may present the content corresponding to. Thereby, the content recommendation system 1 can present both the content close to the region where the user's interest is high and the content of the region where the interest is low, and can provide either one according to the user's selection.

In addition, the content search method according to the present invention is not limited to the content recommendation system 1 realized by the above-described website, and may be implemented as a program recommendation function in a smart TV, for example.
Furthermore, although the present embodiment has been described assuming video content such as a program as content, the target content is not limited to this, and the present invention is not limited to content services provided by content providers, The present invention can be applied to intangible contents or tangible contents that are handled in general retail services for actual products.

For example, when the target content is “travel tour information”, the content recommendation system 1 generates a word map from travel tour information (summary, place name, etc.) selected by the user in the past, and the distance from this word map is A tour corresponding to another distant word map may be recommended. As a result, a travel tour that has never been performed in the past and seems to be new to the user is recommended.
For example, when the target content is a tangible product, the content recommendation system 1 can recommend a product with a different viewpoint from the product purchased in the past.

  In this embodiment, each part with which the content search device 10 is provided may be distributed to a plurality of devices. For example, the content providing device corresponding to the content providing unit 124 may be configured to receive content ranking information from the content search device 10 and present recommended content to the content viewing device 20. The history storage unit 113 may be included in the content providing device or the content viewing device 20.

  In the present embodiment, the configuration and operation of the content search apparatus have been mainly described. However, the present invention is not limited to this, and includes each component and is configured as a method or program for recommending appropriate content to the user. May be.

  Further, it may be realized by recording a program for realizing the function of the content search apparatus on a computer-readable recording medium, causing the computer system to read and execute the program recorded on the recording medium. .

  The “computer system” here includes an OS and hardware such as peripheral devices. The “computer-readable recording medium” refers to a storage device such as a portable medium such as a flexible disk, a magneto-optical disk, a ROM, and a CD-ROM, and a hard disk built in the computer system.

  Furthermore, “computer-readable recording medium” means that a program is dynamically held 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. It is also possible to include one that holds a program for a certain time, such as a volatile memory inside a computer system that becomes a server or client in that case. Further, the program may be for realizing a part of the above-described functions, and may be capable of realizing the above-described functions in combination with a program already recorded in the computer system. .

DESCRIPTION OF SYMBOLS 1 Content recommendation system 10 Content search apparatus 20 Content viewing apparatus 111 Word network storage part 112 Content storage part 113 History storage part 114 Selection map storage part 121 Word information acquisition part 122 Word map generation part 123 Word map comparison part 124 Content provision part 125 Comparison information generator

Claims (13)

A word network storage unit for storing word relationship data describing a relationship between a plurality of words;
A content storage unit for storing content information;
A word information acquisition unit that extracts words included in the content information;
For each content, a word map generation unit that generates a word map with the word relation data added to the words extracted by the word information acquisition unit;
A word map comparison unit that compares a word map with respect to content selected by the user with a word map with respect to other content, and preferentially extracts content with few common parts of the word map from the other content. Content search device.   The word map comparison unit includes at least a part of important words determined to be relatively high among words included in the word map for the content selected by the user, and the word maps The content search apparatus according to claim 1, wherein content with a long distance is preferentially extracted.   The content search device according to claim 2, wherein the word map comparison unit weights the distance between the word maps based on a type of relation link in the word relation data.   The said word map production | generation part excludes the word which is less than predetermined | prescribed importance among the words extracted by the said word information acquisition part, The said word map is produced | generated in any one of Claims 1-3. The content search device described.   The content search apparatus according to claim 4, wherein the importance is an index based on a TF-IDF for a word extracted by the word information acquisition unit.   The content search apparatus according to claim 4, wherein the importance is an index based on the number of related links in the word relationship data for the word extracted by the word information acquisition unit. A history storage unit for storing a selection history of content for each user;
The content search device according to any one of claims 1 to 6, wherein the word map comparison unit preferentially extracts content not included in the selection history. A selection map storage unit that stores words that match the word map for the content selected by the user in the word relation data in association with the user;
The content according to any one of claims 1 to 7, wherein the word map comparison unit preferentially extracts content corresponding to a word map having few common parts with the word group stored in the selection map storage unit. Search device. The selection map storage unit stores words that match the word map for the content selected by the user, together with the number of selections,
The content search apparatus according to claim 8, wherein the word map comparison unit preferentially extracts content corresponding to a word map including a word with a small number of selections.   The content search apparatus according to claim 8 or 9, further comprising a comparison information generation unit that generates comparison information between the word relation data and the word data for each user in the selection map storage unit.   The content search apparatus according to any one of claims 1 to 10, wherein the content information is a posted message of a user with respect to the content. A word network including word relationship data describing a relationship between a plurality of words, and a computer capable of referencing content information;
A word information acquisition step of extracting words contained in the content information;
For each content, a word map generation step for generating a word map with the word relation data added to the words extracted in the word information acquisition step;
A word map comparison step that compares the word map for the content selected by the user with the word map for the other content, and preferentially extracts content with a small number of common parts of the word map from the other content. To search for content. A word network including word relationship data describing a relationship between a plurality of words, and a computer capable of referencing content information;
A word information acquisition step of extracting words contained in the content information;
For each content, a word map generation step for generating a word map with the word relation data added to the words extracted in the word information acquisition step;
A word map comparison step that compares the word map for the content selected by the user with the word map for the other content, and preferentially extracts content with a small number of common parts of the word map from the other content. Content search program to let you.

Images