JP2005250648A - Article summarizing device and news distributing device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To simplify the overall image of a relevant article group to which articles retrieved from a plurality of news articles under predetermined retrieval conditions belong. <P>SOLUTION: An article DB10 acquires and stores a plurality of articles. In this case, each of the plurality of articles is applied with attribute information for specifying the article group constituted of a plurality of articles preliminarily associated with one another to which the articles belong. An article retrieving part 30 retrieves the articles matched with retrieval conditions stored in a profile DB20 from the plurality of articles stored in the article DB10. A relevant article group extracting part 40 extracts the article group to which the articles belong as a relevant article group from the article DB10 based on the attribute information of the retrieved articles. An article summarizing part 50 generates the summary of the extracted relevant article group. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to an article summarization apparatus that generates a summary of a plurality of articles, and a news distribution apparatus that distributes news on which summaries of a plurality of articles are posted.

  A news distribution service that distributes news articles to users via a network is known. For example, Patent Document 1 discloses a system that distributes articles based on a user's selection.

  There is also a known system that searches for articles that match a user profile that expresses user requests and interests, and displays a list of these headings or the entire article and presents it to the user. In this system, when a large incident occurs and many related articles occur in a short period of time, many related articles may be selected. However, even if similar articles are posted at random, it is difficult for users to read.

  Therefore, in Patent Document 2, an article that matches the search condition is selected, the degree of similarity between the selected articles or the selected article and other articles is calculated, and each article is related according to the degree of similarity. A system for determining an article, adding information on the determined related article to a selected article, and presenting the information to a user is disclosed. Here, as information added to an article, the heading of the article, the information source, the number of characters, and the similarity to the article in which the current text is presented are listed. In addition, a configuration is shown in which the text portion of the additional information is structured with hypertext or the like to enable access to the text of the excluded duplicate article.

  Patent Document 3 also describes a system that generates a summary having a length corresponding to the type (priority) of a searched article and presents the summary to the user.

  Patent Document 4 discloses a technique for creating appropriate summaries and keywords for a plurality of overlapping documents. Patent Document 4 discloses a technique for collectively summarizing the contents of a plurality of text sets having similar contents.

JP 2002-109250 A Japanese Patent Laid-Open No. 9-101990 JP-A-11-39334 JP 2000-305928 A

  The system described in Patent Document 2 has the following problems.

  When adding text such as the article headline, information source, number of characters, etc. as additional information to the searched article, it is possible to know the existence of the related article, but it is difficult to grasp the overall picture of the related article group. . Moreover, even if it is possible to access the related article body from the text portion of the additional information, it is not possible to grasp the overall image of the related article group.

  Further, when a summary of retrieved articles is generated and this summary is presented, it is easy to grasp each retrieved article, but it is not possible to grasp the whole image of related articles.

  Therefore, the present invention provides an article summarizing apparatus that can easily grasp the overall image of related articles to which an article is retrieved from articles retrieved from a plurality of articles.

  In addition, the present invention provides a news distribution device that can distribute to a user news that allows easy understanding of the overall image of related articles to which an article belongs, among articles searched from a plurality of articles.

  The article summarizing apparatus according to the present invention includes an article acquisition unit that acquires a plurality of articles, and an article search that searches for articles that match a preset search condition from among a plurality of articles acquired by the article acquisition unit. And a related article group extracting means for extracting a related article group consisting of a plurality of mutually related articles to which the article searched by the article searching means belongs, and the related article group extraction And article summary means for generating a summary of the related article group extracted by the means.

  In the present invention, each of the plurality of articles is provided with attribute information for specifying an article group including a plurality of articles that are associated with each other in advance, and the related article group extracting unit Preferably, the article group to which the article belongs is extracted as a related article group based on attribute information given to the searched article.

  Preferably, the article summary means divides a plurality of articles belonging to the related article group into a plurality of article groups according to a time series order, and summarizes each article group to generate a summary composed of a plurality of summary parts. .

  Here, it is preferable that the article summarizing means summarizes each article group by a summarizing method according to the order of each time series.

  In a preferred embodiment of the present invention, the article summarizing means generates a summary portion that is simpler than the article group after the predetermined time point for the article group before the predetermined time point.

  Further, in the present invention, the related article group extraction unit determines whether or not there is a relation between the searched articles, and when the search result includes a predetermined number or more of articles that are mutually related, the predetermined article It is preferable to extract a related article group to which more than a few articles belong.

  A news creation apparatus according to the present invention includes any one of the above-described article summarization apparatuses and news creation means for creating news on which a summary generated by the article summary means is posted.

  Further, the news creation device according to the present invention divides a plurality of articles belonging to the related article group into a plurality of article groups in time series order, and summarizes each article group to generate a summary composed of a plurality of summary parts. A news creation that creates a news in which a summary is formed by laying out a summary device and a plurality of summary parts generated by the article summary device in a display form corresponding to the order of each time series. And means.

