JP2001134600A - System and method for information extraction and recording medium stored with recorded program for information extraction - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately extract information matching a purpose from a set of documents having different contents and formats. SOLUTION: A document classifying means 1 classifies the document set into specified categories. An information extracting means 3 judges which information is extracted according to the classifications of the documents and extracts information from the documents. An information classifying means 5 classifies the extracted information. A result selecting means 7 selects only necessary information from the extracted and classified information. A result output means 186 divides and outputs the information according to the classification result of the information.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information extraction system for extracting specific information from a document, an information extraction method, and a recording medium on which an information extraction program is recorded (hereinafter, referred to as an information extraction system). Classify documents,
The present invention relates to an information extraction system that can change information to be extracted according to a result of the classification.

[0002]

2. Description of the Related Art Conventionally, an example of this type of information extraction system is described in Japanese Patent Application Laid-Open No. 8-329165.
FIG. 23 shows a flowchart of the operation of this conventional information extraction system. The operation of this conventional technique will be described below with reference to FIG.

When a document to be processed is input (4801
To 4803), extract a character string having a specific pattern from the document as numerical data (4804), and further extract a character string based on a certain rule before and after the numerical data as numeric character string data ( 4805). Noun data is extracted from the numeric character string data, classified for each predetermined item, and associated with the numeric data (4806). The numerical data of each item obtained in this way is tabulated for each predetermined period such as one month (4807), and the tabulated result data is displayed.
(4808).

[0004] By using this conventional system, it is possible to extract the amount spent for shopping from a document in a specific format such as a household account book.

[0005] Next, as a conventional system for extracting information from a document of any format, a dissertation titled "Development of a proper noun extraction system and evaluation in IREX-NE" (Preprints of the IRX-NE Workshop, pp.171-178) , 1999). In this system, a place name, a person name, a price, and the like can be extracted from a document.

[0006] One of the information services provided on the World Wide Web is to collect employment information and present information, and manually extract information such as company names and prizes on the homepage. It has been done recently.

[0007]

SUMMARY OF THE INVENTION The first of the prior arts described above.
The problem is that information cannot be extracted with sufficient accuracy from a set of documents that vary in content and format. The reason is that the conventional information extraction system assumes a document of a specific format.

A second problem is that unnecessary information is also extracted. The reason is that what kind of information is needed depends on the purpose of creation and use of each document. The necessity depends on the context. Therefore, if information is extracted from all documents or the entire document, unnecessary information is included.

[0009] The third problem is that if manual work is introduced to compensate for the first and second problems, it is impossible to cope with a large number of documents, or very much time and cost will be required. That is to say.

SUMMARY OF THE INVENTION It is therefore an object of the present invention to provide an information extraction system capable of extracting information according to a user's purpose of use from an irregular document set in view of the above-mentioned problems of the related art.

Another object of the present invention is to provide an information extraction system for easily selecting a required document.

[0012]

An information extracting system according to the present invention comprises a document classifying means for classifying a document, and information extracting means capable of changing the information to be extracted and the extracting method in accordance with the classification of the document. , And operates to extract information according to the classification result of the document.

[0013] Further, there is provided an information extracting means capable of classifying the information extracted according to the classification of the document.

By adopting such a configuration and extracting only necessary information accurately from each document, the object of the present invention can be achieved.

[0015]

Next, embodiments of the present invention will be described in detail with reference to the drawings.

(First Embodiment) [Explanation of Configuration] Referring to FIG. 1, a first embodiment of the present invention comprises a document classification unit 1, an information extraction control unit 2, a plurality of information extraction units. It comprises means 3, information classification control means 4, a plurality of information classification means 5, and result output means 6.

There are a plurality of information extracting means 3 corresponding to the types of the document classifications.
32 and extracted knowledge storage means 33.

There are a plurality of information classifying means 5 according to the type of document classification.
Including 52.

Each of these means operates as follows.

The document classification means 1 receives a plurality of documents and classifies the documents into designated categories.

