JP2009116469A - Information extraction program and information extraction device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a user with information precious to the user by properly managing information contributed from each user. <P>SOLUTION: An information management server 100 is configured to acquire an article (including a blog or the like) created in a prescribed period from a storage device, and extracts a keyword included in the acquired article as a feature keyword. On the basis of an article and the feature keyword created by an object user or a community configured of a plurality of users, the information management server 100 calculates the featured quantity (first TF/IDF value) of each feature keyword for the user or the community, and extracts the feature keyword whose calculated featured quantity is equal to a threshold or more as a profile keyword showing the feature (profile) of the user or the community. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an information extraction program and an information extraction apparatus for extracting predetermined information from a storage device storing article information.

  In recent years, companies have established network environments, and for example, in-house SNS (Social Network Service) as disclosed in Non-Patent Document 1 and information tools such as groupware are routinely used. Users (employees, etc.) use such information tools to publish articles (including blogs, etc.) created by themselves to other employees, and refer to articles created by other users. ing.

  In Patent Document 1, in order to improve the efficiency of the user's article search, a plurality of blog articles are analyzed, a word as a keyword is extracted, and when one of the keywords is selected, the selected keyword is selected. A technique for extracting related keywords from the latest predetermined number of blog articles is published.

JP 2007-233438 A FUJITSU SOFTWARE TECHNOLOGIES, “SNS“ Chisokukan ””, [online], search on October 26, 2007, Internet <URL: http://jp.fujitsu.com/group/fst / services / chisokukan / ＞

  However, information tools such as in-house SNS and groupware are used on a daily basis, but there is no mechanism for efficiently reusing that information (for example, articles created by users), so there is valuable information. Nevertheless, there is a problem that this valuable information is not reused by the user and is buried in other information.

  This problem can be attributed to the fact that articles and blogs posted by users can be freely posted, and that information is not managed. However, in the past, it has not been possible to uniquely determine how to manage these pieces of information, and the information is not properly managed.

  The present invention has been made to solve the above-described problems caused by the prior art, and can appropriately manage information posted by each user and provide the user with valuable information. An object is to provide an information extraction program and an information extraction device.

  In order to solve the above-described problems and achieve the object, this information extraction program acquires an article created within a predetermined period from a storage device and extracts keywords included in the acquired article as feature keywords. A feature amount of each feature keyword for the user or the community is calculated based on a feature keyword extraction procedure to be performed, an article created by a user or a community composed of a plurality of users, and the feature keyword. It is a requirement to execute a quantity calculation procedure and a profile keyword extraction procedure for extracting a feature keyword whose feature quantity is equal to or greater than a threshold value as a profile keyword indicating a feature of the user or community.

  Further, in the information extraction program, in the information extraction program described above, the feature amount calculation procedure extracts an article created by the user or community from the storage device, and extracts the feature keyword included in the extracted article. A first counting procedure for counting a first value indicating a number, a second counting procedure for counting a second value indicating the total number of articles created by the user or community, and created by the user or community A third counting procedure for counting a third value indicating the number of articles including the feature keyword in the posted articles, and a calculation procedure for calculating a feature value based on the first, second, and third values Is a requirement.

  In addition, the information extraction program includes, in the information extraction program described above, a reference profile keyword group that associates each profile keyword indicating characteristics of a user or community serving as a reference and the number of appearance of the profile keyword in the article, Similarity is calculated based on each profile keyword indicating the characteristics of other users or communities and other profile keyword groups including the number of appearance of the profile keyword in the article, and based on the similarity It is a requirement that the computer further execute a similarity calculation procedure for extracting other users or communities similar to the reference user or community.

  Further, in the information extraction program, the information extraction program extracts articles within a predetermined period from the storage device, and counts a fourth value indicating the number of the feature keywords included in the extracted articles. A fourth counting procedure, a fifth counting procedure for counting a fifth value indicating the total number of articles in a predetermined period, and a sixth number indicating the number of articles including the characteristic keyword in the articles in the predetermined period Based on the sixth counting procedure for counting values and the fourth, fifth, and sixth values, the second feature amount of each feature keyword is calculated, and the calculated second feature amount is equal to or greater than a threshold value. It is a requirement that the computer further execute a trend keyword extraction procedure for extracting a keyword as a trendy keyword.

  The information extraction apparatus is an information extraction apparatus that manages an article created by a user or a community composed of a plurality of users and extracts predetermined information from the article, the user or community Article storage means for storing articles created by the above, feature keyword extraction means for acquiring articles created within a predetermined period from the article storage means, and extracting keywords included in the acquired articles as feature keywords; Feature quantity calculating means for calculating a feature quantity of each feature keyword for the user or community based on an article created by the user or community and the feature keyword, and a feature keyword having the feature quantity equal to or greater than a threshold value As a profile keyword indicating the characteristics of the user or community It is a requirement in that it comprises a profile keyword extracting means, a to.

  According to this information extraction program, an article (including a blog or the like) created within a predetermined period is acquired from the storage device, and a keyword included in the acquired article is extracted as a feature keyword. Then, the feature amount (first TF / IDF value) of each feature keyword for the user or community is calculated based on the article created by the target user or the community composed of a plurality of users and the feature keyword. Since the feature keyword whose calculated feature amount is equal to or greater than the threshold is extracted as a profile keyword indicating the feature (profile) of the user or community, information useful to the user can be provided to the user.

  Further, according to this information extraction program, an article created by a user or a community is extracted from the storage device, and the first value indicating the number of feature keywords included in the extracted article is created by the user or the community. A second value indicating the total number of articles, a third value indicating the number of articles including a feature keyword among the articles created by the user or the community, respectively. Since the feature amount is calculated based on the profile keyword, it is possible to efficiently extract the profile keyword that represents the feature of the user.

  Further, according to the information extraction program, each profile keyword indicating characteristics of a reference user or community and a reference profile keyword group that associates the number of appearance of the profile keyword in an article with another user or The similarity is calculated based on each profile keyword indicating the characteristics of the community and another profile keyword group including the number of occurrences of the profile keyword in the article, and a user or a reference user based on the similarity Since other users or communities similar to the community are extracted, a user or community similar to the reference user or community can be extracted more accurately.