  In addition, the news creation device according to the present invention, when the search result includes a predetermined number or more of articles related to each other, an article summarization apparatus that extracts a related article group to which the predetermined number of articles or more belongs, News generating means for generating a summary generated by the article summarizing apparatus and news on which articles other than the predetermined number or more of the searched articles are posted.

  In any of the news creation devices described above, the news creation means preferably lays out the summary or article within a predetermined number of pages.

  Further, in a preferred embodiment of any of the above news creation devices, the article search means searches for an article for each user based on a search condition set for each user to whom the news is distributed. The news creation means creates news for each user.

  A news distribution device according to the present invention includes the news creation device according to any one of the above-mentioned items, and news distribution means for distributing news created by the news creation device to a user. .

  The article summarizing method according to the present invention is a method executed by the article summarizing apparatus, and is set in advance from an article acquiring step for acquiring a plurality of articles and a plurality of articles acquired in the article acquiring step. An article search step for searching for articles that match the search conditions, and a related article group consisting of a plurality of mutually related articles to which the article searched in the article search step belongs is extracted from the plurality of articles. An article group extraction step and an article summary step for generating a summary of the related article group extracted in the related article group extraction step.

  The article summary program according to the present invention searches an article that matches a preset search condition from among an article acquisition step of acquiring a plurality of articles and a plurality of articles acquired in the article acquisition step. An article search step, a related article group extraction step for extracting a related article group consisting of a plurality of mutually related articles to which the article searched in the article search step belongs, and the related article And an article summary step for generating a summary of the related article group extracted in the article group extraction step.

  In the present invention, an article that matches a predetermined search condition is searched from a plurality of articles. Then, a related article group including a plurality of mutually related articles to which the searched article belongs is extracted, and a summary of the extracted related article group is generated. Therefore, according to the present invention, it is possible to obtain a summary of a related article group to which the article belongs with respect to an article searched from a plurality of articles, and to easily grasp the overall image of the related article group. Can do.

  In addition, according to the present invention, since the news on which the summary of the related article group is posted is created and this news is distributed to the user, the article to which the article belongs is searched for from among a plurality of articles. News that makes it easy to grasp the overall picture of an article group can be distributed to users.

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

  FIG. 1 is a block diagram showing an overall configuration of a news distribution system including a news distribution apparatus 1 according to an embodiment. In this news distribution system, a news distribution apparatus 1 managed by a news provider and one or more user terminals 2 used by a user to acquire news are connected to each other via a network 3 such as the Internet. Yes. As an example of the user terminal 2, there is a printer multifunction device that is installed in a store or the like and has a user interface. However, the user terminal 2 may be another information communication terminal such as a personal computer or a mobile phone.

  The news distribution apparatus 1 searches for articles that match a predetermined search condition from a plurality of news articles (hereinafter simply referred to as articles), creates news using the searched articles, and creates the news Is distributed to the user (specifically, the user terminal 2). In the present specification, news means that one or more articles or summaries are posted, and in particular, one or more articles or summaries are laid out on paper or on a screen. Here, the summary is a summary of the contents of a plurality of articles.

  In the present embodiment, article search and news creation are performed for each user. That is, the news distribution device 1 distributes news customized for each user to each user. However, news does not necessarily have to be customized, and article search and news creation may be performed uniformly for all users.

  Here, articles handled in this embodiment will be described. The articles supplied to the news distribution apparatus 1 are roughly classified into two types: a new article not related to an article before the article (also referred to as an already published article) and a subsequent article related to the already published article. Then, there are two types of follow-up articles: a follow-up article that is a continuation of the already published article in a series of article reports and the like, and a corrected article that corrects an error in the already published article. However, the type of subsequent articles is not particularly limited.

  FIG. 2 is a schematic diagram for explaining the relationship between a plurality of articles. In FIG. 2, a circle indicates an article. The arrows indicate the relationship between articles. The article on the rear end side of the arrow is a subsequent article (follow-up article or correction article) with respect to the article on the front end side. Specifically, the article A1 is a new article, the article A2 is a succeeding article of the article A1, and the article A4 is a succeeding article of the article A2.

  Hereinafter, the basic concept of the present embodiment will be described with reference to FIG. First, in FIG. 2, a case where articles A20, A30, A40, and A50 are searched under a predetermined search condition will be described as an example. In this case, if the articles A20, A30, A40, and A50 are posted in the news as they are, many similar articles are posted, which is very difficult for the user to read. In order to avoid this, it is conceivable to summarize the articles A20, A30, A40, and A50 and post the obtained summary in the news. However, in this method, a summary of some articles in the article groups A1 to A50 is posted, and the user cannot grasp the entire image of the article groups A1 to A50.