The information extraction control means 2 determines the information extraction means 3 to be used according to the document classification result of the document classification means 1, and causes the determined information extraction means 3 to extract information from the document. The information extraction means 3 operated by the information extraction control means 2 extracts information from the document specified by the information extraction control means 2.

The information classification control means 4 determines the information classification means 5 to be used in accordance with the classification result of the document, and causes the determined information classification means 5 to classify the information extracted from the document. The information classification means 5 operated by the information classification control means 4 classifies the information specified by the information classification control means 4.

The result output means 6 outputs a document classification result, extracted information, and information classification result.

Here, the information extraction control means 2 and the information extraction means 3 will be described in detail with reference to FIG.

A plurality of information extracting means 3 are prepared in accordance with the number of document classification categories. In FIG. 2, three information extracting means 3a, 3b, 3c are described, but this does not limit the number of information extracting means to three.

The information extraction control means 2 selects the information extraction means 3 according to the classification result of the document classification means 1.

The information extracting means 3 comprises an extracting executing means 31, an extracting information defining means 32, and an extracted knowledge storing means 33.

The extraction executing means 31 refers to the extracted information defining means 32 and the extracted knowledge storing means 33 and extracts information corresponding to the category of the document from the document. Extraction information definition means 32
Stores the type of information corresponding to the classification category of the document. The extracted knowledge storage means 33 stores a method for determining each piece of information stored in the extracted information definition means 32.

The information classification control means 4 and the information classification means 5 will be described with reference to FIG.

A plurality of information classifying means 5 are prepared in accordance with the number of document classification categories. In FIG. 3, three information classifying means 5a, 5b, 5c are described, but this does not limit the information classifying means to three.

The information classifying means 5 comprises a classification executing means 51 and a classification knowledge storing means 52.

The classification executing means 51 classifies the information extracted by the information extracting means 3 with reference to the classification knowledge storing means 52.
The classification knowledge storage means 52 stores rules for classifying information for each type of extracted information.

[Explanation of Operation] Next, the overall operation of the present embodiment will be described in detail with reference to the flowcharts of FIGS.

First, documents are classified into categories by the document classifying means 1 (step 401 in FIG. 4). As a classification method, a type classification technology (Japanese Patent Application No. 10-200171) utilizing the characteristics of a Web page filed by the present applicant before filing the present invention is proposed.
Or a classification technique using word information can be used, but the configuration of the document classification means 1 is not limited. Various existing technologies can be used as the configuration of the document classifying means 1, and those skilled in the art can fully understand. One document may be classified into a plurality of categories, or may not be classified anywhere.

Next, the information extraction control means 2 selects the information extraction means 3 corresponding to the document classification result in step 401, and classifies the document (step 402).

The information extracting means 3 extracts information from the document (Step 403). Each information extracting means 3 extracts information corresponding to the corresponding document category from the document. The extraction executing means 31 obtains information to be extracted from the document by referring to the extraction information defining means 32. Further, the extraction executing means 31 obtains a method for determining each piece of information to be extracted with reference to the extracted knowledge storage means 33. The methods for this determination include:-the format of the information to be extracted itself-item names and description positions in tables and bullets-notation methods of words and symbols that co-occur before and after the information to be extracted-directly described There are rules for estimating missing information, and a plurality of methods can be combined.

Next, the information classification control means 4 selects the information classification means 5 corresponding to the document classification result, and classifies the information extracted from the document (step 404).

The information classifying means 5 classifies the extracted information according to the corresponding document category (step 40).
Five). The classification executing means 51 classifies the information for each type of the extracted information with reference to the classification knowledge storage means 52. Information classification knowledge includes: ・ Dividing numerical and time information into several ranges ・ Preparing a correspondence table between words and classification ・ Using a dictionary with a hierarchical structure ・ Classification based on pattern matching of character strings Further, a plurality of methods can be combined.

Finally, the result is output (step 406).
Outputs the classification result of the document name and the extracted information. Further, the document category and the extracted information may be output.

Next, effects of the present embodiment will be described.

In the present embodiment, since documents are classified and necessary information is extracted according to the classification result, necessary information is obtained even when various types of documents are mixed. Only can be extracted.