  Further, according to the information extraction program, articles within a predetermined period are extracted from the storage device, the fourth value indicating the number of feature keywords included in the extracted article, and the total number of articles within the predetermined period are indicated. A fifth value and a sixth value indicating the number of articles including the feature keyword in articles within a predetermined period are counted, and each feature keyword is counted based on the counted fourth, fifth, and sixth values. Since the second feature amount is calculated and the feature keyword whose calculated second feature amount is equal to or greater than the threshold is extracted as the trendy keyword, the trendy keyword can be efficiently extracted.

  Exemplary embodiments of an information extraction program and an information extraction apparatus according to the present invention will be explained below in detail with reference to the accompanying drawings.

  First, features of the information sharing system according to the present embodiment will be described. In the information sharing system according to the present embodiment, the information management server acquires articles (including blogs and the like) created within a predetermined period from the storage device, and extracts keywords included in the acquired articles as feature keywords. Then, the information management server uses the feature amount (first TF / first TF) for each user or community based on the article and the feature keyword created by the target user or the community composed of a plurality of users. IDF value) is calculated, and a feature keyword whose calculated feature amount is equal to or greater than a threshold value is extracted as a profile keyword indicating a feature (profile) of the user or community.

  As described above, in the information sharing system according to the present embodiment, the information management server extracts the feature keyword, and extracts the profile keyword based on the extracted feature keyword and the article created by the target user or community. Therefore, information useful for the user can be provided to the user.

  In the information sharing system according to the present embodiment, the similarity is calculated based on the profile information of each user (including each community), and other uses similar to the user based on the calculated similarity. Since users and communities (similar profiles) are extracted, information useful to the users can be provided to the users.

  Further, in the information system according to the present embodiment, an article created within a predetermined period is extracted, and a feature amount (second TF / IDF value) is calculated based on the extracted article and a feature keyword. Since keywords whose feature amount is equal to or greater than the threshold are extracted as trendy keywords (trend keywords), information useful to the user can be provided to the user.

  Next, the configuration of the information sharing system according to the present embodiment will be described. FIG. 1 is a block diagram illustrating the configuration of the information sharing system according to the present embodiment. As shown in the figure, this information sharing system includes user terminals 10 to 30 used by a user and an information management server 100 that manages various types of information. The user terminals 10 to 30 and the information management server 100 are connected to each other via a network 50. In addition, although the user terminals 10-30 are shown here for convenience of explanation, this information sharing system may have other user terminals.

  Here, the user terminals 10 to 30 are devices that execute data communication with the information management server and output information on articles created by the user to the information management server 100. Also, the user terminals 10 to 30 acquire from the information management server 100 information to be displayed on a screen including information on articles created by users or communities and the above-described profile keywords, trend keywords, similar profiles, and the like. Display on the monitor.

  FIG. 2 is a diagram illustrating an example of a screen image displayed on the monitor of the user terminals 10 to 30. As shown in the figure, in this screen image, a profile keyword of Mr. Fujitaro, a similar profile of Mr. Taro Fuji, and a trend keyword are displayed. By referring to this screen, the user can efficiently acquire various useful information without performing keyword search by himself / herself, so that important information is not missed.

  The information management server 100 acquires and manages article information from the user terminal 10, extracts the above-described feature keywords, profile keywords, similar profiles, and trend keywords, and displays information for display on the user's screen. It is an apparatus provided to the terminals 10-30. FIG. 3 is a functional block diagram illustrating the configuration of the information management server 100 according to the present embodiment. As shown in the figure, the information management server 100 includes an input unit 110, an output unit 120, a communication control IF unit 130, an input / output control IF unit 140, a storage unit 150, and a control unit 160. Configured.

  Among these, the input unit 110 is a means for inputting various types of information, and includes a keyboard, a mouse, a microphone, and the like. For example, the input unit 110 accepts update information of each table 150a to 150l stored in the storage unit 150 described later. Enter. A monitor (output unit 120) described later also realizes a pointing device function in cooperation with the mouse.

  The output unit 120 is an output unit that outputs various types of information. The output unit 120 includes a monitor (or display, touch panel), a speaker, and the like. For example, the output unit 120 outputs information on each table 150a to 150l stored in the storage unit 150 described later. To do.

  The communication control IF unit 130 is means for controlling data communication with the user terminals 10 to 30 mainly via the network 50. The input / output control IF unit 140 is a unit that controls input / output of data by the input unit 110, the output unit 120, the communication control IF unit 130, the storage unit 150, and the control unit 160.

  The storage unit 150 is a storage unit (storage unit) that stores data and programs necessary for various processes performed by the control unit 160. In particular, as shown in FIG. Table 150a, article information table 150b, execution management table 150c, synonym / joint word table 150d, keyword table 150e, article keyword table 150f, unnecessary word table 150g, article space management table 150h, profile An information table 150i, a profile keyword table 150j, a trend keyword table 150k, and a similar profile table 150l are provided.

  Below, each table 150a-150l which the memory | storage part 150 memorize | stores is demonstrated in order. The article management table 150a is a table for managing information on the article and the user or community that created the article. FIG. 4 is a diagram illustrating an example of the data structure of the article management table 150a.

  As shown in FIG. 4, the article management table 150a updates an article ID (Identification) for identifying an article, a poster (including a user and a community), a profile ID for identifying the poster, a posting destination, and an article body. Content update date and time indicating the date and time, the article text, the total number of comments indicating the number of comments on the article, and a deletion flag. The deletion flag is information indicating whether or not the information of the corresponding line is valid. When the deletion flag is “on”, the information of the corresponding line is valid. When the deletion flag is “off”, the information of the corresponding line is displayed. Is invalid (the same applies to the deletion flag of the table described below, so the description of the deletion flag below is omitted).

  The article information table 150b is a table for managing whether or not a feature keyword has been extracted for each article. FIG. 5 is a diagram illustrating an example of a data structure of the article information table 150b. As shown in the figure, the article information table 150b includes an article ID, a poster (including a user and a community), a posting destination, a content update date, a keyword extraction flag, and a deletion flag. Among these, the keyword extraction flag is information indicating whether or not the feature keyword has been extracted from the article identified by the article ID of the corresponding line. When the keyword extraction flag is “on”, it indicates that the feature keyword has been extracted from the article, and when it is “off”, it indicates that the feature keyword has not been extracted from the article.

