JP2001325272A - Information arrangement method, information processor, storage medium and program transmitter - Google Patents

Abstract

PROBLEM TO BE SOLVED: To freely combine plural information sources and to display information which is the topic of the conversation in an easy-to-understand form. SOLUTION: By periodically observing the dynamically changing plural information sources acquired from the Internet, support relation between sites and the degree of the interest of an individual, etc., are taken into consideration, the more important topics of the conversation are automatically extracted from extracted information elements and they are gathered and easily understandably visualized. That is, this method is provided with a step (S102) for periodically going round the registered plural information sources and gathering the information, a step (S103) for selecting words to be the elements of the topic from the gathered information, a step (S104) for executing clustering to the set of the selected words and a step (S105) for displaying the information elements in respective clusters based on a time base on the basis of the result of clustering and displaying a main keyword from the set of the words in the respective clusters.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to information retrieval from information sources, and more particularly to a method for extracting topical information from a plurality of information sources on the Internet and visualizing the information.

[0002]

2. Description of the Related Art In recent years, with the development of the Internet, the amount of information available to users has become enormous. Among these enormous sources, information search technology for organizing and providing information desired by users as quickly as possible, accurately, and in a form that is easy for users to use is becoming increasingly important.

[0003] As a conventional information retrieval technique, for example, there is a technique of extracting an element (link and its title, continuation of text, etc.) that conveys information from a registered information source (site), and performing a language analysis of the text part. . There is also a technology for extracting topics using a portal site that provides information providing services such as search services and news. This portal site provides a service that provides a keyword that has been a topic by manual creation. For example, there is a service that provides a user with a keyword ranking or the like that is a topic for a searcher.

[0004] Reference 1 (J. Kleinberg. Authoritati)
ve sources in a hyperlinked environment.Proc. 9th
ACM-SIAM Symposium on Discrete Algorithms,. Also
appears as IBM Research Report RJ 10076, May 199
7.) discloses a technique for calculating importance in consideration of a static structural reference relationship (support) at a certain moment on the Internet. Here, the authoritative page (Authority) for the specified search expression,
Pages (Hub) containing many authoritative pages are extracted. Also, in Reference 2 (Yanagase, Nakao, "Automatic Extraction of Hot News Using Email Magazine," 57-20, p. 151-158, Proc. Of the Information Processing Society of Japan, 3/22/2000.) Is
A technology is disclosed in which a plurality of information sources (e-mail magazines) are used as information sources, and a noticeable news is automatically extracted therefrom. Here, the number of information sources is large as the importance of the clustered result (many types of e-mail magazines)
The index is used.

[0005] On the other hand, in Japanese Patent Application Laid-Open No. Hei 8-287074, the frequency of occurrence of unregistered words appearing in recent documents, such as continuously issued documents, is monitored in real time and relates to topics that are currently attracting attention. A technique for defining terms and documents to a user is disclosed. Japanese Patent Application Laid-Open No. H11-143892 discloses a technique of generating a keyword weight by combining a keyword weight appearing in a sentence and a weight in consideration of category information. Furthermore, Japanese Patent Application Laid-Open No. H11-143796 discloses a technique for extracting main topics exchanged in each mailing list in a mailing list service.

[0006]

As described above, it is very useful to sort information and automatically extract topics that are topical, and to display them in an easily understandable manner. The proposal has been made. However, in the portal sites mentioned above, topics are extracted manually for news in important fields, but there is a possibility that the evaluation criteria for information is biased in a single site, and important information is overlooked There is a fear that there may be fear or that all information on one topic cannot be obtained. If you try to view multiple sites to avoid this fear, there is a problem that information is duplicated. In addition, since the viewpoints vary, it is necessary to rearrange from another viewpoint in order to facilitate understanding for the user. In addition, for news in fields that many readers cannot expect, there is no service such as manually organizing information, and users need to collect and organize multiple sites themselves. .

On the other hand, the above-mentioned Document 1 does not include a technique for extracting a topic of interest,
Although the keywords in the search expression are used to weight the reference relationship, the results do not include the words themselves. Reference 2 does not consider whether a word has newly appeared. In addition, since the index indicating that the number of information sources is large is used for determining the importance of a cluster and not for determining the importance of a word, the introduction of support cannot affect the result of clustering.