Further, in this embodiment, since a method for extracting information is selected in accordance with the classification result of the document, highly accurate information extraction becomes possible.

Further, by referring to a document based on the extracted information, the user can easily find a necessary document.

[Example] Next, the operation of this embodiment will be described using a specific example.

In this embodiment, a Web page (HTML file)
A case where information is extracted from the information will be described as an example.

The Web pages to be classified are stored on a storage device in advance by an automatic collection program or a download program. Alternatively, a list of Web page URLs may be prepared and downloaded as needed.

First, the Web page is classified by the document classifying means 1.

In this embodiment, an example in which a structured document search system described in Japanese Patent Application No. 10-200171 is used as the document classification means 1 will be described. However, as described above, the document classification means 1 of the present invention is not limited to this example. In this system, several types are set for a document, and the relevance of each document and the type is calculated. By setting a reference value of the relevance, the document can be classified into a type having a relevance higher than the reference value.

FIG. 5 is a diagram showing an example in which the degree of conformity is calculated for a Web page. In FIG. 5, “x.html”, “y.htm”
"l" and "z.html" are page names. Domain names and directory names are appended to URLs, but they are omitted in the figure. "Job information", "events", and "presents" are the types that result in classification. By name, the numerical value is the relevance, and the higher the numerical value, the more the content contained in the document is compatible with the type. Here, assuming that the reference value of the goodness of fit in this example is 70, “x.html” is “event”,
“Y.html” is classified as “job information” and “z.html” is classified as “present”.

Next, the information extraction control means 2 selects the information extraction means 3 corresponding to the classification result of each page.
From pages classified as "job information", "event", and "present", information extraction means 3a, 3b, 3c respectively
To extract information.

FIG. 6 is a diagram showing an example of the type of information stored in each information name defining means 32 of the information extracting means 3. Extraction information definition means 32a, 3 of information extraction means 3a, 3b, 3c
2b and 33c store the information shown in 6a, 6b and 6c in FIG. 6, respectively. For example, when the information extracting means 3a has a role of extracting information from a document classified as "job information", the extracted information defining means 32a has the information shown in 6a of FIG. Extracts information according to the contents of the extraction information definition means 32a (in this example, [work location], "occupation type") from the document. Further, in the case of this example, the information extraction control means 2 needs information for judging which category of information is assigned to each of the plurality of information extraction means.

FIG. 7 is a diagram showing an example of a method of extracting information stored in the extracted knowledge storage means 33. (A), (b),
(c) corresponds to the extracted knowledge storage means 33a, 33b, 33c, respectively.

In FIG. 7, the extraction information defining means shown in FIG.
It describes how to find information for each of the 32 types of "information name". In FIG. 7, "description type" designates where information written is extracted. For example, in the case of "heading", a portion having a heading written in "pattern" in a table or a bullet point is extracted. In addition, if it is “specified tag”, a specific tag is extracted, and if it is “text”, a part matching the “pattern” in the text is extracted.

In the pattern shown in FIG. 7, "% d" or "% s" means a variable, meaning that an arbitrary character string is entered. By using a format that allows detailed specification such as the range of numeric values, the type and length of characters, etc., extraction can be performed more accurately. If there are multiple extraction methods, they are ordered by “priority”.

In the example of FIG. 7, the "job information" type "working place" and the "event" type "holding place" are both information relating to a place, but are extracted using different patterns. , Can be distinguished and extracted. Similarly, the “event” type “holding date” and the “present” type “application deadline” can be extracted separately, although they are dates.

Therefore, “y.html” classified as “recruitment information” in the example of FIG.
And "occupation type" are extracted, and the extraction is performed with reference to the extraction method 7a in FIG.

FIG. 8 is an example of a display image of a Web page classified as "job information". In the actual file,
Tables and bullets are represented by tags. A case where information is extracted from here will be described.

As for "work location", an item with a heading "work location" or "work location" or "sales office" is searched from the table or the list in FIG. 7 (a). Is extracted. Similarly, “system engineer” is extracted from the “occupation” section.