  The execution management table 150c is a table that stores the execution start date and time of processing for extracting feature keywords from articles (feature keyword extraction processing) and processing for extracting user and community profile keywords (profile keyword extraction processing). FIG. 6 is a diagram illustrating an example of the data structure of the execution management table 150c. As shown in the figure, the execution management table 150c has a process type and an execution start date and time.

  The synonym / joint word table 150d is a table that manages various synonyms / joint words and stores the replacement source keyword and the replacement keyword in association with each other. FIG. 7 is a diagram illustrating an example of a data structure of the synonym / joint word table 150d. As shown in the figure, the synonym / joint word management table 150d has a replacement source keyword, a replacement keyword, and an update date and time. According to the synonym / joint word table 150d shown in FIG. 7, for example, the keyword “computer” extracted from the article is replaced with “computer”.

  The keyword table 150e is a table that stores various types of information related to keywords (feature keywords) included in articles. FIG. 8 is a diagram illustrating an example of the data structure of the keyword table 150e. As shown in the figure, the keyword table 150e includes a keyword ID, a keyword (feature keyword), a normalized keyword, the total number of keyword appearances, the number of contained contents, and an unnecessary word flag.

  Among these, the normalized keyword indicates a normalized keyword. For example, “abc main store” is normalized to “ABC main store”. The total number of keyword appearances indicates the total number of occurrences of a predetermined keyword in all article texts in the article management table 150a. In the example shown in the first row of FIG. 8, it is stored that the keyword “computer” appears “10000 times” in the article text.

  The number of contained contents indicates the number of articles including a predetermined keyword among the articles in the article management table 150a. In the example shown in the first row of FIG. 8, it is stored that the article including the keyword “computer” is “50”.

  The unnecessary word flag is information indicating whether or not the keyword of the corresponding line is the target of the feature keyword. When the unnecessary word flag is “on”, it indicates that the keyword of the corresponding line is not a candidate for the feature keyword, and when it is “off”, it indicates that the keyword of the corresponding line is a candidate for the feature keyword. Yes.

  The article keyword table 150f is a table that stores a keyword ID of a feature keyword and an article including the feature keyword in association with each other. FIG. 9 is a diagram illustrating an example of the data structure of the article keyword table 150f. As shown in the figure, the article keyword table 150f has an article ID, a keyword ID, a keyword appearance count, and a deletion flag.

  Here, the keyword identified by the keyword ID in FIG. 9 indicates a feature keyword. The keyword appearance count indicates the number of times that the feature keyword identified by the keyword ID appears in the article identified by the article ID.

  The unnecessary word table 150g is a table that stores keywords (unnecessary words) that are not candidates for feature keywords. FIG. 10 is a diagram illustrating an example of a data structure of the unnecessary word table 150g. As shown in the figure, the unnecessary word table 150g stores various unnecessary words.

  The article space management table 150h is a table for managing personal information of contributors (including users and communities). FIG. 11 is a diagram illustrating an example of the data structure of the article space management table 150h. As shown in the figure, the article space management table 150h has a profile ID for identifying a contributor, article space identification information, a name, an update date and time indicating the date and time when the corresponding line is updated, and a deletion flag. Of these, the article space identification information is information indicating whether the poster corresponding to the profile ID is an individual (single user) or a community.

  The profile information table 150i is a table that stores various types of information corresponding to profile IDs. FIG. 12 is a diagram illustrating an example of the data structure of the profile information table 150i. As shown in the figure, this profile information table 150i includes a profile ID, article space identification information, keyword extraction flag, similarity calculation flag, total number of articles indicating the total number of articles created by the poster, and articles created by the poster. The total number of commented on and a deletion flag.

  The keyword extraction flag in FIG. 12 is information indicating whether or not the profile keyword corresponding to the poster on the corresponding line has been extracted. When the keyword extraction flag is “on”, it indicates that the profile keyword has been extracted, and when it is “off”, it indicates that the profile keyword has not been extracted.

  The similarity calculation flag is information indicating whether or not a similarity (a value used when a similar profile is extracted and details will be described later) corresponding to the poster of the corresponding line has been calculated. When the similarity calculation flag is “on”, it indicates that the similarity has been calculated, and when it is “off”, it indicates that the similarity is not calculated.

  The profile keyword table 150j is a table for managing profile keywords of contributors (including users and communities). FIG. 13 is a diagram illustrating an example of the data structure of the profile keyword table 150j. As shown in the figure, the profile keyword table 150j has a profile ID, article space identification information, keyword ID, first TF / IDF value, total number of keyword appearances, number of contained contents, unnecessary word flag, and deletion flag.

  The keyword ID included in FIG. 13 indicates the keyword ID of the profile keyword for the poster. For example, the profile keyword of the poster identified by the profile ID “p0001” is the keyword identified by the keyword ID “w0001” and the keyword identified by “w0002”.

  The first TF / IDF value indicates the feature amount of the feature keyword for the poster (including the user and the community). The larger the first TF / IDF value, the more the feature keyword that better represents the feature of the corresponding poster. . A method for calculating the first TF / IDF value will be described later. The total number of keyword appearances and the number of contained contents are the same as described above.

  The trend keyword table 150k is a table for managing trend keywords. FIG. 14 is a diagram illustrating an example of a data structure of the trend keyword table 150k. As shown in the figure, the trend keyword table 150k has a keyword ID, a second TF / IDF value, a total number of keyword appearances, and the number of contained contents.

  The keyword ID included in FIG. 14 indicates a keyword ID of a keyword that becomes a trend keyword. For example, the keyword identified by the keyword ID “w0001” is a trend keyword. The second TF / IDF value indicates the feature amount of the corresponding trend keyword. The larger the second TF / IDF value, the more popular the keyword. The total number of keyword appearances and the number of contained contents are the same as described above.

  The similar profile table 150l is a table for managing users and communities (similar profiles) similar to posters (including users and communities). FIG. 15 is a diagram illustrating an example of a data structure of the similar profile table 150l. As shown in the figure, the similar profile table 150l has a profile ID, a sequence number indicating the order of information, a similar profile ID, and a similarity.

  Of these, the similar profile ID indicates the profile ID of the user and community similar to the poster identified by the profile ID. For example, the similar profile of the poster identified by the profile ID “p0001” is the poster (user or community) identified by the profile ID (similar profile ID) “p0022”, “p0087”. The degree of similarity is information indicating how similar a poster with a similar profile ID is to a poster with a profile ID, and the greater the numerical value, the more similar.