  Therefore, in the present embodiment, a related article group including a plurality of mutually related articles to which the searched article belongs is extracted from the plurality of articles shown in FIG. I will summarize. Here, various methods can be considered as a method for extracting the related article group, but in order to facilitate understanding of the overall image of the article group A1 to A50, the article group A1 to A50 is surely included as the related article group. Preferably extracted and summarized.

  Therefore, in this embodiment, a plurality of articles related to each other are associated with each other in advance. Then, attribute information for specifying an article group composed of a plurality of articles associated with each other in advance is assigned to each of the plurality of articles. Then, based on the attribute information given to the searched article, an article group to which the searched article belongs is specified, and this article group is extracted as a related article group and summarized.

  Specifically, the articles A1 to A50 are associated with each other in advance. Then, attribute information for specifying the article groups A1 to A50 is assigned to each of the articles A1 to A50. For example, when an article A20 is searched, based on the attribute information given to the article A20, the article group A1 to A50 to which the article A20 belongs is specified, and the article group A1 to A50 is extracted as a related article group. And to summarize.

  As described above, in this embodiment, when any article is searched from among the article groups A1 to A50 including a plurality of mutually related articles, the article groups A1 to A50 are surely extracted and summarized. The Thereby, it becomes easy to grasp the whole image of the article groups A1 to A50.

  In the present embodiment, the following three types of attribute information are given to each article. First, an article ID that is identification information for identifying the article is assigned. Also, the article ID of an already published article related to the article (hereinafter referred to as the original article of the article) is assigned as the original article ID. Further, an article ID of an article whose original article is the article is assigned as a subsequent article ID. Specifically, in the article A3, the article ID of the article A3, the article ID of the original article A1 of the article A3 (original article ID), and the article IDs of the articles A5 and A6 having the article A3 as the original article (following) Article ID). Note that since the original article does not exist in the article A1, which is a new article, attribute information indicating that the original article does not exist is assigned instead of the original article ID. Further, since there is no subsequent article having itself as the original article, the article A50 is provided with attribute information indicating that there is no subsequent article instead of the subsequent article ID. A plurality of articles A1 to A50 are associated with each other by such attribute information. However, the attribute information added to each article may be anything as long as it can identify an article group including a plurality of articles associated with each other in advance. For example, the article group ID “A” may be assigned to the article groups A1 to A50, and the article group ID “A” may be assigned to each of the articles A1 to A50 as attribute information.

  Next, a method for summarizing related articles will be described. The related article group may be summarized in a lump, but in the present embodiment, a plurality of articles belonging to the related article group are summarized in consideration of the time series order. Specifically, the articles A1 to A50 are divided into a plurality of article groups in time series order, and each article group is summarized by a summarization method according to the order of each time series, and a summary composed of a plurality of summary parts is generated. More specifically, the article group A1 to A50 is divided into an article group before the predetermined time point (for example, the article group A1 to A20) and an article group after the predetermined time point (for example, the article group A21 to A50). For the article group, a summary part simpler than the article group after a predetermined time point is generated. Thereby, the user can grasp | ascertain an outline about a comparatively old part among article groups A1-A50, and can grasp | ascertain in detail about a comparatively new part.

  Note that the summarization method corresponding to the predetermined time point or the order of the time series may be appropriately set by the distributor or the user. For example, it is possible to divide the article groups A1 to A50 into article groups for each date and generate a dated summary for each article group.

  Next, the case where the article B4 is searched with a predetermined search condition in FIG. 2 will be described as an example. In this case, similarly to the above, it is one of preferred embodiments to extract and summarize the article groups B1 to B5 based on the attribute information of the article B4, and to post the obtained summary in the news. However, when there are a small number of articles searched (one in this example), even if the searched articles are posted as they are, the amount is small, so it is considered that it is not so difficult for the user to read. In addition, since the number of articles searched is small, it is presumed that this is not a big incident.

  Therefore, in the present embodiment, based on the attribute information given to each article, it is determined whether or not there is a relation between searched articles, and the search result includes a predetermined number of articles that are related to each other. The related article group to which the predetermined number or more of articles belong is extracted and summarized. On the other hand, among the retrieved articles, articles that do not constitute a predetermined number or more of mutually related articles are not included in the summary, and the articles are directly posted in the news. Specifically, when the predetermined number is “3” and the articles A20, A30, A40, A50, B3, and B4 are searched, the summary of the article groups A1 to A50 and the articles B3 and B4 are posted in the news. Will be. In addition, what is necessary is just to set said predetermined number suitably by the distributor side or a user.

  Hereinafter, the configuration of the news distribution apparatus 1 will be described.

  FIG. 3 is a block diagram showing a functional configuration of the news distribution apparatus 1. 3, the news distribution apparatus 1 includes an article database 10 (hereinafter referred to as an article DB), a profile database 20 (hereinafter referred to as a profile DB), an article search unit 30, a related article group extraction unit 40, and an article summary unit 50. A news creation unit 60 and a news distribution unit 70.

