JP2010108519A - Communication system, program and method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To present more useful information to a user by acquiring articles corresponding to roles of the user in a community without performing a special operation by the user. <P>SOLUTION: A KL information DB 9 stores a KL record showing relation and a type of relation between articles stored in an article DB 8. A user role derivation part 11 derives roles of each user in the community on the basis of the KL record. A user role information DB 12 stores information showing the roles of the user derived by the user role derivation part 11 by associating the information with the user. An information retrieval part 14 acquires, according to a request of retrieval of the articles or reference to the articles from the user, articles matched to the request from the article DB 8. The information retrieval part 14 determines priority when the acquired articles are presented to the user at a request source on the basis of the roles shown by the information stored in the user role information DB 12 by associating the priority with a degree of matching to the request and the user at the request source. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a communication system and program capable of electronically communicating with a plurality of users.

  As a system in which a plurality of users communicate electronically, a communication system using a mailing list or an electronic bulletin board is used. In such a communication system, in response to a request from a user, an article that the user wants to know or an article that seems important to the user is generally searched from articles (article data) in the community. It has been broken.

  As a conventional article search method, a method of searching for an article from a keyword added at the time of registering the article or a method of searching using a natural language is known. There is also a method for searching for articles based on the number of utterances in the article.

  However, in the above-described method of searching for articles from keywords, the search target article is uniquely determined by the keyword specified by the user. There are cases where it is not possible to reliably search for a document that is truly required simply by including a specified keyword in an article. Therefore, there arises a problem that a highly reliable document search function cannot be realized. Further, when there are a large number of corresponding articles depending on the content, number of keywords, etc., a desired article must be extracted from among them by artificial visual inspection. Therefore, there is a problem that a search operation for a desired article requires a lot of time and labor, and an efficient search operation cannot be performed. The same problem occurs in the method of searching using the natural language described above. Even when a search is performed based on the number of utterances of an article, whether an article presented as a search result is important or not depends on the person viewing the article. For this reason, the reliability of the search result is not necessarily guaranteed, and the user's request may not be met.

  Thus, for example, Patent Document 1 discloses the following technique (hereinafter referred to as prior art) for narrowing down search targets from a large number of electronic files. According to this prior art, when searching for an electronic file containing desired contents, a search target is narrowed down based on the history of operations on the electronic file, and a document file containing a keyword specified by the user is searched. Can be executed.

JP 2000-163439 A

  As described above, according to the prior art, it is possible to execute the search process after narrowing down the search target by using the history of operations on the electronic file.

  However, in the prior art, when searching for a document file including desired contents, in order to narrow down the search target, the user needs to input a history of operations on the electronic file himself as one of the search conditions. . Also, articles that the user determines to be important differ depending on the role of the user in the community. However, the prior art does not consider the role of users in the community. For this reason, even if the search process is executed after narrowing down the search target based on the history of operations on the electronic file, it is not always possible to present a search result that is useful to the search requesting user.

  The present invention has been made in consideration of the above circumstances, and its purpose is to derive what role a user has in a given community without any special operation by the user. It is to provide a communication system and program capable of presenting more useful information.

  According to one aspect of the present invention, there is provided a communication system having article storage means for storing articles created by users in a community and including articles related to the already created articles. The communication system includes a knowledge link information storage unit that stores, for each article related to each other among the articles stored in the article storage unit, knowledge link information indicating a relationship between the articles and a type of the relationship, Based on the type of relationship indicated by the knowledge link information, the user role deriving means for deriving the role of the user in the community who created the articles related to each other indicated by the knowledge link information, and derived by the user role deriving means User role storage means for storing information indicating the user's role in association with the user, and in response to an article search or article reference request from the user, based on the degree of matching the request with the request Acquired from the article storage means and acquired by the article acquisition means The degree of matching with the request for each article is acquired by the article acquisition unit, a changing unit that changes the degree of matching with the requesting user based on the role indicated by the information stored in the user role storage unit Priority determining means for determining the priority for each article when the article is presented to the requesting user based on the degree of matching with the request for each article changed by the changing means.

  According to the present invention, by deriving what role a user has in a predetermined community and acquiring articles according to the role of the user, more useful information is presented to the user. Is possible.

1 is a schematic configuration diagram of a communication system 3 and user terminals 2-1, 2-2,..., 2-n according to an embodiment of the present invention. The block diagram which mainly shows a function structure of the communication system 3 of FIG. The figure which shows an example of the article contribution screen displayed on the user terminal 2-i (i = 1, 2, ..., n) of FIG. The block diagram which shows the structure of the user role derivation | leading-out part 11 of FIG. The block diagram which shows the structure of the information search part 14 of FIG. The figure which shows the example of a data structure of article DB8 of FIG. The figure which shows the example of a data structure of KL information DB9 of FIG. The flowchart which shows the process sequence of the user role derivation | leading-out part 11. FIG. The figure which shows an example of the combination information of pattern information. The figure which shows an example of the weight information of pattern information. The figure which shows an example of the result of having totaled the weight with respect to a role as a point for every user. The flowchart which shows the process sequence of the pattern collation part 11c. The figure which shows an example of the conditions corresponding to the hierarchy k. FIG. 14 is a diagram for explaining an article that meets the conditions of FIG. 13. The figure which shows an example of the conditions corresponding to the hierarchy k + 1. FIG. 16 is a diagram for explaining an article that meets the conditions of FIG. 15. The flowchart which shows the process sequence of the information search part 14 with respect to an article search request. The flowchart which shows the process sequence of the information search part 14 with respect to an article reference request.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
First, with reference to FIG.1 and FIG.2, the structure of the communication system which concerns on one Embodiment of this invention is demonstrated.

  FIG. 1 is a diagram illustrating a relationship between a communication system and a user terminal according to an embodiment of the present invention. As shown in FIG. 1, user terminals 2-1, 2-2,..., 2-n operated by a user can communicate with a communication system 3 via a network 1 such as the Internet or a local area network. So connected. The communication system 3 provides a community such as an electronic bulletin board to the users of the user terminals 2-1 to 2-n.