[0008] Further, Japanese Patent Laid-Open No. 8-287070 discloses a static information source (document).
Information from the same source, such as the Internet, is recognized as a separate document. For this reason, only the number of documents is considered in the calculation of importance. In addition, it does not include structuring between documents and cannot utilize the characteristics of document classification such as identifying words that frequently appear in a certain document set. Furthermore, only words or compound words that are not registered in the dictionary are to be extracted, and if a sentence is created by combining the words after registration, it cannot be extracted as a new concept. As a result, for example, sentences such as "Japan IBM announces a new database product" are all composed of registered words, and cannot be extracted as new concepts by this method. Furthermore, even if there is a related new word, since the similarity is not considered, the related new words cannot be viewed at the same time, and the registration work of the related new words is separated, and the same document is obtained each time. A set is presented, and work efficiency cannot be improved. Further, in Japanese Patent Application Laid-Open No. 11-143892, there is no consideration in terms of time and no consideration is given to dynamic information sources. Furthermore, in JP-A-11-143796, the target is limited to a mailing list,
In addition, it is only for a single mailing list, and it is difficult to retrieve information by extracting topics from a plurality of information sources.

The present invention has been made in order to solve the above technical problems, and an object of the present invention is to freely combine a plurality of information sources and become a topic from there. Displaying information in an easy-to-understand form. Another object is to obtain a result of clustering according to the user's interest.

[0010]

SUMMARY OF THE INVENTION With this object in mind, the present invention provides a method for periodically observing a plurality of dynamically changing information sources (referred to as URLs or the like) obtained from the Internet or the like. , Automatically extract more important topics from the extracted information elements in consideration of the support relationship between sites, the degree of personal interest, etc., and collectively visualize them in an easy-to-understand manner . That is, the present invention is an information organizing method for organizing information from information sources connected via a network, comprising an information collecting step of periodically patrol a plurality of registered information sources to collect information. A word selecting step of selecting words that are topic elements from the collected information; a clustering step of performing clustering on a set of selected words; and a clustering step based on a result of the performed clustering. And a display step of displaying, as a representative keyword of the cluster, a main keyword from a set of words in each cluster, while displaying the information elements of the cluster based on the time axis.

This display step is characterized by displaying, from the information elements in each cluster, supplementary information based on the keywords contained in the text portion. When a plurality of words can be reduced to one, the method further includes a reduction step of using the reduced form as one reduced expression, and the display step displays the reduced expression newly appearing in each cluster as supplementary information. It is characterized by:
According to these inventions, it is preferable in that the obtained information can be visualized and expressed more easily for the user.

Further, if the word selecting step is characterized by selecting a newly appearing word with a high weight, the newly issued word can be provided to the user preferentially. Further, in the word selecting step, for a specific information source from which a specific word has been selected, a word which is a topic element is selected in consideration of support from other information sources among a plurality of information sources at a word level. The feature is that it is possible to select the topic information and provide it to the user.

[0013] When the present invention is grasped from another viewpoint, the information organizing method to which the present invention is applied accepts a user's registration of an information source from which information is to be obtained and a word of interest to the user, and registers the registered information. Information elements are obtained by periodically patrol the information source, and words are selected with increasing importance from the obtained information elements to words considered to be of interest to the user. Clustering is performed on a set of information elements having words, and the clustered information elements are displayed together with the cluster result. Furthermore, it is possible to determine the degree of interest in the individual information source by the user, and select words with increased importance for words that appear in the determined information source with a high degree of interest. . As the determination of the degree of interest by the user, for example, in addition to the designation of a specific site by the user, for example, it is possible to determine that the site whose information element corresponding to the user has been selected in the past has a high degree of interest. It is.

From another viewpoint, the information organizing method to which the present invention is applied is such that a plurality of sites from which information is to be obtained are registered, and the registered plurality of sites are periodically visited.
For example, by collecting information from multiple sites visited by examining changes in content due to words etc. newly added in a specified period, information collected from specific sites can be An important topic can be extracted in consideration of support from other sites in the site. In addition, clustering is performed on the extracted information elements having an important topic, and the obtained information elements are displayed together with the clustering result. The display of the result of the clustering includes, for example, displaying a time-series display for each cluster, displaying a representative keyword and supplementary information of each cluster, and the like. Further, the amount of topics provided by each site is calculated based on the number of extracted information elements, and an index indicating the topic supply capability of the site is accumulated based on the calculated amount of topics. For example, it is preferable in that it can be used for weighting sites and words based on the accumulated topic supply capability. Furthermore, as an application of this, the sites are arranged in the order of the topic supply ability index,
In addition, it is possible to display the numerical value.

[0015] On the other hand, in order to achieve the above object, an information processing apparatus to which the present invention is applied includes a designating means for designating a plurality of sites to be visited, a storage means for storing the designated sites, An information collection unit that periodically visits a plurality of stored sites to collect information, a word selection unit that selects words that are topic elements from the collected information, and a set of selected words. Clustering means for performing clustering on the cluster, and output means for outputting a keyword present in a set of words in each cluster together with information elements in each cluster based on a result of the performed clustering. can do.