  Returning to the description of FIG. 3, the control unit 160 is a control unit that has an internal memory for storing programs and control data that define various processing procedures, and executes various processes using these programs. As shown in FIG. 3, the information management unit 160a, the feature keyword extraction processing unit 160b, the profile keyword extraction processing unit 160c, the similarity calculation processing unit 160d, and the trend keyword extraction are closely related to the invention. A processing unit 160e and a service provision processing unit 160f are provided.

  The information management unit 160 a is a unit that manages the tables 150 a to 150 l stored in the storage unit 150. For example, when the update data for the tables 150a to 150l is acquired, the information management unit 160a updates the information stored in the tables 150a to 150l with the acquired update data.

  The feature keyword extraction processing unit 160 b is a means for extracting feature keywords using information in the tables 150 a to 150 l stored in the storage unit 150. 16 and 17 are diagrams illustrating specific processing of the feature keyword extraction processing unit 160b.

  First, referring to FIG. 16, the feature keyword extraction processing unit 160b acquires the execution start time <A> corresponding to the process type “feature keyword extraction processing” from the execution management table 150c (see (1) in FIG. 16). Then, various pieces of article information (article ID, contributor, posting destination, deletion flag, content update date and time) whose content update date is newer than <A> are acquired from the article management table 150a (see (2) in FIG. 16).

  Then, the feature keyword extraction processing unit 160b registers the various article information acquired in (2) in FIG. 16 in the article information table 150b (see (3) in FIG. 16), and executes the execution start date and time in the current process. Register in the management table 150c (see (4) in FIG. 16).

  Thereafter, the feature keyword extraction processing unit 160b compares the update date / time of the synonym / joint word table 150d with the content update date / time of the article information table 150b, and the content update date / time is the update date / time of the synonym / joint word table 150d. The keyword extraction flag for the article IDs that follow is set to “off” (not shown).

  Subsequently, in the description of FIG. 17, the feature keyword extraction processing unit 160b acquires an article ID list whose keyword extraction flag is “off” from the article information table 150b (see (1) of FIG. 17). Article text corresponding to each article ID acquired in (1) is acquired (see (2) in FIG. 17).

  The feature keyword extraction processing unit 160b extracts keywords from the article text ((3) in FIG. 17). In addition, when extracting a keyword from an article text, a morphological analysis or the like that is a well-known technique may be executed.

  Subsequently, the feature keyword extraction processing unit 160b acquires a list of replacement source keywords / replacement keywords from the synonym / joint word table 150d, and compares the acquired list with the keywords extracted in (3) of FIG. Thus, each keyword is replaced (see (4) in FIG. 17).

  Then, the feature keyword extraction processing unit 160b creates various information (normalized keyword, part of speech tag, keyword appearance total number) corresponding to the keyword (replaced keyword) (see (5) in FIG. 17). Note that the regular keyword includes a table (not shown) in which the regular keyword and the keyword are associated with each other, and the feature keyword extraction processing unit 160b uses the table to store the feature keyword extraction processing unit 160b. Normalize keywords. Further, it is assumed that the feature keyword extraction processing unit 160b counts the total number of keyword appearances by comparing the keyword and the article text stored in the article management table 150a.

  Subsequently, the feature keyword extraction processing unit 160b registers various information created in (5) of FIG. 17 in the keyword table 150e (see (6) of FIG. 17). Each keyword registered in the keyword table 150e is a feature keyword. Note that the feature keyword extraction processing unit 160b counts the number of contained contents by comparing the keyword and the article text stored in the article management table 150a.

  The feature keyword extraction processing unit 160b also stores various information (article ID, keyword ID, keyword appearance total number, deletion flag) in which the article ID is associated with the keyword (feature keyword) extracted from the article ID. It is registered in 150f (see (7) in FIG. 17), and the keyword extraction flag of the article information table that has been processed is set to “ON” (see (8) in FIG. 17).

  Thereafter, the feature keyword extraction processing unit 160b compares each unnecessary word stored in the unnecessary word table 150g with each keyword (or regularized keyword) in the keyword table 150e, and corresponds to a keyword line that matches the unnecessary word. The unnecessary word flag to be set is set to “ON” (not shown).

  Returning to the description of FIG. 3, the profile keyword extraction processing unit 160 c extracts a profile keyword indicating the feature of the poster based on the article and the feature keyword created by the poster (including the user and the community). Means. 18 and 19 are diagrams illustrating specific processing of the profile keyword extraction processing unit 160c.

  First, referring to FIG. 18, the profile keyword extraction processing unit 160c acquires the execution start time <A> corresponding to the processing type “profile keyword extraction processing” from the execution management table 150c (see (1) in FIG. 18). Then, various profile information (profile ID, article space identification information, deletion flag) whose update date is newer than <A> is acquired from the article space management table 150h (see (2) in FIG. 18).

  Then, the profile keyword extraction processing unit 160c registers the various profile information acquired in (2) of FIG. 18 in the profile information table 150i (see (3) of FIG. 18), and manages the execution start date and time in the current process. It is registered in the table 150c (see (4) in FIG. 18).

  Further, the profile keyword extraction processing unit 160c sets the keyword extraction flag of the profile ID updated after the execution start time <A> corresponding to the “profile keyword extraction processing” in the execution management table 150c to “off” (illustration shown). (Omitted).

  Subsequently, in the description of FIG. 19, the profile keyword extraction processing unit 160c acquires a list of profile IDs whose keyword extraction flag is “off” from the profile information table 150i (see (1) in FIG. 19), and article management. The total number of articles corresponding to the profile ID (the total number of articles created by the poster identified by the profile ID) and the total number of comments are obtained from the table 150a (see (2) in FIG. 19). The acquired various information, that is, the profile ID, the total number of articles, and the total number of comments are associated and registered in the profile information table 150i (see (3) in FIG. 19).

  Then, the profile keyword extraction processing unit 160c acquires an article (article ID) keyword associated with the profile ID acquired in (1) of FIG. 19 from the article keyword table 150f and the article information table 150b ((4 in FIG. 19). )reference). The article ID associated with the profile ID can be determined by referring to the article management table 150a, for example.