  The article DB 10 acquires and stores a plurality of articles. Here, attribute information for specifying an article group including a plurality of articles associated with each other in advance, to which the article belongs, is assigned to each of the plurality of articles. The configuration of data stored in the article DB 10 will be described in detail later with reference to FIG.

  The articles are supplied to the article DB 10 by appropriate means. For example, it may be supplied via a network from an article supply source such as a newspaper company, or may be input via a user interface by a news distribution service provider. The article attribute information is given in advance by a news distribution service provider or article supplier.

  The profile DB 20 stores profile information for each user as information regarding individual users who receive news distribution. Here, the profile information for each user includes at least a search condition for each user used for searching for articles. Here, the profile information for each user is generated based on the user registration information received from each user prior to the provision of the news distribution service. The configuration of data stored in the profile DB 20 will be described in detail later with reference to FIG.

  The article search unit 30 searches for articles that match the search conditions stored in the profile DB 20 from among a plurality of articles stored in the article DB 10. In the present embodiment, the article search unit 30 searches for articles for each user based on search conditions set for each user.

  The related article group extraction unit 40 extracts a related article group including a plurality of mutually related articles to which the article searched by the article search unit 30 belongs, from a plurality of articles stored in the article DB 10. In the present embodiment, the related article group extraction unit 40 determines whether or not there is a relation between searched articles based on the attribute information of each article, and a predetermined number or more of articles related to each other are included in the search result. If included, the related article group to which the predetermined number or more of articles belong is extracted. At this time, the related article group extraction unit 40 extracts the related article group based on the attribute information given to the searched article.

  The article summary unit 50 generates a summary from the related article group extracted by the related article group extraction unit 40. In the present embodiment, the article summary unit 50 divides the related article group into an article group before the predetermined time point and an article group after the predetermined time point, and the article group before the predetermined time point is simpler than the article group after the predetermined time point. A simple summary part. Note that the summarization algorithm used in the article summarization unit 50 is not particularly limited, and various summarization techniques can be applied. For example, an automatic summary generation technique using natural language analysis may be used, or a simple one in which only the first paragraph is used as a summary sentence may be used.

  The news creation unit 60 uses the one or more summaries generated by the article summarization unit 50 or one or more articles that were not included in the retrieved articles to be posted one or more summaries or articles. Create news. In the present embodiment, the news creation unit 60 lays out a summary or article within a preset number of pages. Further, when a summary composed of a plurality of summary parts is generated by the article summary unit 50, the plurality of summary parts are laid out in a display form corresponding to the order of each time series.

  The news distribution unit 70 distributes the news created by the news creation unit 60 to the user (specifically, the user terminal 2) via the network 3.

  In the present embodiment, news distribution apparatus 1 having the above configuration is realized by a general computer. Each of the units 10 to 70 is realized by a program corresponding to each function being executed by computer hardware resources such as a CPU, ROM, RAM, and hard disk drive. For example, the article search unit 30, the related article group extraction unit 40, and the article summary unit 50 are realized by the article summary program according to the present embodiment stored in the storage medium being executed by the CPU. However, the implementation mode of the news distribution apparatus 1 is not particularly limited.

  Next, the structure of data stored in the article DB 10 and the profile DB 20 will be described.

  FIG. 4 is a diagram illustrating a configuration example of data stored in the article DB 10. As shown in FIG. 4, the data in the article DB 10 is composed of one or more “article information” records. The “article information” record includes “article ID”, “article genre”, “registration date”, “distribution type”, “original article ID”, “subsequent article ID list”, “keyword”, “heading”, And a plurality of fields such as “article”.

  The “article ID” field holds an article ID that is identification information for identifying an article. The “article genre” field holds the genre of articles such as politics, society, and sports. The “registration date” field holds the date when the article is registered in the article DB 10.

  The “distribution type” field holds information indicating whether the article is a new article, a follow-up article, or a corrected article. The “original article ID” field holds the article ID of the original article when the article is a related article (follow-up article or corrected article). The “subsequent article ID list” holds an article ID of a subsequent article whose original article is the article, that is, an article ID of an article that refers to the article in the “original article ID” field. For example, the article ID of the article A1 is described in the “original article ID” field of the article A2 in FIG. 2, and the article ID of the article A3 is described in the “original article ID” field of the article A5. When there is no original article to be referred to like article A1, the “original article ID” field is empty (NONE) as shown in “article information 1” in FIG. Further, the article IDs of articles A2 and A3 referring to the article A1 as the original article are displayed in the “subsequent article ID list” field of the article A1, and the article A3 is stored in the “subsequent article ID list” field of the article A3. The article IDs of articles A5 and A6 referred to as the original article are described. When there is no article that refers to itself, such as the articles B2, B4, and B5, the “subsequent article ID list” field is empty (NONE) as shown in “Article information 2” in FIG. . When neither the original article referred to as the article C1 nor any other article referring to itself exists, both the “original article ID” field and the “subsequent article ID list” field are empty (NONE).