In the example of FIG. 5, “x.
As for “html”, “name”, “venue”, and “date” are extracted from 6b in FIG.

FIG. 9 is an example of a display image of a web page of “event”. From this page, extract the "name" and "venue" based on the bulleted headings. Since "Date" is not described in a bulleted list, a character string that matches the pattern is searched from the text in FIG.
October 10 "is extracted.

FIG. 10 is a diagram showing an example of information extracted from each page. The page name, classification result, information name, and extracted contents are described.

Note that not all information defined by the extraction information definition means 32 can be extracted. For example,
The “name” and “date” may be extracted from the document classified as “event” and the “venue” may not be extracted.

Next, the information classification control means 4 selects the information classification means 5 corresponding to the classification result of each page.
Information extracted from pages classified as “job information”, “event”, and “present” is classified by the information classifying means 5a, 5b, and 5c, respectively.

FIG. 11 is a diagram showing an outline of a classification method described in the classification knowledge storage means 52. Describe the type and method of classification based on each information. (A), (b), and (c) of FIG. 11 correspond to the classification knowledge storage means 53a, 53b, and 53c, respectively.

FIG. 12 is a diagram showing the classification result. FIG.
The classification of the date of “Event” in 2 is expressed by year and month with 6 digits. The “present” deadline is classified on a weekly basis, and is represented by a Sunday date of the week corresponding to the classification.

The classification executing means 51 classifies information according to the description of the classification knowledge storage means 52. As a method of classification, there is a method of using a list of words included in each classification, using a thesaurus in which the names of municipalities and facilities are described for each prefecture, and using a word thesaurus to put together a general concept.

By specifying a classification method for each type of information, information of the same date, such as “event date” of “event” type and “application deadline” of “present” type in FIG. Can also be re-classified.

Finally, the result output means 6 outputs the extracted information and the classification result.

As an output method, in addition to outputting the extracted information to a storage device in a format such as a CVS file format that can be registered in a search system or a database system, a list table as shown in FIG. You can use any structured document format such as XML, SGML, etc.

Note that the present embodiment is not limited to HTML,
Works similarly for structured documents such as XML.

[Another Example] Next, the operation of this embodiment will be described using another example.

Here, a case where information is extracted from newspaper articles will be described.

First, newspaper articles are classified into predetermined categories. A conventional system that classifies documents based on words contained in articles (for example, Just System CB
Classifier (trademark)) can be used. For example, articles are classified into "international politics", "new product information", and "sports".

Next, the information extraction control means 2 selects the information extraction means 3 corresponding to the classification of each article and extracts information.

FIG. 13 is a diagram showing an example of the types of information stored by the information name definition means 32. For example, "international politics"
The information of “place name” and “related person” is extracted from the page classified as (13a in FIG. 13).

FIG. 14 is a diagram showing an example of an information extraction method stored in the extracted knowledge storage means 23. Since a newspaper article is an unstructured text, information is extracted by pattern matching in the text. “Related parties” extracted from articles on “international politics” are limited to persons at the national representative level using titles (FIG. 14 (a)).

In FIGS. 14 (a) and 14 (c), lists of "place names" and "competition names" are created in advance to search for text.

FIG. 15 is a diagram showing an example of information extracted from newspaper articles. The article number, classification result, information name, and extracted contents are described.

Next, the information classification control means 4 selects the information classification means 5 corresponding to the article classification result, and classifies the information extracted from the article. FIG. 16 is a diagram showing an outline of a classification method described in the classification knowledge storage means 52. Describe the type and method of classification based on each information. FIG.
It is a figure showing a classification result.

Finally, the result output means 6 outputs the extracted information and the classification result.

(Second Embodiment) Next, a second embodiment of the present invention will be described.
An embodiment will be described in detail with reference to the drawings.

Referring to FIG. 18, according to a second embodiment of the present invention, a document classification unit 1, an information extraction control unit 2, a plurality of information extraction units 3, an information classification control unit 4, It comprises information classification means 5, result selection means 7, and result output means 6.