Subsequently, the profile keyword extraction processing unit 160c calculates the first TF / IDF value of each keyword (feature keyword) acquired in (4) of FIG. 19 (see (5) of FIG. 19). The specific formula for calculating the first TF / IDF value is:
First TF / IDF value = (Number of appearances of keywords of article with profile ID (contributor) <Keyword appearance total number>) × log {(Total number of articles of article created by profile ID (contributor)) / (Profile ID ( Number of articles that contain keywords in the articles created by contributors) <number of contents contained>)}
Can be represented by It should be noted that each value of the above expression is counted in advance by the profile keyword extraction processing unit 160c with reference to the tables 150a to 150l.

  Then, the profile keyword extraction processing unit 160c obtains various information related to the profile ID (profile ID, article space identification information, keyword ID, first TF / IDF value, keyword appearance total number, content number contained, unnecessary word flag, deletion flag). It is registered in the profile keyword table 150j (see (6) in FIG. 19), and the keyword extraction flag (keyword extraction flag in the profile information table 150i) of the profile ID registered in the profile keyword table 150j is set to “ON” (FIG. 19). (See (7)).

  Thereafter, the profile keyword extraction processing unit 160c compares each unnecessary word stored in the unnecessary word table 150g with each keyword (keyword identified by the keyword ID) in the profile keyword table 150j, and matches the unnecessary word. The unnecessary word flag corresponding to the line is set to “ON” (not shown).

  Of each keyword (keyword <characteristic keyword> identified by keyword ID) stored in the profile keyword table 150j, a keyword whose first TF / IDF value is equal to or greater than a threshold is extracted as a profile keyword corresponding to the corresponding profile ID. Will be.

  Returning to the description of FIG. 3, the similarity calculation processing unit 160 d is a means for calculating a similarity based on the profile keyword of each profile ID and extracting a similar profile ID based on the calculated similarity. . FIG. 20 is a diagram illustrating specific processing of the similarity calculation processing unit 160d.

  As illustrated in FIG. 20, the similarity calculation processing unit 160d acquires a list of profile IDs whose similarity calculation flag is “off” from the profile information table 150i (see (1) in FIG. 20), and (1) in FIG. The profile keyword list (profile ID, keyword ID, first TF / IDF value, total number of keyword appearances, number of contained contents, unnecessary word flag) of the profile ID acquired in (1) is acquired from the profile keyword table 150j ((2 in FIG. 20). )reference).

  Subsequently, the similarity calculation processing unit 160d acquires all profile ID lists from the profile information table 150i (see (3) in FIG. 20), and each profile keyword list of profile IDs acquired in (3) in FIG. (Profile ID, keyword ID, first TF / IDF value, keyword appearance total number, content content number, unnecessary word flag) are acquired from the profile keyword table 150j (see (4) in FIG. 20).

  Then, the similarity calculation processing unit 160d calculates the similarity between the profiles from the profile keyword list acquired in (2) and (4) in FIG. 20 (see (5) in FIG. 20). Here, as an example, there are posters A, B, and C identified by each profile ID, and A's profile keywords are (keyword 1 <10>, keyword 2 <5>, keyword 3 <8). >), B's profile keyword is (Keyword 1 <18>, Keyword 2 <56>, Keyword 9 <6>), and C's profile keyword is (Keyword 2 <12>, Keyword 7 <6>) In addition, a case where the similarity between Mr. B and Mr. C with respect to Mr. A is calculated will be described (the numerical values in <> are the total number of keyword appearances).

First, the similarity calculation processing unit 160d calculates the vectors of Mr. A, Mr. B, and Mr. C with Mr. A as a base point.
A's vector A = (10, 5, 8),
B's vector B = (18, 56, 0),
Mr. C's vector C = (0, 12, 0).

  Then, the similarity calculation processing unit 160d calculates the distance between the vectors as the similarity, and determines that the poster of each vector is similar when the calculated similarity is equal to or greater than the threshold. For example, when the distance between the vector A and the vector B is equal to or greater than the threshold value, Mr. A and Mr. B are similar, and the profile ID of Mr. B is set in the similar profile of Mr. A. In this case, the similar keyword targeting Mr. B is the keyword 2 that is the maximum total number of keyword appearances among the keywords of Mr. B.

  On the other hand, when the distance between the vector A and the vector C is less than the threshold value, the Mr. A and the Mr. B are not similar, and the profile ID of Mr. C is not set in the similar profile of Mr. A.

  The similarity calculation processing unit 160d sets the calculated similarity (and information associated therewith, that is, profile ID, sequence number, similar profile ID, similarity) in the similar profile table 150l (see (6) in FIG. 20). ). Here, the profile ID is the profile ID of the poster who has become the base point (Mr. A in the above example), and the similar profile ID is that of the poster (Mr. B in the above example) whose similarity is greater than or equal to the threshold value. Profile ID. Then, the similarity calculation processing unit 160d sets the similarity calculation flag of the processed profile ID to “ON” (see (7) in FIG. 20).

  Returning to the description of FIG. 3, the trend keyword extraction processing unit 160 e is a processing unit that extracts popular keywords (that is, trend keywords) using each table stored in the storage unit 150. FIG. 21 is a diagram illustrating specific processing of the trend keyword extraction processing unit.

  As shown in FIG. 21, the trend keyword extraction processing unit 160e acquires the total number of registered articles from the article information table 150b (see (1) in FIG. 21), and the content update date / time from the article information table 150b is within a predetermined range. Article ID is acquired (see (2) in FIG. 21).

  Then, the trend keyword extraction processing unit 160e acquires the keyword of the article corresponding to the article ID acquired in (1) of FIG. 21 from the article keyword table 150f (see (3) of FIG. 21), and for each of the acquired keywords. The second TF / IDF value is calculated (see (4) in FIG. 21).

Here, the specific formula for calculating the second TF / IDF value is:
2nd TF / IDF value = (total number of keyword appearances) x log (total number of articles / number of contained contents)
Can be represented by It should be noted that each value of the above equation is counted in advance by the trend keyword extraction processing unit 160e with reference to the tables 150a to 150l.

  The trend keyword extraction processing unit 160e registers various information related to the trend keyword (keyword ID, second TF / IDF value, keyword appearance total number, content content count) in the trend keyword table 150k (see (5) in FIG. 21).