  The communication system 3 includes a computer 31 and an external storage device 32 such as a hard disk drive. The external storage device 32 stores a program 320 necessary for realizing the communication system 3.

Next, mainly the functional configuration of the communication system 3 will be described with reference to FIG.
As shown in FIG. 2, the communication system 3 includes an input interface 4, an information registration management unit 5, an article storage processing unit 6, a knowledge link storage processing unit 7, an article database 8, a knowledge link information database 9, and a role pattern dictionary 10. , A user role derivation unit 11, a user role information database 12, a rule database 13, an information search unit 14, a result generation unit 15, and an output interface 16. The article database 8, knowledge link information database 9, role pattern dictionary 10, user role information database 12, and rule database 13 are stored in the external storage device 32 shown in FIG.

  The input interface 4 receives, for example, an article search or article reference request or article data (hereinafter referred to as “article search”) sent from the user terminal 2-i (i = 1, 2,..., N) in response to a user operation. The communication system 3 is input.

  FIG. 3 shows an example of an article posting screen displayed on the user terminal 2-i when the user posts (inputs) an article to the communication system 3 from the user terminal 2-i. FIG. 3A shows an article posting screen 17 in the case of creating an article that replies to a certain article (here, “item information (Mr. XXX)”). When the user presses an add button 17a for adding a related article existing on the article posting screen 17, another screen 18 as shown in FIG. 3B is opened. On this screen 18, related articles, for example, recently viewed articles and bookmark articles are displayed. For example, recently viewed articles are displayed as a list in the column 18b indicating recently viewed articles. Bookmark articles are displayed as a list in the column 18c indicating articles registered in the bookmark. Each article in the list displayed in the columns 18b and 18c of the screen 18 has a check box. Here, it is assumed that, for example, the user has checked a check box of an article to be referred to when creating a reply article, and has pressed the add button 18a. Then, the checked article is added to the related article column 17b of the article posting screen 17 as a related article. The article posting screen 17 includes an attachment field 17c. By using this field 17c, an attached file can be designated. The user operates the user terminal 2-i to create a title and text on the article posting screen 17 based on the related article or the attached file, and posts the article to the communication system 3.

  Referring to FIG. 2 again, the information registration management unit 5 accepts an article input by the input interface 4. The information registration management unit 5 outputs the accepted article to the article storage processing unit 6. The information registration management unit 5 instructs the knowledge link storage processing unit (KL storage processing unit) 7 to acquire knowledge link information (hereinafter referred to as KL information) described later. In addition, the information registration management unit 5 notifies the user role deriving unit 11 that a new article has been added and that an instruction for acquiring KL information has been issued.

  The article storage processing unit 6 performs processing for storing the article output by the information registration management unit 5 in the article database (DB) 8. This article includes information indicating the user who created the article.

  The KL storage processing unit 7 acquires KL information in response to an instruction from the information registration management unit 5. The KL information is a mutual article between an article (first article) newly created by the user and stored in the article DB 8 and an article (second article) already stored in the article DB 6 related to this article. Represents the relationship between and the type of relationship. The KL information includes information indicating a direction for identifying the first article as the previous article and the second article as the original article. The KL storage processing unit 7 determines the type of relationship between the articles based on the type of user operation for associating the previous article with the original article when the user creates the previous article.

  Here, the type of user operation for associating the previous article with the original article is to add an article to the related article column 17b when the user creates an article, as described in the explanation of FIG. 3 described above. For example, an operation or an operation for specifying an attached file in the attachment column 17c. The KL storage processing unit 7 obtains KL information (KL record) indicating the type of relationship between the determined articles, and stores the KL information in a knowledge link information database (KL information DB) 9.

  In the role pattern dictionary 10, pattern information having combination information and weight information is stored in advance. The combination information is a feature of a combination of types of relations between the plurality of articles included in a standard path indicated by a connection of a plurality of articles created in a chain by a user in the community. Indicates. The weight information corresponds to the user who created an article corresponding to each of the connection depth (hereinafter referred to as a hierarchy) of articles from an article located at the end of a typical path in the plurality of articles. Role and weight.

  The user role deriving unit 11 refers to the article DB 8, the KL information DB 9, and the role pattern dictionary 10, and derives the role of the user in the community who created the articles related to each other by the KL information. The user role deriving unit 11 stores information indicating the derived user role in the user role information database (DB) 12 in association with the user. The roles derived by the user role deriving unit 11 include, for example, “actual charge”, “combining”, and “management” described later.

  The rule database (DB) 13 is associated with each role indicated by the information stored in the user role information DB 12, and the degree of matching with an article search or article reference request from the user (hereinafter referred to as the degree of matching). ) (Rule information) for changing the) based on each of the roles. Note that the rules stored in the rule DB 13 include a rule that is applied when an article search is requested by the user and a rule that is applied when an article reference is requested by the user.

  In response to an article search or article reference request from the user, the information search unit 14 acquires an article that matches the request from the article DB 8 based on the degree of match. Here, an article having a matching degree higher than a certain value is acquired as an article that matches the request. In addition, a certain number of articles with the highest degree of match may be acquired as articles that match the request. For the degree of match, a score is used for an article search request, and a weight is used for an article reference request. The information search unit 14 refers to the rule DB 13 and changes the degree of coincidence of the acquired article according to the rule corresponding to the role of the user who requested the article search request or the article reference request.

  The information search unit 14 sets the priority (presentation priority) for each article when presenting the acquired article to the requesting user based on the matching degree changed based on the rules stored in the rule DB 13. To decide. The information search unit 14 outputs information acquired from the article DB 8 and information indicating the priority of each article to the result generation unit 15.

  The result generation unit 15 generates a result screen corresponding to the role indicated by the information stored in the user role information DB 12 in association with the requesting user based on the article and priority output by the information search unit 14. To do. The result generation unit 15 outputs information on the generated result screen to the output interface 16.

  The result screen generated by the result generation unit 15 displays a list of articles to be presented to the user as a result of an article search or article reference request from the requesting user. This list of articles includes summary sentences of the articles. The result generation unit 15 changes the number of articles displayed on the result screen and the length of the summary sentence of the article according to the role of the requesting user.