Next, the overall operation of the present embodiment will be described in detail with reference to the flowcharts of FIGS.

Operation of Document Classification Means 1, Information Extraction Control Means 2, Information Extraction Means 3, Information Classification Control Means 4, Information Classification Means 5
(Steps 1901-1905) are the same as in the first embodiment.

The result selecting means 7 selects only specific information from the extracted and classified information and transfers it to the result output means 186 (step 1906). The selection criteria include:-Specifying the classification of the document-Specifying the classification of the information-Specifying the conditions for the extracted information-Selecting only the information from the document from which the specific information was extracted.

The result output means 186 receives the output of the result selection means 5 and divides it into a plurality of pieces according to the result of classification of information and outputs it (step 1907). Examples of the division method include: • division for each document classification; • division for each information classification; and • division for each combination of a plurality of classifications.

Next, effects of the present embodiment will be described.

In this embodiment, since only the necessary information is selected and output by the result selecting means 7 from the extracted or classified information, the information extraction according to a specific purpose or a user is performed. Can provide results.

Further, in the present embodiment, the result output means 186 is configured to classify the extracted information, divide it, and output it. By combining the classification of the document and the classification of the extracted information, a document having specific information can be distinguished from others, and the purpose can be easily achieved when searching for a document with a specific purpose.

[Example] Next, the operation of this embodiment will be described using a specific example.

Here, information extracted from a Web page (HTML file) will be described as an example.

As a classification result by the information classification control means 4 and the information classification means 5, information in the format shown in FIG. 7 is obtained.

Next, the result selecting means 5 selects specific information from the result. As an example of the selection method: ・ Select the information extracted from the “Event” page ・ Select the information extracted from “Job information” where the “Work location” is Kanto ・ The “Date” is from the current date and time Select information extracted from the previous “Event” page. ・ Select information extracted from the “Present” page from which both “Prizes” and “Application deadline” have been extracted.

Finally, the result output means 184 receives the output of the result selection means 5 and divides the output into a plurality of pieces according to the result of classification of the information and outputs it. The division method is as follows:-Divide by page classification-Divide information on "Event" page by prefecture of "Place"-Divide "Event" page by month of "Date"-"Job information" The page information is divided according to the prefectural classification of “work location” and the classification combination of “occupation type”. In this case, the division may be performed in (47 prefectures × the number of job classifications).

Further, by using an output format in which page classification and extracted information classification are hierarchically combined, it is possible to efficiently find a document containing specific information. As a method of realizing the hierarchical structure, for example, it can be realized by using a hypertext function of HTML or XML.

FIG. 20 is a diagram showing an example of the output result of the result output means 186. In this example, a list of document classification results
From (2001), there is a link to the classification of information extracted from documents of each classification. Further, a list of individual document names can be referenced from the information classification.

For example, as one of the links from the “job information”, a list of combinations of “work place” and “occupation” (200
2) is linked, and further to the list of relevant job postings (2003). For "Event", "Date"
It links to the event list (2005) from the classification (2004).

By using such an output format, it is easy to find a target document from the output result.

(Third Embodiment) Next, a third embodiment of the present invention will be described.
An embodiment will be described in detail with reference to the drawings.

Referring to FIG. 21, according to a third embodiment of the present invention, a document classification unit 2101 and an information extraction control unit 2
, An information extraction means 3, an information classification control means 4, an information classification means 5, a result selection means 7, and a result output means 6.

Next, the overall operation of the present embodiment will be described in detail with reference to FIG.
2, information extraction means 3, information classification control means 4, information classification means
5. The operation of the information output means 6 is the same as that of the first embodiment, and the description is omitted.

The information classifying means 2101 of this embodiment classifies documents and transfers only documents included in a specific category to the information extraction control means 2.

Next, effects of the present embodiment will be described.

In the present embodiment, since information extraction processing is performed only on documents classified into a specific category, necessary information can be efficiently extracted.

As a special case of the present embodiment,
When there is only one specific category, it is not necessary to select an information extraction unit and an information classification unit based on the category, so that the information extraction control unit and the information classification control unit can be omitted (FIG. 22).