  Of each keyword (keyword <characteristic keyword> identified by keyword ID) stored in the trend keyword table 150k, a keyword having a second TF / IDF value equal to or greater than a threshold value is extracted as a trend keyword.

  Returning to the description of FIG. 3, the service provision processing unit 160 f is a means for providing various services in response to service requests from the user terminals 10 to 30, and particularly as closely related to the present invention. When a predetermined poster (including a user and a community) is designated, a profile keyword, a similar profile, and a trend keyword corresponding to the designated poster are output (for example, the screen information described in FIG. 2) Is output to the user terminal that is the service request source).

  Specifically, when the service provision processing unit 160f extracts the poster keyword of the poster, the posting is performed by comparing the profile ID of the designated poster, the profile keyword table 150j, and the keyword table 150e. The profile keyword of the user. Note that the service provision processing unit 160f extracts and outputs a keyword whose first TF / IDF value is equal to or greater than a threshold value as a profile keyword.

  Further, when the service providing processor 160f extracts a similar profile of the poster, the poster ID is compared by comparing the profile ID of the designated poster, the similar profile table 150l, and the article management table 150a. Similar profiles (other users and communities similar to the designated contributor) are extracted and output.

  When the service provision processing unit 160f extracts a trend keyword, the service keyword processing unit 160f refers to the trend keyword table 150k, extracts a keyword whose second TF / IDF value is equal to or greater than a threshold value as a trend keyword, and outputs it.

  Next, a processing procedure of the information management server 100 according to the present embodiment will be described. FIG. 22 is a flowchart illustrating the processing procedure of the information management server 100 according to the present embodiment. As shown in the figure, the information management server 100 determines whether or not the article stored in the storage unit 150 has been updated (step S101). If the article has not been updated (No in step S102), the processing is performed. Exit.

  On the other hand, if the article has been updated (step S102, Yes), the feature keyword extraction processing unit 160b executes the feature keyword extraction processing (step S103), and the profile keyword extraction processing unit 160c executes the profile keyword extraction processing. (Step S104).

  Then, the similarity calculation processing unit 160d executes similarity calculation processing (step S105), and the trend keyword extraction processing unit 160e executes trend keyword extraction processing (step S106).

  Next, the feature keyword extraction process shown in step S103 of FIG. 22 will be described. FIG. 23 is a flowchart illustrating the processing procedure of the feature keyword extraction process. As shown in the figure, the feature keyword extraction processing unit 160b acquires the updated article (step S201), morphologically analyzes the article text, and decomposes the part of speech into units of “words” (step S202).

  Then, the feature keyword extraction processing unit 160b extracts the “word” that is the target of the feature keyword (step S203), and the “word” that is registered in the unnecessary word table 150g from the “word” that is the target of the feature keyword. Exclude (step S204).

  Subsequently, the feature keyword extraction processing unit 160b replaces the “word” that is the target of the feature keyword with the “word” registered in the synonym / joint word table 150d (step S205), and the feature keyword is replaced with the keyword table 150e. (Step S206), the characteristic keyword and the article are associated with each other and registered in the article keyword table 150f (step S207).

  Next, the profile keyword extraction process shown in step S104 of FIG. 22 will be described. FIG. 24 is a flowchart showing a processing procedure of profile keyword extraction processing. As shown in the figure, the profile keyword extraction processing unit 160c extracts an updated article from the article keyword table 150f, and extracts a feature keyword associated with the article (step S301).

  The profile keyword extraction processing unit 160c includes the total number of feature keywords in the articles for each target person (user) and community, the total number of articles for each target person and community, and the keywords for each target person and community. Based on the number of articles, the first TF / IDF value is calculated (step S302).

  The profile keyword extraction processing unit 160c sets n feature keywords whose first TF / IDF value is equal to or greater than a threshold as profile keywords (step S303), and registers the profile keywords in the profile keyword table 150j (step S304).

  Next, the similarity calculation process shown in step S105 of FIG. 22 will be described. FIG. 25 is a flowchart illustrating a processing procedure of similarity calculation processing. As shown in the figure, the similarity calculation processing unit 160d acquires a profile keyword of a user (including a community) as a similarity calculation target from the profile keyword table 150j (step S401), and a user as a comparison target Are obtained from the profile keyword table 150j (step S402). The comparison target users correspond to all users other than the calculation target users.

  Then, the similarity calculation processing unit 160d calculates the similarity based on the profile keyword of the user as the similarity calculation target and the profile keyword of the user as the comparison target (step S403), and the similarity is equal to or greater than the threshold value. Each profile ID is extracted as a similar profile (step S404). The similarity calculation processing unit 160d registers the extracted similar profile in the similar profile table 150l (step S405).

  In step 403, the top n number of profile keywords corresponding to the users to be calculated (in descending order of the first TF / IDF value) are taken as dimensions, and the total number of appearance of profile keywords for each user to be calculated and compared is calculated. The cosine distance between two vectors as values is calculated as the similarity.

  Next, the trend keyword extraction process shown in step S106 of FIG. 22 will be described. FIG. 26 is a flowchart illustrating a processing procedure of trend keyword extraction processing. As shown in the figure, the trend keyword extraction processing unit 160e extracts the latest n articles (or articles within a predetermined period) from the article keyword table, and extracts feature keywords associated with the articles (steps). S501).

  Then, the trend keyword extraction processing unit 160e calculates the second TF / IDF value based on the total number of feature keywords appearing in the articles within the target period, the total number of articles within the target period, and the number of articles including the feature keywords within the target range. Calculate (step S502).

  The trend keyword extraction processing unit 160e sets n feature keywords whose second TF / IDF value is equal to or greater than the threshold value as the trend keyword (step S503), and registers the trend keyword in the trend keyword table 150k (step S504).

  As described above, the feature keyword extraction processing unit 160b extracts the feature keyword, and uses the extracted feature keyword to extract the profile keyword, the similar profile, and the trend keyword. Therefore, information useful for the user is efficiently extracted. It comes out.

  As described above, the information management server 100 according to the present embodiment acquires articles (including blogs and the like) created within a predetermined period from the storage unit 150, and uses keywords included in the acquired articles as feature keywords. Extract. Then, the information management server 100 uses the feature amount (first TF) of each feature keyword for the user or community based on the article and the feature keyword created by the target user or the community composed of a plurality of users. / IDF value) is calculated, and feature keywords whose calculated feature amount is equal to or greater than the threshold are extracted as profile keywords indicating the features (profiles) of the user or community, so that information useful to the user is provided to the user. be able to.