Here, the output means outputs the information elements in each cluster in chronological order and outputs supplementary information using a keyword included in the text part of the information element. It is excellent in that it can easily output how the topic has changed. Note that this output means may be configured to output electronic information to a terminal connected via a network, in addition to the mode of displaying on a display device.

From another point of view, the information processing apparatus to which the present invention is applied includes a registration accepting unit for accepting a user to register an information source from which information is to be obtained and a word of interest to the user; Means for periodically circulating information sources to obtain information elements, and selecting words from the obtained information elements with increasing importance for words considered to be of interest to the user Selection means, clustering means for performing clustering on a set of information elements having the selected words, and display means for displaying the clustered information elements together with the result of the cluster. Can be. Setting means for setting the information source registered by the user or the information element corresponding to the user to be higher in importance than the information source selected in the past; Therefore, it is possible to increase the degree of importance for words appearing in the information source set to have a high degree of importance and select words.

On the other hand, according to the present invention, a program to be executed by a computer is input to a computer by input means (for example, a CD-R).
On a storage medium (for example, a CD-ROM or the like) readable by an OM driver or the like, this program executes a process of periodically circulating through a plurality of registered information sources and collecting information. A process of selecting words that are topic elements from among the above, a process of performing clustering on a set of selected words, and, based on a result of the performed clustering, an information element in each cluster based on a time axis. And displaying the main keyword from a set of words in each cluster, for example, as a representative keyword of the cluster, as a predetermined keyword. And displaying the supplementary information based on the keyword included in the text portion from the information element in each cluster by using a degenerated expression newly appearing in each cluster. This is preferable because it is possible to provide a display that is easy to understand.

According to the present invention, there is provided a storage medium in which a program to be executed by a computer is stored so as to be readable by an input means of the computer. Process of periodically traversing the site, collecting information from multiple sites traversed by examining changes in content, and supporting the collected information from other sites at the word level And causing the computer to execute a process of extracting an important topic in consideration of the above.

Further, the present invention provides a program transmission comprising storage means for storing a program to be executed by a computer, and transmission means for transmitting the program stored in the storage means to a user terminal via the Internet or the like. The program stored in the storage means is a device that periodically traverses a plurality of registered information sources to collect information, and a word that is a topic element from the collected information. A process of selecting, a process of performing clustering on a set of selected words, and displaying information elements in each cluster on a time axis based on a result of the performed clustering, and Displaying a predetermined keyword from the set, and transmitting the keyword by the transmitting means. Can.

[0021]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below in detail based on an embodiment shown in the accompanying drawings. First, before the description of the system configuration in the present embodiment, an outline of the system will be described to facilitate understanding of the system.
FIG. 1 is a diagram showing an outline of an information extraction / display method according to the present embodiment. In this method, individuals can freely select information sources and further organize them by freely assigning importance according to the degree of interest to them, and their own information site (Personal Portal) or A site (Vertial Portal) dedicated to a specific field is automatically realized. For this purpose, the user first registers a favorite site (step 101). When registering, for example, its name and its reference (URL: Uniform Res
ource Locators). Next, the system periodically traverses the registered site at the designated time, and compares the contents with those registered in the database. If the contents are different, it is registered as a new version and metadata is created (step 102). This metadata is obtained by extracting elements for selecting information from the contents referred to by the URL.

Next, at each site in the set of registered sites, the keywords appearing in the version immediately before the designated period and in the version in the designated period are counted, and the set of keywords is weighted. To extract new words (keywords) (step 1
03). Thereafter, the set of selected keywords is clustered using the inclusion relation of the information element set including the individual keywords and the assigned weight (step 104). This clustering can be said to be an operation of dividing into a set that is meaningful from some viewpoint. Then, based on the result of the clustering, the main representative keyword (hot word) of the keyword set of each cluster is displayed, the information element set is displayed in chronological order, and the clustering is performed using the keyword (subword) as supplementary information. The result is displayed (step 105). Through this series of processes, we can automatically extract more important topics in consideration of the support relationship between sites, the degree of personal interest, etc.
In addition, it is possible to collectively visualize them collectively. Then, for the cluster extracted in this way, a topic supply capability index, which is an index indicating the ability of each site to provide a topic, is calculated based on the importance of the keyword (step 10
6). Thus, a site having a high topic supply capability or a site having a higher topic supply capability for a specific word can be presented using the importance calculated at the time of topic extraction.