  The “Heading” field holds the heading (title) of the article. The “article” field holds the article itself, that is, the article body. The article may include still images, moving images, sounds, and the like.

  As described above, article attribute information is registered in fields other than the “article” field among the plurality of fields constituting the “article information” record.

  FIG. 5 is a diagram illustrating a configuration example of data stored in the profile DB 20. FIG. 5 shows the profile information of a single user. This profile information is created and stored for each user.

  As shown in FIG. 5 (A), each user's profile information includes “user authentication information”, “personal information”, “selection conditions”, “distribution method”, and “summary information”. It is composed of fields.

  The “user authentication information” field holds a user ID and a password that are used to identify individual users. The “personal information” field holds the user's address, telephone number, and the like. The “selection condition” field holds a search condition for searching for an article, the number of sheets at the time of distribution, and the like. The “distribution method” field holds information on how to distribute to the user. In this example, a distribution request from a user is received, and distribution to a printer multifunction device designated at that time is shown.

  The “summary information” field holds information for dividing the related article group into an article group before and after a predetermined time point. Specifically, in the “summary information” field, the “detailed summary period” field holds information indicating a period after a predetermined time point, that is, a period during which a detailed summary portion is to be generated. The “last delivery date / time” field holds the date / time when the previous delivery was performed. In the example shown in FIG. 5A, the “detailed summarization period” field is after the last delivery date and time, and this indicates that the date after the last delivery is the subject of the detailed summarization. . Here, since the “last delivery date / time” field is 17:59 on December 31, 2002, articles after this date / time are detailed summaries, and articles before this date / time are simple summaries. It becomes. In the “detailed summary period” field, another period, for example, a fixed period such as within the past 24 hours may be designated.

  FIG. 5B is a diagram illustrating an example of search conditions for searching for articles. In this example, the search condition includes two fields of “keyword” and “type”. The keyword of the record whose “type” field is the article genre is treated as a search key for the “article genre” field of the article DB 10 shown in FIG. The keyword of the record whose “type” field is the article text is handled as a search key for the “article” field of the article DB 10 shown in FIG. 4, that is, a search key for the entire text.

  Hereinafter, the operation of the news distribution apparatus 1 having the above configuration will be described in detail.

[Whole news distribution process]
Here, the overall flow of the news distribution process will be briefly described.

  FIG. 6 is a flowchart showing an operation procedure of news distribution processing. The news distribution apparatus 1 receives input of the user ID and password of the user A from a user who wants to distribute news (hereinafter referred to as user A), and identifies the user A (S601). Next, an article that matches the search condition of the user A stored in the profile DB 20 is searched from the articles stored in the article DB 10 (S602). When the retrieved articles include a predetermined number or more of mutually related articles, a related article group to which the predetermined number or more of articles belong is extracted from the article DB 10, and a summary of the extracted related article group is generated ( S603). Next, the summary generated in step S603 and the article searched in step S602 and not the summary target in step S603 are laid out to create news (S604), and user A uses the created news. Is distributed to the existing user terminal 2 (S605). Then, the updating process of the profile information of the user A such as updating the article delivery date and time is performed (S606). For example, the information in the “last delivery date” field shown in FIG. 5 is updated.

  Note that in step S602 described above, the articles to be searched may be a part or all of a plurality of articles stored in the article DB 10. However, since the search is intended for news distribution, it is desirable that the articles to be searched are within a predetermined period of time such as the search day, within a predetermined period from the search day, or after the most recent distribution date and time. . Further, such a search time condition may be set as a search condition for each user.

  In this example, the news distribution apparatus 1 performs article layout in step S604. However, article layout may be performed on the user terminal 2 side.

  When the related article group is summarized in step S603, the amount of articles laid out in step S604 changes. When laying out articles in a specific number, it is also possible to obtain margin areas that can be laid out after summarization processing or layout processing, and repeat the processing of adding distribution articles (S602 to S604) based on this.

[Related article summary processing]
FIG. 7 is a flowchart showing an example of related article group summarization processing in step S603 of FIG. This related article group summarization process is performed by the related article group extracting unit 40 and the article summarizing unit 50. In the present embodiment, the summary is text, and the summary is hereinafter referred to as a summary sentence.

  First, the related article group extraction unit 40 identifies the root article of each article searched in step S602 (S701). The root article of an article means an article that can be traced through the original article of the article in order, and in the example of FIG. 2, articles A1, B1, and C1 correspond to this. That is, all the root articles from articles A1 to A50 in FIG. 2 are articles A1, and all the root articles from articles B1 to B5 are articles B1. Therefore, it can be said that a plurality of articles having a common root article are a plurality of articles related to each other.