  Further, the information management server 100 according to the present embodiment calculates the similarity based on each profile keyword of each user (including each community), and other similar to the user based on the calculated similarity. Since users and communities (similar profiles) are extracted, information useful to the users can be provided to the users.

  Further, the information management server 100 according to the present embodiment extracts an article created within a predetermined period, calculates a feature amount (second TF / IDF value) based on the extracted article and a feature keyword, Since keywords whose calculated feature amount is equal to or greater than the threshold are extracted as trendy keywords (trend keywords), information useful to the user can be provided to the user.

  In this embodiment, when a service provider 160f designates a poster (including a user / community), the user extracts a profile keyword and a similar profile associated with the profile ID of the poster. Although outputting to the terminal, it is not limited to this.

  For example, when a predetermined profile keyword or trend keyword is specified, the service provision processing unit 160f extracts the profile ID and similar profile information associated with the profile keyword or trend keyword, and outputs them to the user terminal May be.

  By the way, among the processes described in the present embodiment, all or a part of the processes described as being automatically performed can be manually performed, or the processes described as being performed manually can be performed. All or a part can be automatically performed by a known method. In addition, the processing procedure, control procedure, specific name, and information including various data and parameters shown in the above-described document and drawings can be arbitrarily changed unless otherwise specified.

  Further, each component of the information management server 100 shown in FIG. 3 is functionally conceptual and does not necessarily need to be physically configured as illustrated. In other words, the specific form of distribution / integration of each device is not limited to that shown in the figure, and all or a part thereof may be functionally or physically distributed or arbitrarily distributed in arbitrary units according to various loads or usage conditions. Can be integrated and configured. Furthermore, each processing function performed by each device may be realized by a CPU and a program that is analyzed and executed by the CPU, or may be realized as hardware by wired logic.

  FIG. 27 is a diagram illustrating a hardware configuration of a computer constituting the information management server according to the embodiment. As shown in FIG. 27, the computer (information management server) 200 includes an input device 201, a monitor 202, a RAM (Random Access Memory) 203, a ROM (Read Only Memory) 204, and a medium reading device 205 that reads data from a storage medium. An interface 206 that transmits and receives data to and from other devices (user terminals), a CPU (Central Processing Unit) 207, and an HDD (Hard Disk Drive) 208 are connected by a bus 209.

  The HDD 208 stores an information management program 208b that exhibits the same function as that of the information management server 100 described above. When the CPU 207 reads and executes the information management program 208b, the information management process 207a is activated. The information management process 207a includes an information management unit 160a, a feature keyword extraction processing unit 160b, a profile keyword extraction processing unit 160c, a similarity calculation processing unit 160d, a trend keyword extraction processing unit 160e, and a service provision processing unit shown in FIG. Corresponds to 160f.

  The HDD 208 stores various data 208a corresponding to the tables 150a to 150l shown in FIG. The CPU 207 reads out various data 208a stored in the HDD 208, stores it in the RAM 203, and extracts feature keywords, profile keywords, similar profiles, and trend keywords using the various data 203a stored in the RAM 203.

  By the way, the information management program 208b shown in FIG. 27 is not necessarily stored in the HDD 208 from the beginning. For example, a “portable physical medium” such as a flexible disk (FD), a CD-ROM, a DVD disk, a magneto-optical disk, or an IC card inserted into a computer, or a hard disk drive (HDD) provided inside or outside the computer. The information management program 208b is stored in the "fixed physical medium" of the computer, and also in the "other computer (or server)" connected to the computer via the public line, the Internet, LAN, WAN, etc. The computer may read out and execute the information management program 208b from these.

(Supplementary note 1)
A feature keyword extraction procedure for acquiring an article created within a predetermined period from a storage device and extracting a keyword included in the acquired article as a feature keyword;
A feature amount calculation procedure for calculating a feature amount of each feature keyword for the user or the community based on an article created by a user or a community composed of a plurality of users and the feature keyword;
A profile keyword extraction procedure for extracting a feature keyword whose feature quantity is equal to or greater than a threshold value as a profile keyword indicating a feature of the user or community;
An information extraction program characterized by causing

(Additional remark 2) The said feature-value calculation procedure extracts the article produced by the said user or community from the said memory | storage device, and counts the 1st value which shows the number of the said feature keywords contained in the extracted article. A counting procedure, a second counting procedure for counting a second value indicating the total number of articles created by the user or community, and the feature keyword included in the articles created by the user or community Supplementary note 1 characterized in that a third counting procedure for counting a third value indicating the number of articles and a calculation procedure for calculating a feature value based on the first, second, and third values are executed. The information extraction program described.

(Supplementary Note 3) Each profile keyword indicating characteristics of a user or community serving as a reference, a reference profile keyword group in which the number of appearance of the profile keyword in an article is associated, and each characteristic indicating characteristics of another user or community Similarity is calculated based on the profile keyword and other profile keyword group including the number of occurrence of the profile keyword in the article, and similar to the reference user or community based on the similarity The information extraction program according to appendix 1 or 2, further causing a computer to execute a similarity calculation procedure for extracting a user or community.

(Additional remark 4) The 4th count procedure which extracts the article | item in the predetermined period from the said memory | storage device, counts the 4th value which shows the number of the said characteristic keywords contained in the extracted article, and the article | item in a predetermined period A fifth counting procedure for counting a fifth value indicating the total number of articles, a sixth counting procedure for counting a sixth value indicating the number of articles including the feature keyword in articles within a predetermined period, and Trend keyword extraction procedure for calculating a second feature amount of each feature keyword based on the fourth, fifth, and sixth values, and extracting a feature keyword having the calculated second feature amount equal to or greater than a threshold as a trendy keyword The information extraction program according to appendix 1, 2, or 3, wherein the computer is further executed.

(Additional remark 5) The said similarity calculation procedure performs the vector operation which used the said reference profile keyword group as the 1st vector, said other profile keyword group as the 2nd vector, and performed between the said 1st and 2nd vector The information extraction program according to supplementary note 3, wherein a distance is calculated as the similarity.