Next, the present method will be described in further detail using a system configuration. FIG. 2 is a diagram for explaining the overall configuration of the system according to the present embodiment. This system is a personal computer connected to the Internet 10
It is executed as a processing program of application software by a (PC) or the like. Moreover, it is also possible to configure as a server that provides information to a user's PC terminal connected to the Internet 10. The output by the processing program is configured to be provided to the user's PC terminal via the Internet 10 in the case of a server, in addition to being displayed on the display at the user's PC terminal.
In the present embodiment, the flow of processing in the user's PC terminal is mainly described. Further, a processing program for executing this system is generally stored in a hard disk drive (not shown), and is loaded into a main memory (not shown) at the time of execution, and is generally processed by a CPU (not shown). is there. The processing program may be supplied to a user's PC terminal or the like via a storage medium such as a CD-ROM (not shown), or may be downloaded by the user via the Internet 10, for example. The form provided is also conceivable.

In FIG. 2, reference numeral 11 denotes a registered site DB (database) for storing sites registered by the user;
Reference numeral 2 denotes a metadata DB that stores the above-described metadata, and 1
Reference numeral 3 denotes a site topic supply capability DB that stores the importance of the site obtained by calculation from the importance of the keyword, and 17 denotes a user-designated weighting DB that stores a favorite keyword designated by the user or the importance of the site. Uses a part of storage means such as a hard disk drive provided in a PC, for example. Reference numeral 14 denotes a crawler that automatically traverses registered sites from the Internet 10. Reference numeral 15 denotes a DBMS (database management system) with a version management function for storing and managing metadata of registered sites, and is an HTML (Hypertext).
An information element is extracted from the Markup Language, a text part thereof is analyzed, and a metadata generation mechanism 20 is provided for storing a keyword included therein and a classification thereof.
Reference numeral 16 denotes a metadata access method that provides access to data stored in the metadata DB 12. Reference numeral 30 denotes a new information extraction and display mechanism that extracts and displays a new topic based on information accumulated in the metadata DB 12.

The sites registered in the registration site DB 11 are sites registered according to the user's preference as described above. When registering, the user specifies its name and its reference (URL). FIG. 3 shows an example of a registered site. In the example shown in FIG. 3, four sites are registered, and the registration format is XML (extensible).
Markup Language). It should be noted that, for example, a method in which a directory list of a specific portal site is cut and pasted and registered is a simple operation for the user.

The crawler 14 periodically visits sites registered in the registered site DB 11 at a designated time. For example, the tour goes around at 7:30 am every day.
You may visit the entire designated site at the same time,
It is also possible to specify different times for individual sites. The DBMS 15 with the version management function manages the content as a new version when the content is different at the time of patrol by the crawler 14, creates metadata for the new version by the metadata creation mechanism 20, and saves the result in the metadata DB 12. ing. Thus, when a new version of your site is created,
The metadata is created. As described above, the metadata is obtained by extracting elements that carry information from the contents referred to by the URL. It has a link and its text portion, or a continuous text portion. Attribute extraction is applied to the text portion of these information elements, and keywords and their classifications are extracted.

FIG. 4 shows the configuration of the metadata creation mechanism 20 in more detail. This metadata creation mechanism 20
Creates metadata from an input file such as HTML and outputs it as an output file, as shown in FIG. Reference numeral 21 denotes an information element extraction mechanism, which is HTML
By analyzing the contents such as, the information elements (links, text, etc.) are extracted. An attribute extraction mechanism 22 extracts a keyword from the text of the information element extracted by the information element extraction mechanism 21 and assigns a category to the keyword. The attribute extraction mechanism 22 includes a morphological analysis mechanism 23, a keyword extraction mechanism 24, and a keyword classification mechanism 25. The morphological analysis mechanism 23 divides the text part of the information element extracted by the information element extraction mechanism 21 into words. The keyword extracting mechanism 24 extracts only those that become keywords from the word strings resulting from the division by the morphological analyzing mechanism 23. The keyword classification mechanism 25 has a function of assigning a classification of the keyword extracted by the keyword extraction mechanism 24.

FIG. 5 is a diagram showing an example of a link as an example of the created metadata. FIG. 6 shows an example of a text block as an example of the created metadata. In FIG. 5, the expression in the HTML file in the case of a link is indicated by an a tag using a tag indicating a link destination, and the extracted information element is configured by an anchor tag. In FIG. 6, the expression in the HTML file in the case of the text block is a text expression, and the extracted information element is configured by a text tag. With the above processing, if there is a change during the patrol by the crawler 14 in the site registered in the registration site DB 11, all the contents and the metadata created from the metadata by the metadata creation mechanism 20 are stored in the metadata. Registered in DB12. In addition, the date and time when the content has been changed (the date and time when the update date and time is obtained from the web server, and the date and time when the content has been visited when the update date and time are not obtained) are similarly stored in the metadata DB 12.