  Next, compare each root article to find the number of articles with the same root article, and in the search results, more than a predetermined number of articles with the same root article, that is, more than a predetermined number related to each other. It is determined whether an article exists (S702).

  If it is determined that there are more than a predetermined number of articles (S702: YES), a related article group to which the predetermined number or more articles belong is extracted based on the root article (S703). Based on this, a summary sentence is generated (S704). Then, after deleting the article to be summarized from the search result (S705), the process returns to step S702.

  On the other hand, if it is determined that there are no more than a predetermined number of articles (S702: NO), the process is terminated. In this manner, related article groups are extracted and summarized for all sets of a predetermined number or more of articles related to each other included in the search result.

  The route article specifying process in step S701, the related article group extracting process in step S703, and the summary sentence generating process in step S704 will be described in detail below with reference to FIGS.

  FIG. 8 is a flowchart showing a more detailed process example of the route article specifying process in step S701 of FIG. This route article specifying process is performed by the related article group extraction unit 40.

  First, the related article group extraction unit 40 substitutes the article ID of the article that is the target of the root article specifying process into the variable id (S801), and acquires the original article ID of the article that uses id as the article ID (S802). Specifically, the “article information” record whose id is stored in the “article ID” field is specified, and the original article ID is acquired from the “original article ID” field of the record. Next, it is determined whether the original article ID is empty, that is, whether it is a root article (S803). If it is not empty (S803: NO), the original article ID is set to the variable id (S804), and the process goes to step S802. Return. On the other hand, if the original article ID is empty (S803: YES), the article with the variable id as the article ID is set as the root article (S805), and the process is terminated.

  FIG. 9 is a flowchart showing a more detailed processing example of the related article group extraction processing in step S703 of FIG. This related article group extraction processing is performed by the related article group extraction unit 40.

  First, the related article group extraction unit 40 acquires a subsequent article ID list set in the root article (S901), and extracts articles corresponding to the article IDs included in the list as articles constituting the related article group. (S902). Specifically, the article ID held in the “subsequent article ID list” field of the “article information” record of the root article is acquired, and the article identified by this article ID is extracted. Next, for each subsequent article in the subsequent article list acquired in step S901, each subsequent article is further acquired (S903, S904). Here, the details of the processing in step S904 are the processing in FIG. That is, the process of FIG. 9 is recursively executed in step S904, and articles registered in the “subsequent article ID list” are sequentially extracted as articles constituting a related article group.

  FIG. 10 is a flowchart showing a more detailed process example of the summary sentence generation process in step S704 of FIG. This summary sentence generation process is performed by the article summary unit 50.

  First, the article summary unit 50 refers to the profile information of the user A, acquires the value of the “detailed summary period” field (S1001), and determines whether a period is specified (S1002).

  When the period is specified (S1002: YES), the corresponding date and time is specified based on the current date and time, and this is set as a predetermined time t (S1003). For example, when the detailed summary period information is “24 hours”, the predetermined time t is 24 hours before the current date and time. If “after the last delivery date and time” is specified, the value of the “last delivery date and time” field becomes the predetermined time t with reference to the profile information of the user A. Next, among a plurality of articles belonging to the related article group to be summarized, articles registered before a predetermined time t are acquired (S1004), and summary processing is performed on these articles (S1005). Further, among a plurality of articles belonging to the related article group, articles registered after a predetermined time t are acquired (S1006), and summary processing is performed on these articles (S1007). Here, the summary process in step S1005 generates a simple summary sentence as compared with the summary process in step S1007. For example, only the heading part of the root article can be used as a summary sentence, or only the first sentence or paragraph of the root article can be used as a summary sentence.

  On the other hand, if the period has not been designated (S1002: NO), all articles belonging to the related article group to be summarized are acquired (S1008), and summary processing is performed on these articles (S1009).

[News creation process]
FIG. 11 is a flowchart showing an example of the news creation process in step S604 of FIG. This news creation process is performed by the news creation unit 60.

  First, the news creation unit 60 selects one of the summary sentences generated in step S603 or the articles that have not been laid out from the articles retrieved in step S602 and not summarized in step S603. (S1101).

  If the selected item is an article (S1102: NO), a layout (allocation on the page) process is performed for the article (S1106), and the process proceeds to step S1107 to be described later.

  On the other hand, if the selected sentence is a summary sentence (S1102: YES), the simplified summary part generated in step S1005 is laid out (S1103), and then the normal summary part generated in step S1007 is laid out (S1104). ). Furthermore, information indicating that the part is a summary sentence is laid out (S1105).

  In step S1107, it is determined whether layout processing has been completed for all summary sentences and articles. Then, when all the processes are not completed (S1107: NO), the process returns to step S1101, and when all the processes are completed (S1107: YES), the process is terminated. That is, the above processing (S1101 to 1106) is performed for all articles and summary sentences.