Further, in order to realize the information extraction device in this embodiment by a computer, in the case of the first embodiment, the document classification means 1, the information extraction control means 2, the information extraction means 3, the information classification control Means 4, information classification means 5,
A computer program for realizing each function of the result output means 6 is created, and the computer program is stored in a CD-R
If the above functions are generated in a computer by recording the program on a recording medium represented by an OM, a floppy (registered trademark) disk or a semiconductor memory, and reading out the recording medium on which the program is recorded on the computer side,
The information search device of the present invention can be realized by a computer. Also, this computer program
For example, the program may be recorded in a recording device in the server, or a program provided in the server may be provided via a network.

[0107]

A first effect of the present invention is that information can be accurately extracted. The reason is to classify documents and define an information extraction method according to the classification result.

A second effect of the present invention is that only necessary information can be extracted. The reason is that documents are classified and only important information is extracted according to the classification.

A third effect of the present invention is that information can be extracted from a large number of documents. The reason is that the information can be extracted accurately, so that the information can be automatically extracted without using any manual operation.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a first exemplary embodiment of the present invention.

FIG. 2 is a block diagram illustrating a configuration of an information extracting unit 3 according to the first embodiment.

FIG. 3 is a block diagram illustrating a configuration of an information classification unit 5 according to the first embodiment.

FIG. 4 is a flowchart showing the operation of the first embodiment.

FIG. 5 is a diagram illustrating an example of calculating a degree of suitability for a Web page in the first embodiment.

FIG. 6 is a diagram illustrating an example of types of information stored in an extraction information defining unit 32 according to the first embodiment.

FIG. 7 is a diagram illustrating an example of an information extraction method stored in an extracted knowledge storage unit 33 according to the first embodiment.

FIG. 8 is an example of a display image of a web page of “job information” in the first embodiment.

FIG. 9 is an example of a display image of a web page of “event” in the first embodiment.

FIG. 10 is a diagram showing an example of information extracted by the information extracting means 3 of the first embodiment.

FIG. 11 is a diagram showing an outline of a classification method described in the classification knowledge storage means 52 of the first embodiment.

FIG. 12 is a diagram illustrating a classification result of the first embodiment.

FIG. 13 is a diagram illustrating an example of types of information stored by an extraction information defining unit 32 according to the second embodiment.

FIG. 14 is a diagram illustrating an example of an information extraction method stored in an extracted knowledge storage unit 33 according to the second embodiment.

FIG. 15 is a diagram illustrating an example of information extracted by the information extracting unit 3 of the second embodiment.

FIG. 16 is a diagram showing an outline of a classification method described in the classification knowledge storage means 52 of the second embodiment.

FIG. 17 is a diagram illustrating a classification result of the second embodiment.

FIG. 18 is a block diagram showing a configuration of a second exemplary embodiment of the present invention.

FIG. 19 is a flowchart showing the operation of the second embodiment.

FIG. 20 is a diagram illustrating an output example of an output unit 1806 according to the example of the second embodiment.

FIG. 21 is a block diagram showing a configuration of a third exemplary embodiment of the present invention.

FIG. 22 is a block diagram showing another configuration of the third embodiment.

FIG. 23 is a flowchart showing the operation of a conventional document search device.

[Explanation of symbols]

 1, 2101 Document classification means 2 Information extraction control means 3 Information extraction means 31 Extraction execution means 32 Extracted information definition means 33 Extracted knowledge storage means 4 Information classification control means 5 Information classification means 51 Classification execution means 52 Classification knowledge storage means 6, 186 Result output means 7 Result selection means

Claims (18)