Next, a new word is extracted and clustered by the new information extracting and displaying mechanism 30. FIG.
FIG. 3 is a diagram for explaining the configuration of the new information extraction and display mechanism 30. In the figure, reference numeral 31 denotes a keyword statistical mechanism. From the metadata for a designated site obtained from the metadata DB 12, a keyword included in an information element newly appearing in a version within a designated period and a designated keyword are designated. And the keywords included in the information elements included in the version immediately before the specified period. Judgment as to whether an information element has newly appeared can be made based on whether a link having a different URL has appeared or if the same URL already exists but its corresponding text is different. It is determined to be a link. For a text block, it is determined by whether a different text has appeared. 32
Is a keyword importance calculation mechanism that assigns importance to extracted keywords. The keyword importance calculation mechanism 32 can set the importance in consideration of the importance of the site with reference to the site topic supply ability DB 13. Reference numeral 33 denotes a clustering mechanism, which performs clustering using the extracted keywords with importance. For the extracted cluster, the importance is calculated based on the importance of the keyword as described later, and the result is stored in the site topic supply capability DB 13.
Is stored in Reference numeral 34 denotes a clustering result display mechanism, which has a function of displaying a result of clustering.

FIG. 8 is a diagram showing the relationship between the designated period and the version. In the keyword statistical mechanism 31 shown in FIG. 7, at each site in the set of sites registered in the registration site DB 11, a version immediately before the specified period and a version during the specified period shown in FIG. 8 appear. Keywords are counted. Here, the previous version (Version N
-3) contains the count (Fs (w)) and subsequent versions
(Fn included in (Version N-2 to Version N))
(w)). The keyword importance calculation mechanism 32 determines whether the keyword set is a new keyword by weighting these keyword sets. As a sorting method, for example, a method in which weights such as the importance of a word and the importance of a site are used alone or in combination, and those in which the weight is equal to or less than a threshold can be considered.

The following study examples can be considered as the importance of words. (a) Consider the ratio of simple new words (Fn (w) / (Fs (w) + Fn (w))). (b) Calculate the information amount of keywords in the past version (all versions before Version N-3), and reduce the importance of keywords with low information amount. This makes it possible to reduce the importance of words that are always added to individual information, such as “release” in “new product release information”. (c) Consider whether a word is included in multiple sites (is supported by multiple sites). (d) Weighting by the user is performed. That is, the user registers a word of particular interest (or not interest) along with the importance, and when it appears, increases the importance (lower)
I do. The specified method is a method in which the user explicitly describes the importance for each site, or when the corresponding information element is selected when displaying the finally displayed clustering result For example, a method of increasing the weight of a site including the information element can be considered.

As the importance of a site, there is a method based on the degree of importance of a user to each site. For example, a user registers a site of particular interest (or not interest), and words appearing on that site are given higher (lower) importance. The specified method is a method in which the user explicitly describes the importance for each site, or when the corresponding information element is selected when displaying the finally displayed clustering result There is a method of increasing the weight of the site including the information element.

Next, clustering of the selected keyword set will be described. In the clustering mechanism 33 shown in FIG. 7, the keyword set selected by the keyword statistical mechanism 31 is clustered using the weight given by the keyword importance calculation mechanism 32. Any method may be used for this clustering. However, as a pre-processing of the clustering, if a plurality of keywords include exactly the same keyword set and the keywords can be reduced to one, the keywords are reduced. Are used as one keyword.

Here, the degeneracy includes, for example, the following. -Orthography Convert to an orthography using an orthography dictionary. Notation fluctuation such as "computer" and "computer"
→ Convert to the original "computer". -Synonyms Convert to regular expressions using a synonym dictionary. "United States", "United States"->"UnitedStates"-Compound Words that appear adjacently as compound words in all texts are converted into one compound word. "Obuchi", "Prime Minister" → "Prime Minister Obuchi"-Dependency structure Convert words that have the same dependency in all texts into one expression. If a case marker is obtained, it is also added. The case marker corresponds to a particle or the like in Japanese and a preposition or the like in English. In the following example, the particle “GA” is added as a case marker. "Cabinet", "General resignation" → "Cabinet resignation"