  Here, it is preferable that a plurality of summary portions generated from a plurality of article groups are laid out in a display form corresponding to the order of each time series. In the above example, the simplified summary portion and the normal summary portion are preferably laid out in different display forms.

  FIG. 12 is a diagram illustrating an example of news (article layout) created by the news creation unit 60 as described above. In FIG. 12, information such as a title and delivery date, and one or more articles and a summary sentence are laid out on a printed sheet 1200. Here, a summary sentence is laid out in the area 1201. Specifically, the area 1201 includes a headline part, a simple summary part generated in step S1005 and laid out in step S1103, a normal summary part generated in step S1007 and laid out in step S1104, and a summary sentence. A sign 1202 indicating that the above is laid out. Here, the simple summary portion is displayed with a smaller character size than the normal summary portion.

  The news created as described above is distributed to the user A by the news distribution unit 70.

  According to the present embodiment described above, the following effects can be obtained.

  (1) In this embodiment, an article that matches a predetermined search condition is searched from a plurality of articles. Then, a related article group including a plurality of mutually related articles to which the searched article belongs is extracted, and a summary of the extracted related article group is generated. Therefore, according to the present embodiment, when an article that matches a predetermined search condition is searched from a plurality of articles, a summary of the related article group to which the article belongs can be obtained. It is possible to facilitate the grasp of the whole image. Moreover, in the news distribution apparatus that searches for articles that match a predetermined search condition as distribution candidates among a plurality of articles by posting the obtained summary in the news, the entire related article group to which the searched articles belong It is possible to distribute news that allows easy grasp of images to users.

  From another perspective, it is possible to generate summaries that include other related articles that could not be obtained by keyword search, etc., so that even a relatively small amount of text can be easily understood. It becomes possible to do.

  (2) Attribute information for identifying an article group consisting of a plurality of articles that are associated with each other in advance and belong to each of the plurality of articles is assigned to the searched article. Based on the information, the article group to which the article belongs is extracted as a related article group. For this reason, an article group consisting of a plurality of articles related to each other is defined in advance, and when an article belonging to the defined article group is searched, the defined article group is reliably set as a related article group. Can be extracted. Therefore, an appropriate article group can be extracted as a related article group by appropriately setting an article group such as the article group A1 to A50 or the article group B1 to B5 in FIG. For example, in FIG. 2, when the article A40 is searched, there is no inconvenience that the articles A1 to A10 are leaked from the related article group, and that the related article group includes the articles B3 and C1.

  (3) Multiple articles belonging to related articles are divided into multiple articles in chronological order, and each article is summarized to generate a summary consisting of multiple summaries, facilitating a series of events. A summary that can be grasped can be generated.

  Regarding this, a case where the registration date of an article belonging to a related article group of a certain case is in a range from January 1, 2004 (case occurrence date) to February 20, 2004 (news distribution date) is taken as an example. I will explain it. In this case, these articles are a series of articles related to the case, but the contents of the articles around the date of the incident and the articles around the news distribution date are considered to be significantly different. Therefore, it is not possible to grasp the transition of incidents simply by summarizing these related articles. Alternatively, when an attempt is made to generate an easy-to-understand summary of incident transitions, it is necessary to use an advanced summary algorithm, and the processing load on the news distribution apparatus becomes excessive. On the other hand, in this embodiment, the related article group is, for example, an article group from January 1 to January 15; an article group from January 16 to January 31; and February 1 to February. The article group is divided into an article group of 10 days and an article group of February 11 to February 20, and each article group is summarized to generate a summary composed of a plurality of summary parts. For this reason, it is possible to generate a summary that makes it easy to understand the transition of an incident while avoiding an excessive processing load. At this time, if time information such as date and period is included in each summary part, it is possible to generate a summary that can easily grasp what happened. Also, at the time of layout, it is preferable that the plurality of summary portions are arranged in chronological order.

  (4) The related article group is divided into a plurality of article groups according to the time series order, and each article group is summarized by the summarization method according to the order of each time series. Can do. For example, it is possible to generate a summary in consideration of a situation where a relatively detailed summary is desired for a relatively new article and a relatively simple summary is desired for a relatively old article.

  (5) When a predetermined number or more of articles related to each other are included in the search result, a related article group to which the predetermined number or more of articles belong is extracted, so that a predetermined number or more of articles related to each other are searched. Only in cases, the related article group extraction process and the article summary process can be performed, and the processing load can be reduced.

  As mentioned above, although embodiment of this invention was described, it cannot be overemphasized that this invention is not limited to said embodiment.

  For example, the article summary device according to the present invention can be applied to uses other than news distribution. For example, when a user accesses an article database and searches for an article by keyword search, if a summary of the entire related article group is provided to the user as a search result together with the searched article, it is very much for the user. Convenient.