  For example, when the role of the requesting user is “actual charge”, the result generation unit 15 reduces the number of articles in the list displayed on the result screen and lengthens the summary sentence of the article. In addition, as the role of the requesting user becomes “summary” or “manager”, the result generation unit 15 increases the number of articles in the list and shortens the summary sentence of the article.

  The length of the summary sentence of the articles in the list displayed on the result screen even when the role of the user is not considered as in the prior art in determining the priority when presenting the acquired article to the requesting user. The size may be changed according to the role of the user.

  The result generation unit 15 includes a weighting unit 151. The weighting unit 151 divides the content of the article acquired by the information search unit 14 into, for example, a paragraph and a sentence in the paragraph with a line feed and a punctuation mark “.”, And performs weighting as follows. For example, the weighting unit 151 assigns a weight (first weight) to a sentence included in each paragraph in the article acquired by the information search unit 14 for a sentence included in the preceding paragraph in the article. Make it high. Also, the weighting unit 151 increases the weight (second weight) of the first and last sentences in the paragraph among the sentences included in each paragraph in the article. Further, the weighting unit 151 calculates the weight of the sentence (third weight) by adding the weight of a word included in the sentence in the article, for example, the weight of TF / IDF for each sentence. The TF / IDF value of a word is well known as an evaluation value that represents the degree to which the word appears. The weighting unit 151 determines the weight for each sentence based on the first, second, and third weights acquired for each sentence included in the article. Here, weighting for each sentence is performed by adding the first, second, and third weights for each sentence.

Based on the weighting result by the weighting unit 151, the result generation unit 15 creates a summary sentence by combining, for example, two or three sentences with high weights according to the role of the requesting user.
Note that the TF / IDF weight calculation need not necessarily be performed for all articles acquired by the information search unit 14. For example, when the role of the requesting user is “actual charge”, the weight of TF / IDF may be calculated only for articles created by the user whose role is “actual charge”. The same applies to the case where the role of the requesting user is “summary” or “administrator”. In this case, since frequently used words are different for each role, for example, by limiting the articles for which weights are calculated by TF / IDF to articles created by a user whose role is “actually responsible” , The weight of a frequently used word used by a user whose role is “actually in charge” is larger than when the weight is calculated for all articles in the article DB 8. As a result, a sentence including words that are frequently used by a user whose role is “actually in charge” is likely to be included in the summary sentence, so that a summary sentence more suitable for each role is created.

  The output interface 16 outputs the result (information on the result screen) for the article search or article reference request from the request source user output by the result generation unit 15 to the user of the request source. -i (i = 1, 2,..., n)

  The input interface 4, the information registration management unit 5, the article storage processing unit 6, the knowledge link storage processing unit 7, the user role derivation unit 11, the information search unit 14, the result generation unit 15 and the output interface 16 are the same as those shown in FIG. It is assumed that 31 is realized by reading and executing a program 320 stored in the external storage device 40. This program 320 can be stored in advance in a computer-readable storage medium and distributed. Further, the program 320 may be downloaded to the computer 31 via the network 1.

  FIG. 4 is a block diagram illustrating a configuration of the user role deriving unit 11. The user role derivation unit 11 includes an article list creation unit 11a, a pattern information list creation unit 11b, a pattern verification unit 11c, an addition unit 11d, and a role selection unit 11e.

  The article list creation unit 11a refers to the article DB 8, acquires an article that has not been processed by the user role deriving unit 11 and is indicated by a calculated flag described later, and creates a list of the article.

  The pattern information list creation unit 11b acquires all the pattern information stored in the role pattern dictionary 10 and creates a list of the pattern information.

  When the article in the list created by the article list creating unit 11a is an article located at the end of the regular route (hereinafter referred to as a starting point), the pattern matching unit 11c uses the pattern information list creating unit 11b. It is determined whether or not the feature is indicated by the combination information of the pattern information of the created list. The pattern matching unit 11c refers to the KL information DB for a combination of links between a plurality of articles in which articles are chained starting from the article i and a feature indicated by the combination information of the pattern information j. To perform the determination process.

  Based on the weight information of the pattern information, the adding unit 11d creates an article that is located in each hierarchy from the starting point in the fixed path among a plurality of articles that match the characteristics indicated by the combination information. Add weight each time.

  The role selection unit 11e selects the role of each user based on the weight for each user added by the addition unit 11c. The role selection unit 11d stores information indicating the role of the user in the user role information DB 12 in association with each user.

  FIG. 5 is a block diagram illustrating a configuration of the information search unit 14. The information search unit 14 includes a user identification unit 21, an article acquisition unit 22, a KL information acquisition unit 23, a creator recognition unit 24, a role acquisition unit 25, an article score / weight change unit 26, and a sort unit 27.

  The user identifying unit 21 identifies the requesting user from the user information added to the article search or article reference request input by the input interface 4.

  The article acquisition unit 22 includes an article search unit 221 and an article list unit 222. In response to an article search request from the user, the article search unit 221 acquires, from the article DB 8, an article having a higher score representing the degree of matching with the request as an article that matches the request. Here, the article search request is a request for searching for an article including a certain keyword, for example. The article search unit 221 acquires, for each acquired article, a score representing the degree of matching of the searched article with respect to the user's article search request.

  In response to an article reference request from the user, the article list unit 222 acquires, from the article DB 8, an article having a higher weight indicating the degree of matching with the request as an article that matches the request. Here, the article reference request is, for example, a request for acquiring a new article in the community. For each acquired article, the article list unit 222 acquires a weight that represents, for example, the novelty of the article as a degree of match.

  The KL information acquisition unit 23 acquires the KL information when the KL information DB 9 stores KL information in which the article acquired by the article search unit 221 or the article list unit 222 is a previous article or an original article.

  The creator recognition unit 24 recognizes the user who created the article from the information indicating the user included in the article acquired by the article search unit 221 or the article list unit 222. Further, the creator recognition unit 24 recognizes the user who created the previous article or the original article from the information indicating the user included in the previous article or the original article indicated by the KL information acquired by the KL information acquisition unit 23. .