Next, an example of clustering will be described.
Here, the keywords selected first are sorted in order of importance. Then, an information element including the keyword is assigned to each keyword. Thereafter, the inclusion relation (strong inclusion relation and weak inclusion relation) is determined. In determining this inclusion relation, it is assumed that each keyword is always included in a keyword having high importance. In the determination of the inclusion relation, the presence / absence of the inclusion relation is checked for all keywords having higher importance than all the keywords. The presence / absence of the inclusion relationship is determined to be a strong inclusion relationship when the information elements associated with the keywords are viewed as a set and the proportion of the common elements is greater than a threshold. If there is some common information element but the ratio is less than the threshold value, it is assumed that the information element has a weak inclusion relation. Keywords deemed to have a strong inclusion relationship are grouped into one cluster. Keywords that are deemed to have a weak containment relationship form another cluster. Here, the information element set included in the weak inclusion relation is excluded from the information element set of the cluster associated with the keyword of higher importance. In addition, the keyword is added to a keyword set of a cluster having higher importance.

FIG. 9 shows an example of the structure of the result obtained as a result of such clustering and its interpretation. In the example shown in FIG. 9, keyword 1 has a strong inclusion relationship with keyword 2 and keyword 3. There is also a strong inclusion relationship between keyword 4 and keyword N-1. The keyword 4 has a weak inclusion relationship with the keyword 3. As a result of the clustering, a set of cluster 1, cluster 2, and cluster m is formed. As a keyword set of the cluster 1, keywords 1 to 3 having a strong inclusion relation are collected, and keywords 4 having a weak inclusion relation are also collected as a supplement. On the other hand, in the information element set, information element sets 1 to 3 corresponding to keywords 1 to 3 having a strong inclusion relation are grouped, and information element set 4 is excluded. Since the information element set 4 outputs a text in a full state, the information element set having a weak inclusion relation in the sense of reducing the amount of information is excluded.

FIG. 10 is a diagram showing a specific example in which clustering has been performed. Here, 3 of clusters 1-3
One cluster is shown, and a keyword set and an information element set are formed respectively. The cluster 2 and the cluster 3 have a weak inclusion relation with the cluster 1.

Next, the display of the result of the clustering will be described. The clustering result display mechanism 34 shown in FIG. 7 displays, as a representative keyword (hot word) of the cluster, a main keyword (keyword with the highest importance) of the keyword set of each cluster from the result of the clustering described above. . Further, the information sets are displayed in chronological order from the information element sets included in the cluster. At this time, the supplementary information is displayed as a subword using the keyword included in the text part of the information element. This supplementary information is displayed when a single reduced expression or a plurality of keywords or reduced expressions of a keyword included in the keyword set of the cluster first appear. The display order of the keywords and the reduced expressions is the same as the order of appearance in the text.

Referring to the specific example shown in FIG. 10, when displaying the cluster 1, the oldest information element is displayed first. The information elements “development tools, e-commerce, operating systems, databases, Lotus products, and networks” include one of the keywords
Since only one keyword, "database", is included, no subword is displayed. The next information element, "XML Access Service Lightweight Extractor (XLE) extracts data from the database and extracts the data using a set of relational tables stored in a JDBC compliant relational database management system (DB2, Oracle, etc.) "Convert and assemble data into XML document." Includes "database" and "DB" in the keyword set. Here, since a plurality of keywords are included, a subword is created using these keywords. Since the display order is displayed in the order of the text in the information element set, "D
B, database ". If these keywords appear consecutively in the text, they are displayed in the reduced expression “DB database” (without commas). This subword is stored, and if only “database” and “DB” are included in the display of cluster 1, it is not displayed again.

Next, if there is a cluster that has a weak inclusion relation with the cluster, it is displayed. In displaying a hot word, "stepping" is performed to indicate that there is an inclusion relationship. The display of the subword is performed in the same manner. In this way, all clusters are displayed. The clusters having a weak containment relationship and the hot words of the clusters having a weak containment relationship are displayed “staged” by the number of levels.

FIG. 11 is a diagram showing a display example obtained by a series of these processes. In the display example shown in FIG. 11, a hot word 51 is displayed at the leftmost end, and a subword 52 is displayed next to it. In addition, as can be understood from the date 53, the oldest information elements are displayed.
In the reference article 54, a text block and a link text indicated by an underline are displayed as information elements. Furthermore, the "version" which has an inclusive relation to the keyword "database" in the first row of the hot word
It can be understood that “DB” and “DB” are displayed “stepped down” and one step lower. As described above, in this embodiment, clustered results are displayed in chronological order, and in addition to the main keywords (hot words) of each cluster, newly appearing degenerated expressions are displayed as supplementary information (subwords). , The corresponding information elements are displayed in chronological order. As a result, it is possible to provide the user with new and desired information in an organized state.