  Further, the article is not limited to a news article, but may be another article such as a magazine.

  In the above embodiment, the related article group is extracted based on the attribute information given to the article, but the extraction of the related article group may be performed by other methods. For example, the degree of similarity between the searched article and another article in the article DB 10 can be calculated, and articles whose similarity is a predetermined value or more can be extracted as a related article group. In this case, when an article A20 is searched in FIG. 2, articles A1 to A50 are not necessarily extracted.

It is a block diagram which shows the whole structure of the news delivery system containing the news delivery apparatus which concerns on embodiment. It is a schematic diagram for demonstrating the relationship of a some article. It is a block diagram which shows the function structure of a news delivery apparatus. It is a figure which shows the structural example of the data stored in article DB. It is a figure which shows the structural example of the data stored in profile DB. It is a flowchart which shows the operation | movement procedure of news delivery processing. It is a flowchart which shows an example of the related article group summary process in step S603 of FIG. It is a flowchart which shows the more detailed process example of the root article specific process in step S701 of FIG. It is a flowchart which shows the more detailed process example of the related article group extraction process in step S703 of FIG. It is a flowchart which shows the more detailed process example of the summary sentence production | generation process in step S704 of FIG. It is a flowchart which shows an example of the news preparation process in step S604 of FIG. It is a figure which shows an example of news.

Explanation of symbols

  1 news distribution device, 2 user terminal, 3 network, 10 article DB, 20 profile DB, 30 article search section, 40 related article group extraction section, 50 article summary section, 60 news creation section, 70 news distribution section.

Claims (14)

Article acquisition means for acquiring a plurality of articles;
Article search means for searching for articles that match a preset search condition from among a plurality of articles acquired by the article acquisition means;
Related article group extracting means for extracting a related article group consisting of a plurality of mutually related articles to which the article searched by the article search means belongs,
Article summary means for generating a summary of related article groups extracted by the related article group extraction means;
An article summarizing apparatus characterized by comprising: The article summary device according to claim 1,
Each of the plurality of articles is provided with attribute information for specifying an article group including a plurality of articles associated with each other in advance, to which the article belongs,
The related article group extracting means extracts the article group to which the article belongs as a related article group based on attribute information given to the searched article. The article summarization apparatus according to claim 1 or 2,
The article summarizing means divides a plurality of articles belonging to the related article group into a plurality of article groups in time series order, and summarizes each article group to generate a summary including a plurality of summary parts. Article summary device. The article summary device according to claim 3,
The article summarizing means summarizes each article group by a summarizing method according to the order of each time series. The article summary device according to claim 4,
The article summarizing unit generates a summary part simpler than an article group after a predetermined time point for an article group before a predetermined time point. The article summary device according to any one of claims 1 to 5,
The related article group extracting means determines whether or not there is a relation between the searched articles, and if the search result includes a predetermined number or more related articles, the predetermined number or more of the articles belong. An article summarization apparatus characterized by extracting a related article group. The article summary device according to any one of claims 1 to 6,
News creation means for creating news on which the summary generated by the article summary means is posted;
A news creation apparatus characterized by comprising: The article summary device according to any one of claims 3 to 5,
A plurality of summary parts generated by the article summary means are laid out in a display form corresponding to the order of each time series, and a news creation means for creating a news on which a summary composed of a plurality of summary parts is posted;
A news creation apparatus characterized by comprising: The article summary device according to claim 6;
A news creation means for creating a summary generated by the article summarizing means and news in which articles other than the predetermined number or more of the searched articles are posted;
A news creation apparatus characterized by comprising: The news creation device according to any one of claims 7 to 9,
The news creation device, wherein the news creation means lays out the summary or article within a predetermined number of pages. The news creation device according to any one of claims 7 to 10,
The article search means searches for articles for each user based on search conditions set for each user to whom the news is distributed,
The news creation device, wherein the news creation means creates news for each user. The news creation device according to any one of claims 7 to 11,
News distribution means for distributing news created by the news creation device to users;
A news distribution apparatus comprising: An article summarization method executed by an article summarization apparatus,
An article acquisition step for acquiring multiple articles;
An article search step for searching for articles that match a preset search condition from a plurality of articles acquired in the article acquisition step;
A related article group extracting step of extracting a related article group consisting of a plurality of mutually related articles to which the article searched in the article searching step belongs;
An article summary step for generating a summary of the related article group extracted in the related article group extraction step;
An article summarizing method characterized by comprising: On the computer,
An article acquisition step for acquiring multiple articles;
An article search step for searching for articles that match a preset search condition from a plurality of articles acquired in the article acquisition step;
A related article group extracting step of extracting a related article group consisting of a plurality of mutually related articles to which the article searched in the article searching step belongs;
An article summary step for generating a summary of the related article group extracted in the related article group extraction step;
An article summarization program characterized in that it is executed.

Images