  The role acquisition unit 25 acquires the role indicated by the information stored in the user role information DB 12 in association with the user who requested the article search or article reference and the user recognized by the creator recognition unit 24.

  The article score / weight change unit 26 includes a score acquisition unit 261, a weight calculation unit 262, and an addition unit 263. The score acquisition unit 261 acquires the score of the previous article or the original article indicated by the KL information acquired by the KL information acquisition unit 23.

  The weight calculator 262 determines the weight of each article acquired by the article search unit 221 or the article list unit 222 according to the rules stored in the rule DB 13. That is, the weight calculation unit 262 determines the relationship between the articles represented by the KL information acquired by the KL information acquisition unit 23 and the type of the relationship, the article search or article reference request source acquired by the role acquisition unit 25. The weight for each article is determined based on the role of the user recognized by the user and the creator recognition unit 24. The weight calculation unit 262 weights the score for each article acquired by the score acquisition unit 261 based on the determined weight for each article. Here, the score for each acquired article is multiplied by the weight of the article. The weight calculator 262 outputs the weighted score for each article (at the time of article search) or the determined weight for each article (at the time of article reference) to the adder 263.

  For each article acquired by the article search unit 221, the adding unit 263 adds the score output by the weight calculation unit 263 to the score of the article acquired by the article search unit 221. For each article acquired by the article list unit 222, the adding unit 263 adds the weight output by the weight calculation unit 262 to the weight of the article acquired by the article list unit 222. The addition unit 263 uses the addition result (score or weight for each article) as information for determining the priority when the article acquired in response to the article search or article reference request is presented to the user. Output.

  The sorting unit 27 sorts the articles to be presented to the user as a result of the article search or article reference request based on the score or weight for each article output by the adding unit 263. That is, the sorting unit 27 determines the priority (priority indicating the order of presentation) of articles in the list when presenting a list of articles to the user based on the score (weighted score) or weight of the article. Functions as a priority determination unit. The sort unit 27 outputs information indicating the sorted result (priority is determined) to the result generation unit 15.

  FIG. 6 shows an example of the data structure of the article DB 8. In the article DB 8, as shown in FIG. 6, for each article, an article ID that is an identifier for uniquely identifying each article, an article title, an article body, a user who created the article (information indicating the), and a calculation A record (article record) including a completed flag is stored. Although omitted in FIG. 6, the article record includes information on the creation date (post date) of the corresponding article. The calculated flag indicates whether the corresponding article is an article processed by the user role deriving unit 11. In FIG. 6, it is assumed that the article record includes the text of the article for the sake of simplicity. However, since the body of the article has a variable length, the body is separated from the article record and stored in the article DB 8, and a pointer indicating the position of the body is set in the article record.

  In the example illustrated in FIG. 6, the first article record includes “article ID: 1”, “title: xxx”, “text: xxx”, “user: A”, and “calculated flag: F1”. Is remembered.

  In the calculated flag, “F1” or “F2” is stored. “F1” indicates that the corresponding article is an article processed by the user role deriving unit 11, and “F2” indicates that the article is not an article processed by the user role deriving unit 11.

  FIG. 7 shows an example of the data structure of the KL information DB 9. In the KL information DB 9, a KL record that represents the relationship between the articles and the type of the relationship is stored for each of the articles related to each other among the articles indicated by the article record stored in the article DB 8. As shown in FIG. 7, the KL record indicates the original article ID indicating the original article, the previous article ID indicating the previous article, and the type of relationship between the original article and the previous article. This is information (KL information) including a link (link type) to be expressed.

  In the example illustrated in FIG. 7, “original article ID: 1”, “previous article ID: 2”, and “link: follow” are stored in the first KL record.

  As shown in FIG. 7, the links include, for example, “follow”, “sum”, and “create”. “Follow” means adding a contextual information such as supplementing, replying or refuting to an article. “Sum” means a relationship for summarizing or summarizing an article. “Create” means that it was created regardless of other articles.

  In addition to the above, for example, there are “consult”, “revise”, and “describe”. “Consult” means a relationship that refers to an article. “Revise” means a relationship in which an article is revised and replaced. “Describe” has a descriptive relationship and means that the previous article and the original article have a strong connection that cannot be understood in context unless they coexist.

Next, the processing procedure of the user role deriving unit 11 will be described with reference to the flowchart of FIG.
First, the article list creation unit 11a refers to the article record stored in the article DB 8, and finds all articles indicated by the article record having a calculated flag “F2” (no calculated flag is set). Acquire and create a list of the articles (step S1). Here, it is assumed that the article list creation unit 11a creates a list of n articles, for example.

  Next, the pattern information list creation unit 11b acquires all the pattern information stored in the role pattern dictionary 10 and creates a list of the pattern information (step S2). Here, it is assumed that the pattern information list creation unit 11b creates a list of m pieces of pattern information, for example.

  In the user role deriving unit 11, the processes from Step S4 to Step S8 described later are repeated for each of the n articles based on the list created by the article list creating unit 11a (Step S3). Here, an article i (i = 1, 2,..., N) among n articles in the list created by the article list creating unit 11a will be described.

  The user role deriving unit 11 repeats the processing from step S5 to step S7 described later for each of the m pieces of pattern information based on the list created by the pattern information list creating unit 11b (step S4). Here, the pattern information j (j = 1, 2,..., M) among the m pieces of pattern information in the list created by the pattern information list creation unit 11b will be described.

  The pattern collating unit 11c has a plurality of articles that are created in a chain manner starting from the article i in order to determine whether or not it matches the feature indicated by the combination information of the pattern information j when the article i is the starting point. The link combination between the articles and the feature indicated by the combination information of the pattern information j are collated based on the KL record stored in the KL information DB 9 (step S5).

  FIG. 9 shows an example of combination information constituting pattern information in the role pattern dictionary 10. The combination information includes, for each layer from the article as the starting point, a number for identifying the layer, a condition for matching the combination indicated by the combination information, and information indicating the end of the combination indicated by the combination information.