Finally, in the present embodiment, the topic supply ability index is calculated. That is, importance can be calculated for the clusters extracted in this manner based on the importance of the keyword. The importance obtained as a result is additionally stored in the site topic supply ability DB 13, updated, and used for calculating the importance of the site. At that time, the latest situation is reflected as much as possible by decreasing the past value. More specifically, the amount of topics provided by each site is calculated by combining the number of words and information elements included in the extracted clusters, or the weight thereof, and the topic supply capability of the site is calculated based on the calculated amount. It is accumulated as an indicator to show. The words included in the cluster are also accumulated as the topic supply ability index for each word in each site. Also, by arranging the sites in the order of the topic supply ability index assigned to each site, or displaying the numerical value, the user is presented with how much new information the site has provided. Furthermore, it is also possible to present an information supply ability index for a specific word in each site by using a topic supply ability index for each word assigned to each site. Furthermore, by displaying a site that matches the keyword specified by the user with respect to the word set with the topic supply capability index assigned to each site, a site having a high topic supply capability for the keyword desired by the user is presented. It is also possible.

As described above, according to the present embodiment, by freely combining a plurality of information sources and extracting information of a topic therefrom, not a single information source but a set of information sources You can get the information that has become a topic. In other words, a more important topic can be extracted by registering a plurality of sites, periodically patrol the sites, and examining changes in the contents. In addition, since the weight for a word is changed by considering the support of a plurality of information sources, the result of clustering is changed, and a more general cluster can be obtained within the site set. That is, by considering support from other sites at the word level, more important topics can be extracted. Similarly, by changing the degree of the user's interest in words and sites, it is possible to obtain a result of clustering according to the user's interest. Further, for example, by displaying the text obtained using the supplementary information together with the result of the clustering, it is possible to easily display how the extracted individual topics have changed.

[0044]

As described above, according to the present invention,
It is possible to freely combine a plurality of information sources and to display information that is a topic from the information sources in an easily understandable manner.

[Brief description of the drawings]

FIG. 1 is a diagram showing an outline of an information extraction / display method according to the present embodiment.

FIG. 2 is a diagram for describing an overall configuration of a system according to the present embodiment.

FIG. 3 is a diagram showing an example of a registered site.

FIG. 4 is a diagram further illustrating the configuration of a metadata creation mechanism 20 in further detail.

FIG. 5 is a diagram showing an example of a link as an example of created metadata.

FIG. 6 is a diagram showing an example of a text block as an example of created metadata.

FIG. 7 is a diagram for explaining a configuration of a new information extracting and displaying mechanism 30.

FIG. 8 is a diagram showing a relationship between a designated period and a version.

FIG. 9 is a diagram showing an example of the structure of a result obtained as a result of clustering and its interpretation.

FIG. 10 is a diagram showing a specific example in which clustering is performed.

FIG. 11 is a diagram showing a display example obtained by a series of these processes.

[Explanation of symbols]

10 Internet, 11 Registration DB, 12
Metadata DB, 13 ... Site topic supply capability DB, 14
... Crawler, 15 ... DBMS with version control function,
16: Metadata access method, 17: User-designated weighting DB, 20: Metadata creation mechanism, 21: Information element extraction mechanism, 22: Attribute extraction mechanism, 23: Morphological analysis mechanism, 24: Keyword extraction mechanism, 25: Keyword classification Mechanism, 30: new information extraction and display mechanism, 31: keyword statistical mechanism, 32: keyword importance calculation mechanism, 33 ...
Clustering mechanism, 34 Clustering result display mechanism, 51 Hot word, 52 Subword, 53 Date, 54 Reference article

Continuation of the front page (72) Inventor Hiroshi Nomiyama 1623-14 Shimotsuruma, Yamato-shi, Kanagawa Prefecture IBM Japan, Ltd. Tokyo Basic Research Laboratory F-term (reference) 5B075 NK31 NR03 NR12 PQ02 PQ34 PR08

Claims (19)