(Additional remark 6) The said memory | storage device further memorize | stores the unnecessary keyword which becomes the object of the said characteristic keyword, The said characteristic keyword extraction procedure removes the characteristic keyword which corresponds to the said unnecessary keyword among the extracted characteristic keywords. The information extraction program according to appendix 1, which is characterized.

(Appendix 7) An information extraction device that manages an article created by a user or a community composed of a plurality of users and extracts predetermined information from the article,
Article storage means for storing articles created by the user or community;
Feature keyword extraction means for acquiring articles created within a predetermined period from the article storage means, and extracting keywords included in the acquired articles as feature keywords;
Feature quantity calculating means for calculating the feature quantity of each feature keyword for the user or community based on the article created by the user or community and the feature keyword;
Profile keyword extraction means for extracting a feature keyword whose feature quantity is equal to or greater than a threshold value as a profile keyword indicating a feature of the user or community;
An information extraction apparatus comprising:

  As described above, the information extraction program and the information extraction apparatus according to the present invention are useful for a system that shares various types of information, and in particular, it is necessary to provide useful information to a user from a vast amount of information. Suitable for cases.

It is a block diagram which shows the structure of the information sharing system concerning a present Example. It is a figure which shows an example of the screen image displayed on the monitor of a user terminal. It is a functional block diagram which shows the structure of the information management server concerning a present Example. It is a figure which shows an example of the data structure of an article management table. It is a figure which shows an example of the data structure of an article information table. It is a figure which shows an example of the data structure of an execution management table. It is a figure which shows an example of the data structure of a synonym and a combined word table. It is a figure which shows an example of the data structure of a keyword table. It is a figure which shows an example of the data structure of an article keyword table. It is a figure which shows an example of the data structure of an unnecessary word table. It is a figure which shows an example of the data structure of an article space management table. It is a figure which shows an example of the data structure of a profile information table. It is a figure which shows an example of the data structure of a profile keyword table. It is a figure which shows an example of the data structure of a trend keyword table. It is a figure which shows an example of the data structure of a similar profile table. It is a figure (1) which shows the specific process of a characteristic keyword extraction process part. It is FIG. (2) which shows the specific process of a characteristic keyword extraction process part. It is a figure (1) which shows the specific process of a profile keyword extraction process part. It is FIG. (2) which shows the specific process of a profile keyword extraction process part. It is a figure which shows the specific process of a similarity calculation process part. It is a figure which shows the specific process of a trend keyword extraction process part. It is a flowchart which shows the process sequence of the information management server concerning a present Example. It is a flowchart which shows the process sequence of characteristic keyword extraction processing. It is a flowchart which shows the process sequence of profile keyword extraction processing. It is a flowchart which shows the process sequence of a similarity calculation process. It is a flowchart which shows the process sequence of a trend keyword extraction process. It is a figure which shows the hardware constitutions of the computer which comprises the information management server concerning an Example.

Explanation of symbols

10, 20, 30 User terminal 50 Network 100 Information management server 110 Input unit 120 Output unit 130 Communication control IF unit 140 Input / output control IF unit 150 Storage unit 150a Article management table 150b Article information table 150c Execution management table 150d Combined word table 150e Keyword table 150f Article keyword table 150g Unnecessary word table 150h Article space management table 150i Profile information table 150j Profile keyword table 150k Trend keyword table 150l Pseudo profile table 160 Control unit 160a Information management unit 160b Feature keyword extraction processing unit 160c Profile Keyword extraction processing unit 160d Similarity calculation processing unit 160e Trend keyword extraction processing unit 1 0f the service providing unit 200 the computer 201 input device 202 monitors 203 RAM
203a, 208a Various data 204 ROM
205 Media reader 206 Interface 207 CPU
207a Information management process 208 HDD
208b Information management program 209 bus

Claims (5)

On the computer,
A feature keyword extraction procedure for acquiring an article created within a predetermined period from a storage device and extracting a keyword included in the acquired article as a feature keyword;
A feature amount calculation procedure for calculating a feature amount of each feature keyword for the user or the community based on an article created by a user or a community composed of a plurality of users and the feature keyword;
A profile keyword extraction procedure for extracting a feature keyword whose feature quantity is equal to or greater than a threshold value as a profile keyword indicating a feature of the user or community;
An information extraction program characterized by causing   The feature amount calculation procedure includes a first counting procedure for extracting an article created by the user or community from the storage device and counting a first value indicating the number of the feature keywords included in the extracted article; A second counting procedure for counting a second value indicating the total number of articles created by the user or community, and the number of articles including the characteristic keyword among the articles created by the user or community. The information according to claim 1, wherein a third counting procedure for counting a third value to be indicated and a calculation procedure for calculating a feature amount based on the first, second, and third values are executed. Extraction program.   Each profile keyword indicating the characteristics of the reference user or community and the number of occurrences of the profile keyword in the article, each profile keyword indicating the characteristics of other users or communities, and the corresponding Similarity is calculated based on other profile keyword groups including the number of profile keywords appearing in the article, and other users similar to the reference user or community based on the similarity or The information extraction program according to claim 1 or 2, further causing a computer to execute a similarity calculation procedure for extracting a community.   A fourth counting procedure for extracting articles from the storage device within a predetermined period and counting a fourth value indicating the number of the feature keywords included in the extracted articles, and a total number of articles within the predetermined period A fifth counting procedure for counting a fifth value, a sixth counting procedure for counting a sixth value indicating the number of articles including the characteristic keyword in articles within a predetermined period, and the fourth, fifth, And a trend keyword extraction procedure for calculating a second feature value of each feature keyword based on the value of 6, and extracting a feature keyword having the calculated second feature value equal to or greater than a threshold value as a trendy keyword. The information extraction program according to claim 1, 2, or 3, wherein the information extraction program is executed. An information extraction device that manages an article created by a user or a community composed of a plurality of users and extracts predetermined information from the article,
Article storage means for storing articles created by the user or community;
Feature keyword extraction means for acquiring articles created within a predetermined period from the article storage means, and extracting keywords included in the acquired articles as feature keywords;
Feature quantity calculating means for calculating a feature quantity of each feature keyword for the user or community based on an article created by the user or community and the feature keyword;
Profile keyword extraction means for extracting a feature keyword whose feature quantity is equal to or greater than a threshold value as a profile keyword indicating a feature of the user or community;
An information extraction apparatus comprising:

Images