  The conditions include, for example, information indicating the direction of the link, the type of link, the number of repetitions, and the creator of the article. In the direction of the link, the first direction (IN) indicating that the link is connected from the article as the starting point to another article, and the link is connected from the other article toward the article as the starting point. There is a second orientation (OUT) to indicate that. Since the type of link has already been described in detail, the description is omitted here. The number of repetitions indicates the number of times the link indicated by the link type is repeated. The creator of the article indicates a user who created an article connected to the article that is the starting point by the link indicated in the type of link.

  The end is a flag “0” or “1” indicating whether the end of the combination information of the pattern information. Here, the flag “0” indicates that the hierarchy corresponding to the flag is not the end of the combination information, and the flag “1” indicates that the hierarchy corresponding to the flag is the end of the combination information.

  In the example shown in FIG. 9, the pattern information corresponds to each of the three levels “1”, “2”, and “3”, and includes a number, a link direction, a link type, a repetition count, an article creator, Contains end information. Specifically, corresponding to “hierarchy: 1”, “number: 1-1”, “link direction: IN”, “link type: follow”, “repetition count: 1”, “article creator” : = Α ”and“ Termination: 0 ”are stored. Corresponding to “hierarchy: 2”, “number: 2-1”, “link direction: IN”, “link type: follow”, “repetition count: ≧ 0”, “article creator: “≠ α (not α)” and “Termination: 0” are stored. Corresponding to “hierarchy: 3”, “number: 3−1”, “link direction: IN”, “link type: sum”, “number of repetitions: 1”, “article creator: ≠ “α” and “Termination: 1” are stored.

  In the role pattern dictionary 10, pattern information in which links such as “create”, “consult”, “revise”, and “describe” described above are combined is also stored in advance.

  Now, as a result of the collation by the pattern collating unit 11c executed in step S5 in FIG. 8, a combination of links between a plurality of articles in which articles are chained starting from the article i and combination information of the pattern information j Is matched (matched) with the feature indicated by (YES in step S6). In this case, based on the weight information of the pattern information j, the adding unit 11d determines the role weight defined in the weight information of the pattern information j for each creator (user) of an article that matches the pattern information j. Are added (step S7).

  On the other hand, when the article i does not match the pattern information j starting from the article i (NO in step S6), the adding unit 11d skips step S7.

  FIG. 10 shows an example of weight information constituting the pattern information described above in the role pattern dictionary 10. The weight information includes, for each user to whom the weight is added, information on an addition destination indicating the user, a role type of the user, and an added value indicating the weight to be added to the user.

  In the example illustrated in FIG. 10, the weight information includes a role type and an addition value corresponding to each of the three addition destinations. Specifically, corresponding to “addition destination: α”, “number of links corresponding to“ type of role: summary ”and“ addition value: (number indicating combination information layer “3”) 3-1 × 2 ” Is stored. Corresponding to “addition destination: 3-1 article creator” and “addition destination: 2-1 article creator”, “type of role: actual charge” and “addition value: number of links” Regardless of 1 "is stored.

  Here, it is assumed that the processing from step S5 to step S7 in FIG. 8 is repeatedly executed for all m pieces of pattern information based on the list created by the pattern information list creation unit 11b. In this case, the loop for the pattern information ends (step S8).

  Similarly, it is assumed that the processes from step S4 to step S8 are repeatedly executed for all n articles based on the list created by the article list creation unit 11a. In this case, the loop for the article ends (step S9).

  For each user, the role selection unit 11e aggregates the weights for the roles as points based on the result added by the addition unit 11d, and selects a role having a higher point as the role of the user (step S10). The role selection unit 11e stores the selected role in the user role information DB 12 in association with each user.

  FIG. 11 shows an example of the result of tabulating the weights for roles in step S10 in a table format. In FIG. 11, for each user in the community, for each role of “actual charge”, “combining”, and “management”, points of the user with respect to the role are shown. For example, for user A, the “actually responsible” point is “3”, the “summary” point is “10”, and the management point is “0”. In this case, the role selection unit 11d selects “Summary” as the role of the user A.

  Similarly, for other users, the role selection unit 11d selects “actual charge” as the roles of the users C and D, and selects “management” as the role of the user X. Further, as in the example of the user B, when the role with a high point is not one, or when there is no role with a high point, other processing is performed without considering the role of the user B.

  As described above, the role of the user includes “actual charge”, “summarization”, and “management”. “Actual charge” is, for example, the role of the user who created an article that replies to a certain article or the user who created the article that is the basis when the collected article is created. “Summary” is, for example, the role of a user who creates an article that summarizes a certain article or a user who creates an explanatory article for a certain article. “Management” is, for example, the role of the user who created an article that becomes a reference source when an article is created with reference to an article.

  Next, with reference to the flowchart of FIG. 12, the detailed procedure of the collation process (step S5 of FIG. 8) of the pattern collation part 11c is demonstrated. Note that the pattern matching unit 11c performs matching processing on the pattern information j.

  The pattern matching unit 11c repeats the processing from step S12 to step S17 described later for each of the layers indicated in the combination information of the pattern information j (step S11). Here, the hierarchy k (k = 1, 2,..., P) will be described on the assumption that p hierarchies are stored in the combination information of the pattern information.

  The pattern matching unit 11c acquires an article list for all articles in the hierarchy k (step S12). Here, it is assumed that the pattern matching unit 11c has acquired an article list for q articles.

  Next, the pattern matching unit 11c repeats the processing from step S14 to step S16 for each of q articles based on the acquired list (step S13). Here, articles 1 (l = 1, 2,..., Q) in the list acquired by the pattern matching unit 11c will be described.

  The pattern matching unit 11c refers to the KL information DB 9 and checks whether there is an article that matches (matches) the condition corresponding to the hierarchy k indicated by the combination information of the pattern information j, starting from the article l (step) S15).

  When there is an article that meets the condition corresponding to the hierarchy k (YES in step S15), the pattern matching unit 11c adds all articles that meet the condition to the article list of the hierarchy k + 1 (step S16).