[Claims] 1. A document classifying means for classifying an input document, information extracting means for changing information to be extracted from the document and a method of extracting the information in accordance with the classification of the document, An information extraction system comprising at least an information extraction unit for classifying the information extracted by the information extraction unit. 2. A document classifying means for classifying individual documents included in an input document set into a plurality of categories; an extraction information defining means defining a type of information to be extracted from a document belonging to a specific category; Information extraction means for extracting information defined in the extraction information definition means from documents classified into the category with reference to the extraction information definition means, and appropriate information extraction according to the document classification result of the document classification means Information extraction control means for selecting information and extracting information from a document; information classification means for classifying information extracted from a document belonging to a specific category; Select the information classification method,
An information classification control unit for performing control for classifying information extracted from the document; a classification result by the document classification unit; information extracted by the information extraction unit; and a classification result by the information classification unit. An information extraction system comprising: a result output unit. 3. The information extraction system according to claim 2, wherein said document classification means includes a type discrimination means for discriminating a document type as a classification method for structured documents. 4. The information extracting system according to claim 2, wherein said document classifying means has a function of extracting a document belonging to a specific category. 5. The information extraction system according to claim 2, wherein said result output means selects and outputs information belonging to a specific document classification or information classification. 6. The information extraction system according to claim 2, wherein the result output unit outputs the information extracted by the information extraction unit in a format having a hierarchical structure. . 7. A document classification step for classifying an input document, an information extraction step for changing information to be extracted from the document and a method for extracting the information in accordance with the classification of the document performed in the document classification step. An information classification step of classifying the information extracted in the information extraction step according to the classification of the document performed in the document classification step. 8. A document classification step of classifying individual documents included in an input document set into a plurality of categories, an extraction information definition step defining a type of information to be extracted from a document belonging to a specific category, An information extraction step of extracting information defined by the extraction information definition step from documents classified into the category with reference to the extraction information definition step; and an appropriate information extraction step according to a document classification result of the document classification step , An information extraction control step of performing control for extracting information from a document by selecting a document, an information classification step of classifying information extracted from a document belonging to a specific category, and an appropriate information according to the document classification result. An information classification control step of selecting a classification step and performing control for classifying information extracted from the document; Classification result by the document classification step, the information extracted information extracted by the step, the information extraction method characterized by comprising the result output step of outputting a classification result, by the information classification step. 9. The information extraction method according to claim 8, wherein said document classification step includes a type determination step of determining a document type as a classification method for structured documents. 10. The information extraction method according to claim 8, wherein said document classification step has a function of extracting a document belonging to a specific category. 11. The information extraction method according to claim 8, wherein in the result output step, information belonging to a specific document classification or information classification is selected and output. 12. The information extraction method according to claim 8, wherein the result output step outputs the information extracted in the information extraction step in a format having a hierarchical structure. . 13. A computer, comprising: a document classifying step of classifying an input document; information extracted from the document in accordance with the classification of the document performed in the document classifying step; and information changing a method of extracting the document. An information extraction program for executing at least an extraction step and an information classification step of classifying the information extracted in the information extraction step according to the classification of the document performed in the document classification step. recoding media. 14. A computer, comprising: a document classification step of classifying individual documents included in an input document set into a plurality of categories; and an extraction information definition step of defining types of information to be extracted from documents belonging to a specific category. An information extraction step of extracting information defined by the extraction information definition step from documents classified into the category with reference to the extraction information definition step; An information extraction control step of selecting an information extraction step to perform control for extracting information from a document; an information classification step of classifying information extracted from a document belonging to a specific category; Information that selects an appropriate information classification step and performs control for classifying information extracted from the document A class control step; a result output step of outputting a classification result by the document classification step; information extracted by the information extraction step; a classification result by the information classification step; and an information extraction program to be executed. Recording medium on which is recorded. 15. The method according to claim 1, wherein the document classification step includes a type determination step of determining a document type as a classification method for structured documents.
A recording medium on which the information extraction program according to item 4 is recorded. 16. The recording medium according to claim 14, wherein said document classification step has a function of extracting a document belonging to a specific category. 17. The information extraction program according to claim 14, wherein the result output step selects and outputs information belonging to a specific document classification or information classification. The recording medium on which it was recorded. 18. The information extracting apparatus according to claim 14, wherein the result outputting step outputs the information extracted by the information extracting step in a format having a hierarchical structure. A recording medium on which a program is recorded.