[Claims] 1. An information organizing method for organizing information from information sources connected via a net, comprising: an information collecting step of periodically patrol a plurality of registered information sources to collect information; A word selecting step of selecting a word that is a topic element from the collected information; a clustering step of performing clustering on the set of the selected words; based on a result of the performed clustering, Displaying an information element in each cluster based on a time axis and displaying a main keyword from a set of words in each cluster as a representative keyword of the cluster. Method. 2. The information organizing method according to claim 1, wherein said displaying step displays supplementary information based on a keyword included in a text portion of the information element in each of the clusters. 3. When the plurality of words can be reduced to one, a reduced step of using the reduced word as one reduced expression is further included. The displaying step further includes the reduced expression newly appearing in each of the clusters. The information arrangement method according to claim 2, wherein the information is displayed as the supplementary information. 4. The information organizing method according to claim 1, wherein said word selecting step selects a newly appearing word with a higher weight. 5. The word selecting step includes, for a specific information source that has selected a specific word, a word that becomes a topic element in consideration of support from another information source among the plurality of information sources at a word level. 2. The information sorting method according to claim 1, wherein the information is sorted. 6. A user receives registration of an information source from which information is to be obtained and a word of interest to the user, and periodically visits the registered information source to obtain an information element. From the information elements that have been selected, words with increasing importance are selected for words considered to be of interest to the user, clustering is performed on a set of the information elements having the selected words, and clustering is performed. An information organizing method characterized by displaying the selected information element together with the result of the cluster. 7. A method for determining a degree of interest in an individual information source by a user, and selecting words that appear in an information source having a high degree of interest and selecting words with increasing importance. The information organizing method according to claim 6. 8. Registering a plurality of sites from which information is to be obtained, periodically patrol the plurality of registered sites, and collecting information from the plurality of visited sites by examining changes in the contents. An information organizing method, wherein an important topic is extracted from information collected from a specific site at a word level in consideration of support from other sites in the plurality of sites. 9. The information organizing method according to claim 8, wherein clustering is performed on the extracted information elements having the important topic, and the obtained information elements are displayed together with a result of the clustering. 10. Calculating the amount of topics provided by each site based on the number of extracted information elements, and accumulating an index indicating the topic supply capability of the site based on the calculated amount of topics. 9. The information organizing method according to claim 8, wherein: 11. A designating means for designating a plurality of sites to be visited, a storage means for storing the plurality of sites designated by the designating means, and a method for periodically storing the plurality of sites stored in the storage means. Information collecting means for circulating and collecting information; word selecting means for selecting words that are elements of topics from the information collected by the information collecting means; and a set of words selected by the word selecting means. Clustering means for performing clustering on, and output means for outputting, based on a result of the clustering performed by the clustering means, a keyword present in a set of words in each cluster together with information elements in each cluster. An information processing apparatus comprising: 12. The information processing apparatus according to claim 11, wherein the output unit outputs information elements in each of the clusters in chronological order, and outputs supplementary information using a keyword included in a text portion of the information element. Information processing device. 13. The information processing apparatus according to claim 11, wherein said output means outputs to a display device or a terminal connected via a network. 14. A registration accepting unit for accepting registration of an information source from which information is to be obtained and a word of interest to the user by the user, and periodically patrols the information source accepted by the registration accepting unit. Circulating means for obtaining an information element by means of: selecting means for increasing the importance of a word considered to be of interest to a user from among the information elements obtained by the circulating means, and selecting the word; Clustering means for clustering a set of information elements having the word selected by the means, and display means for displaying the information elements clustered by the clustering means together with the result of the cluster. Information processing device. 15. A setting means for setting an information source registered by a user or an information element corresponding to the user to be higher in importance of an information source with respect to an information source selected in the past; 15. The information processing apparatus according to claim 14, wherein a word that appears in an information source set with a high degree of importance by the setting unit is selected with a higher degree of importance. 16. A storage medium storing a program to be executed by a computer in a manner readable by an input means of the computer, the program comprising: a process of periodically traveling a plurality of registered information sources to collect information; A process of selecting a word that is a topic element from the collected information; a process of performing clustering on the set of the selected words; and a process for selecting each cluster based on a result of the performed clustering. And displaying the information elements on the basis of the time axis and displaying a predetermined keyword from a set of words in each of the clusters. 17. A process for displaying supplementary information based on a keyword included in a text portion from the information element in each of the clusters using a degenerated expression newly appearing in each of the clusters. 16
The storage medium according to the above. 18. A storage medium storing a program to be executed by a computer in a manner readable by an input means of the computer, the program comprising: a process of registering a plurality of sites from which information is to be obtained; Regular site patrol, collecting information from multiple sites visited by examining changes in content, and providing word-level support for collected information from other sites A storage medium for causing the computer to execute a process of extracting an important topic in consideration of the processing. 19. A program transmission device comprising: a storage unit for storing a program to be executed by a computer; and a transmission unit for transmitting the program stored in the storage unit, wherein the transmission unit stores the program stored in the storage unit. The program includes a process of periodically traveling a plurality of registered information sources to collect information, a process of selecting a word that is a topic element from the collected information, and a process of selecting the selected word. Based on the process of performing clustering on the set and displaying the information elements in each cluster on the time axis based on the result of the clustering performed, a predetermined keyword is selected from the set of words in each cluster. A program transmission device comprising: a display unit for displaying the program;