  Based on the article list, the pattern matching unit 11c repeatedly executes the processes from step S14 to step S16 for all q articles. Thereby, the loop of the articles in the article list ends (step S17).

  Similarly, the pattern matching unit 11c repeatedly executes the processes from step S12 to step S17 for all of the p layers indicated in the combination information of the pattern information. Thereby, the loop about a hierarchy is complete | finished (step S18).

  As described above, when the processing from step S12 to step S17 is executed for the hierarchy k, if the termination flag corresponding to the hierarchy k is “0”, the same processing for the hierarchy k + 1 is performed next. Executed. If the termination flag corresponding to layer k is “1”, the loop for the layer ends.

  When there is an article in which the termination flag included in the combination information of the pattern information has reached the level of “1”, the pattern matching unit 11c determines that the feature indicated by the combination information is matched, and to that effect Is notified to the adder 11d (step S19).

  Next, a specific example of processing of the pattern matching unit 11c will be described with reference to FIGS. FIG. 13 shows an example of conditions corresponding to the hierarchy k, and FIG. 14 is a diagram for explaining articles that meet the conditions of FIG. FIG. 15 shows an example of conditions corresponding to the hierarchy k + 1, and FIG. 16 is a diagram for explaining articles that meet the conditions of FIG. Here, similarly to the above description, it is assumed that the hierarchy is the hierarchy k and the article that is the starting point is the article l.

  As shown in FIG. 13, the link direction, which is a condition corresponding to the hierarchy k, is “IN”, the link type is “sum”, the number of repetitions is “1”, and the creator of the article “ ≠ α (the creator is not α) ”.

  Here, the article in the article list of the hierarchy k is the article 1 as the starting point. The pattern matching unit 11c checks whether there is an article that meets the conditions shown in FIG.

  As shown in FIG. 14, the direction of the link, which is a condition, is “IN”, so the articles that are linked to the article l are articles 101, 102, and 103. Therefore, the articles 104 and 105 are not articles that satisfy the condition corresponding to the hierarchy k.

  Also, as shown in FIG. 14, since the type of link is “sum”, each of the articles 101, 102, and 103 satisfies this condition.

  At this time, the direction and type of the link between the article l and the articles 101, 102, and 103 are specified based on the KL record stored in the KL information DB 9.

  Since the number of repetitions is “1”, each of the articles 101, 102, and 103 satisfies this condition. Since the creator of the article is “≠ α”, the article 103 whose creator is α does not satisfy this condition.

  Therefore, in FIG. 14, articles that meet the conditions of FIG. 13 are articles 101 and 102.

  Articles 101 and 102 that meet this condition are added to the article list at level k + 1. As illustrated in FIG. 12, the pattern matching unit 11 c executes the process of the hierarchy k + 1 after the hierarchy k in order to execute the process for each of the hierarchies.

  The pattern matching unit 11c acquires the article list of the hierarchy k + 1. At this time, articles 101 and 102 that match the conditions of the hierarchy k exist in the article list. Here, the article 101 in the article list will be described.

  First, the pattern matching unit 11c checks whether there is an article that satisfies the condition corresponding to the hierarchy k + 1, starting from the article 101.

  As shown in FIG. 15, the link direction, which is a condition corresponding to the hierarchy k + 1, is “IN”, the type of link is “follow”, the number of repetitions is “≧ 0”, and the creator of the article “≠ α”.

  As shown in FIG. 16, since the direction of the link that is a condition is “IN”, the article that satisfies this condition is the article 106. Also, since the type of link that is a condition is “follow”, the article 106 also satisfies this condition.

  Here, the number of repetitions as a condition is “≧ 0”. “0” means that the article 101 as the starting point is included. Further, “≧ 0” means “0 or more”, so that if the other conditions are satisfied, it means that the number of repetitions from the article 101 as the starting point is not limited.

  Therefore, in this case, the articles 101, 106, 107, and 108 shown in FIG. 16 are all connected by “follow” in the direction of “IN” from the article 101 as the starting point, and therefore all of these satisfy the condition.

  Since the creator of the article is “≠ α”, the article 108 whose creator is α does not satisfy this condition.

  Therefore, in FIG. 16, articles satisfying the conditions shown in FIG. 15 are articles 101, 106, and 107.

  Articles 101 and 102 that meet this condition are added to the article list of the hierarchy k + 2. When the termination flag indicated corresponding to the hierarchy k + 1 is “0”, the process is repeated as described above until the hierarchy where the termination flag indicates “1”.

  Next, a processing procedure of the information search unit 14 when an article search is requested by the user will be described with reference to a flowchart of FIG.

  It is assumed that the user has performed an operation for designating an article search request using the user terminal 2-1. In this case, the user terminal 2-1 sends an article search request designated by the user to the communication system 3 via the network 1. Information of the user who specified the request, that is, the requesting user is added to the article search request.

  The input interface 4 of the communication system 3 inputs an article search request from the user terminal 2-1. The user identification unit 21 of the information search unit 14 identifies the requesting user from the user information added to the article search request input by the input interface 4 (step S110).

  The article search unit 221 searches for an article that matches the article search request from the article DB 8 by, for example, a method of searching for an article from a keyword, and acquires a score representing the degree of match of the searched article with respect to the request (step S120). . The article search unit 221 acquires, for example, the top r articles in the score order of the acquired articles among the searched articles (step S130).

  The KL information acquisition unit 23 extracts an article h (h = 1, 2,..., R) from the acquired top r articles, and stores all KL records indicating the article h as a previous article or an original article. Obtained from the KL information DB 9 (step S140). Here, it is assumed that the KL information acquisition unit 23 has acquired s KL records.

  The creator recognition unit 24 recognizes the user who created the previous article and the original article from the information indicating the previous article and the user included in the original article indicated by the s KL records acquired in Step S140 (Step S140). S150).

  The role acquisition unit 25 is stored in the user role information DB 12 in association with each of the user who has requested the article search request and the user who created the previous article and the original article indicated by the acquired s KL records. The role indicated by the stored information is acquired (step S160).

  The score acquisition unit 261 extracts one of the KL records acquired in step S150, and the article g (g = 1, 2,...) That is the previous article or the original article of the article h indicated by the extracted KL record. The score of r) is acquired (step S170).

  The weight calculation unit 262 is acquired by the relationship acquisition unit 25 between the articles represented by the KL record acquired by the KL information acquisition unit 23 according to the rules stored in the rule DB 13 and the type, direction, and role of the relationship. The weight is determined based on the user who has requested the article search and the role of the user recognized by the creator recognition unit 24 (step S180). The weight calculation unit 262 multiplies the determined weight by the score of the article g acquired by the score acquisition unit 261, and acquires the weighted score of the article g (step S181).

  Here, an example of a rule that is applied when an article search request is specified by the above-described user among the rules stored in the rule DB 13 will be described.

The rule DB 13 stores a plurality of rules including, for example, first to fifth rules shown below as rules applied when an article search request is made.
The first rule is associated with “actual charge” which is the role of the requesting user, the role of the user who created the article h is “summary”, and becomes the original article of “consult” If it is, the score of the previous article (in this case, article g) of the “consult” is weighted with a weight of 0.3.

  Similar to the first rule, the second rule is associated with the role “actual charge” of the user of the request source. When the article h is the original article of “sum”, the previous article of the “sum” This indicates that the score of (article g) is weighted with a weight of 0.5.

  The third rule is associated with “summary” that is the role of the requesting user, and the role of the user who created the article h is “administrator” and is the previous article of “follow” , The score of the original article (here, article g) of the “follow” is weighted with a weight of 0.3.

  As with the third rule, the fourth rule is associated with the role “summary” of the user of the request source, and when the article h is the previous article of “sum”, the original article of the “sum” This indicates that the score of (article g) is weighted with a weight of 0.2.

  The fifth rule is associated with “administrator”, which is the role of the requesting user, and if the article h is the previous article of “revise” and the previous article of “consult”, the “consult” ”Indicates that the score of the original article (here, article g) is weighted with a weight of 0.4.

  Descriptions of rules that are applied when an article search request other than the first to fifth rules are omitted.

  The adder 263 adds the weighted score of the article g acquired by the weight calculator 262 to the score of the article h acquired by the article search unit 221 to obtain a new score for the article h (step S182). That is, the adding unit 263 recalculates the score of the article h.

  The processes of steps S170 and S180-S182 are repeatedly executed for all s KL records acquired in step S150 (step S190).

  When the processes of steps S170 and S180-S182 are executed for all of the s KL records, the processes after step S140 are executed again (NO in step S200). The processing after step S140 is executed for all r articles acquired in step S130.

  When the processing from step S140 onward is executed for all the r articles, the sorting unit 27 sorts the r articles based on the score recalculated in step S181 (step S210).

  Next, a processing procedure of the information search unit 14 when an article reference is requested by a user will be described with reference to a flowchart of FIG.

  First, it is assumed that an article reference request designated by the user is input by the input interface 4. This article reference request is added with the information of the requesting user as in the article search request. The user identifying unit 21 identifies the requesting user from the user information added to the article reference request input by the input interface 4 (step S300).

  In response to an article reference request designated by the user, for example, a request for referring to a new article in the community, the article acquisition unit 22 acquires, for example, u articles that match the request from the article DB 8 (step S310). . Here, it is assumed that an article with a new creation date / time has a higher matching degree with respect to an article reference. In the present embodiment, a weight is used as the degree of match for article reference. That is, the article acquisition unit 22 acquires the top u articles in the order of weight according to the novelty of the articles.

  The KL information acquisition unit 23 extracts an article f (f = 1, 2,..., U) from the acquired u articles, and stores all KL records indicating the article f as a previous article or an original article. Obtained from the information DB 9 (step S320).

  The creator recognition unit 24 recognizes the user who created the previous article and the original article from the information indicating the user included in the previous article and the original article indicated by the KL record acquired in step S320 (step S330).

  The role acquisition unit 25 acquires the role indicated by the information stored in the user role information DB 12 in association with the user who created the previous article and the original article of the article f indicated by the acquired KL record (step S340). .

  The weight calculation unit 262 extracts one of the KL records acquired in step S320, and in accordance with the rules stored in the rule DB 13, the relationship between the articles represented by the extracted KL records and the type of the relationship The weight of the article f is determined based on the role of the user who has requested the article reference acquired by the orientation and role acquisition unit 25 and the user role recognized by the creator recognition unit 24 (step S350).

  Here, an example of a rule that is applied when an article reference request is specified by the above-described user among the rules stored in the rule DB 13 will be described.

The rule DB 13 stores a plurality of rules including, for example, the following sixth to eleventh rules as rules applied at the time of article reference request.
The sixth rule is associated with “actual charge” which is the role of the requesting user, the role of the user who created the article f is “summary”, and the article f is newly created. In the case of an article, this indicates that weighting for preferentially displaying the article f is performed.

  The seventh rule is associated with “actual charge” which is the role of the requesting user, and the article f created by the user whose role is “summary” is the original article of “consult”. When the role of the user who created the previous article of “consult” is “actually responsible”, this indicates that weighting for preferentially displaying the article f is performed.

  The eighth rule is associated with “summary”, which is the role of the requesting user, the role of the user who created the article f is “administrator”, and the article f is newly created. In the case of an article, this indicates that weighting for preferentially displaying the article f is performed.

  The ninth rule is associated with “summary” which is the role of the requesting user, and the article f created by the user whose role is “administrator” is the previous article of “follow”. When the original article of “follow” is the previous article of “sum”, it indicates that weighting for preferentially displaying the article f is performed.

  The tenth rule is associated with “summary” which is the role of the requesting user, and the article f created by the user whose role is “actual charge” is the previous article of “follow”. When the role of the user who created the original article of “follow” is “summarize”, it indicates that weighting for preferentially displaying the article f is performed.

  The eleventh rule is associated with “administrator” which is the role of the requesting user, and is created by the user whose role is “summary” and the article f is the destination of “sum” or “revise”. In the case of an article, this indicates that weighting for preferentially displaying the article f is performed.

  Descriptions of rules other than the sixth to eleventh rules applied at the time of article reference request are omitted.

  For each article acquired by the article list unit 222, the adding unit 263 adds the weight based on the novelty of the article and the weight determined by the weight calculating unit 262 (step S360).

  The processes in steps S350 and S360 are repeatedly executed for all the KL records acquired in step S320 (step S370).

  When the processes in steps S350 and S360 are executed for all the KL records acquired in step S320, the processes after step S320 are executed again (NO in step S380). The processes in and after step S320 are executed for all u articles acquired in step S310.

  When the processing in step S320 and subsequent steps is executed for all the u articles, the sorting unit 27 sorts the u articles based on the weight for each article added in step S360 (step S390). ).

  As described above, in the present embodiment, when an article is created by a user, a KL record that represents the type of relationship between the article and an article already stored in the article DB 8 related to the article is used. The role of each user in the community is derived. And when there is a request for article search or article reference from the user, by presenting the result for the request according to the role of the user to the requesting user, the user does not perform a special operation, It is possible to receive presentation of more useful information for the user.

  Note that the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage. In addition, various inventions can be formed by appropriately combining a plurality of components disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment.

  DESCRIPTION OF SYMBOLS 3 ... Communication system, 4 ... Input interface, 5 ... Information registration management part, 6 ... Article storage process part, 7 ... KL storage process part, 8 ... Article DB, 9 ... KL information DB, 10 ... Role pattern dictionary, 11 ... User role derivation unit, 11a ... article list creation unit, 11b ... pattern information list creation unit, 11c ... pattern verification unit, 11d ... addition unit, 11e ... role selection unit, 12 ... user role information DB, 13 ... rule DB, 14 ... information search unit, 15 ... result generation unit, 16 ... output interface, 21 ... user identification unit, 22 ... article acquisition unit, 23 ... KL information acquisition unit, 24 ... creator recognition unit, 25 ... role acquisition unit, 26 ... Article score / weight change unit, 27... Sort unit.

Claims (8)

In a communication system having an article storage means for storing an article created by a user in a community and including an article related to an already created article,
Knowledge link information storage means for storing knowledge link information representing the relationship between the articles and the type of relationship for each of the articles related to each other among the articles stored in the article storage means;
Based on the type of relationship indicated by the knowledge link information, user role derivation means for deriving the role of the user in the community that created the articles related to each other indicated by the knowledge link information;
User role storage means for storing information indicating the role of the user derived by the user role derivation means in association with the user;
In response to a request for article search or article reference from a user, an article acquisition unit that acquires an article that matches the request from the article storage unit based on the degree of matching the request;
A change unit that changes the degree of matching with the request for each article acquired by the article acquisition unit based on a role indicated by information stored in the user role storage unit in association with a requesting user;
Priority for determining the priority for each article when presenting the article acquired by the article acquisition means to the requesting user based on the degree of matching with the request for each article changed by the changing means A communication system comprising: a degree determining means.   When the first article is newly stored in the article storage means, the relationship between the first article and the second article already stored in the article storage means related to the first article 2. The communication system according to claim 1, further comprising knowledge link storage processing means for acquiring knowledge link information representing a type and storing the knowledge link information in the knowledge link information storage means.   The knowledge link storage processing means is a user operation for associating the first article with the second article when the first article newly stored in the article storage means is created by the user. 3. The communication system according to claim 2, wherein the type of relationship between the articles is determined based on the type of knowledge, and knowledge link information representing the determined type of relationship between the articles is acquired. . The knowledge link information includes information indicating a direction for identifying an article newly stored in the article storage unit as a previous article and an article already stored in the article storage unit related to the article as an original article. Have
The communication according to claim 1, wherein the user role deriving unit derives a role corresponding to each of a user who created the previous article and a user who created the original article indicated by the information indicating the direction. system. Combination information indicating characteristics of a combination of types of relations among the plurality of articles included in a standard path indicated by a connection of a plurality of articles in which articles are created starting from a certain article, and the plurality In the article, the weight information indicating the role and the weight corresponding to the user who created the article corresponding to each of the connection depths of the group of articles from the article located at the end of the regular path, Further comprising role pattern storage means for storing the pattern information possessed,
The user role derivation means includes
Whether each of the articles stored in the article storage means is an article located at the end of the standard path, whether or not there are a plurality of articles that match the characteristics indicated by the combination information is determined by the knowledge link. Pattern collating means for judging based on the type or direction of the relationship indicated by the knowledge link information stored in the information storing means;
When the pattern matching unit determines that there are a plurality of articles that match the feature, an adding unit that adds the weight corresponding to each user who created the plurality of articles to the role;
5. The communication according to claim 4, further comprising role selection means for selecting, as a role for each user, a role having a high weight added for each user based on a result added by the addition means. system. The characteristics of the combination indicated by the combination information include a condition for conforming to the combination set for each connection depth,
The pattern collating unit determines that there are a plurality of articles that match the characteristics indicated by the combination information when there is an article that satisfies the condition for each of the connection depths. Item 6. The communication system according to Item 5.   The role indicated by the information stored in the user role storage means in association with the request source user when an article is presented to the request source user based on the priority determined by the priority determination means The communication system according to claim 1, further comprising: a result generation unit that changes the length of the summary sentence of the article presented to the requesting user in response to the request. In a communication system having an article storage means for storing an article created by a user in a community and including an article related to an already created article, a program executed by a computer of the system,
In the computer,
For each article related to each other among the articles stored in the article storage means, storing knowledge link information representing the relationship between the articles and the type of relation in the knowledge link information storage means;
Deriving the role of the user in the community that created the articles related to each other indicated by the knowledge link information based on the type of relationship indicated by the knowledge link information;
Storing in the user role storage means the information indicating the derived user role in association with the user;
In response to an article search or article reference request from the user, obtaining an article that matches the request from the article storage means based on the degree of matching the request;
Changing the degree of agreement with the request for each acquired article based on the role indicated by the information stored in the user role storage means in association with the requesting user;
Determining a priority for each article when presenting the acquired article to the requesting user based on a degree of matching with the request for each changed article. .

Images