JP2549745B2 - Document search device - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a document search device for searching a document specified by an index from a plurality of stored documents, and in particular, a format is predetermined. The present invention relates to a device for retrieving existing documents.

[Prior Art] In order to construct a document database using a computer system, it is necessary to create an index including a keyword that appropriately represents the contents of a document, depending on the purpose of use. In a conventional document retrieval apparatus, an index is generally created manually, and a person looks at the document itself, finds out a necessary keyword from the document based on a format, etc., and inputs it into a tabular index. Was created by. Therefore, the construction of the document database has required a great deal of cost and time.

Therefore, a database system as shown in FIG. 11 has been put into practical use. In the figure, 1 is a document inputting means such as a word processor for inputting a document, 2 is a document storing means such as a magnetic disk device for storing the input document, and 3 is a keyword automatically stored in the stored document by a natural language processing function. Automatic keyword extracting means for extracting into 4 is a keyword storing means for storing the extracted keywords, 5 is a search request inputting means such as a keyboard or display for inputting keywords as search conditions as a search request, and 6 is input Search means for searching the specified document from the documents stored in the document storage means 2 by referring to the keyword storage means 4 based on the keyword, and 7 is a search result output means such as a display for outputting the search result. .

In this system, the input document is divided into words based on the grammar, and is divided into words. After that, the divided words are evaluated and unnecessary words such as particles are removed. Keywords are automatically extracted and set.

[Problems to be Solved by the Invention] Since the conventional document retrieval apparatus is configured as described above, it is generally necessary to manually create an index as described above, which requires a great deal of cost and time for database construction. There was a problem of this. In addition, automatic extraction of keywords from documents using a natural language processing function has also been put into practical use, but since all business cards and verbs are keywords, the extracted keywords may be ambiguous or inappropriate. However, there are problems that it takes time to extract and set the keywords, and the search cannot be performed efficiently.

The present invention has been made to solve the above-mentioned problems, and it is possible to automatically extract only really necessary keywords, and automatically extract detailed items of the document contents accurately as an index, It is an object of the present invention to obtain a document search device capable of efficiently searching.

[Means for Solving the Problem] A document search device according to the present invention stores a document storage unit for storing a document input via a document input unit and a document format input via a document format input unit. The document format storage means to be stored and the document data stored in the document storage means are decomposed into some element data, and the element data of the element data is stored based on the paragraph title in the format data stored in the document format storage means. Paragraph data candidates are extracted from the paragraphs, paragraph data candidates that conform to the rules for document formats such as consistency of paragraph order and certainty as paragraphs are determined as paragraph data, and the determined paragraph data is combined and documented. The index extraction means for generating the index of and the format data stored in the document format storage means at the time of retrieval Display,
On the other hand, a search means for searching the document using the index based on the input search condition is provided.

[Operation] In the present invention, attention is paid to the fact that a document is generally written according to a fixed format, and this document format is stored in the document format storage means. The index extraction means decomposes the document data stored in the document storage means into some element data, and based on the paragraph title in the format data stored in the document format storage means, the paragraph data from the element data. Are extracted, the paragraph data candidates that conform to the rules regarding the document format are determined as paragraph data, and the determined paragraph data are combined to generate a document index. The index generated in this way is highly reliable, and therefore the reliability at the time of retrieval is also high. Also, the data of each paragraph of the document,
That is, detailed items of the contents of the document can be used as the index. Further, since the format can be presented to the searcher in the search, the searcher can easily give the search condition to the fill-in-the-blank formula without knowing a special search language.

[Embodiment] An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the overall configuration of the document retrieval apparatus of the embodiment. The same or corresponding parts as in FIG. 11 are designated by the same reference numerals and the description thereof is omitted. In the figure, 8 is a document format input means such as a word processor for inputting a predetermined format for a document to be searched, 9 is a document format storage means such as a magnetic disk device for storing the input document format, and 10 is an index extracting means. Yes, this decomposes the document data stored in the document storage means 2 into some element data,
A candidate for paragraph data is extracted from the element data based on the paragraph title in the format data stored in the document format storage means 9, and is related to the document format such as consistency of paragraph order and certainty as a paragraph. The paragraph data candidates that conform to the rules are determined as paragraph data, and the determined paragraph data are combined to generate a document index. Reference numeral 11 is an index storage means such as a magnetic disk device in which the index generated by the index extraction means 10 is stored, and 12 is a display of the search request input means 5 for displaying the document format stored in the document format storage means 9 at the time of retrieval. Is a search means for searching the document in the document storage means 2 using the index in the index storage means 11 based on the search condition input from the keyboard. It is output to the result output means 7. The index extracting means 10 and the searching means 12 are realized by a processor that constitutes a computer system and software that operates on the processor.

 Next, the operation will be described.

As described above, the present application focuses on the fact that documents are generally written according to a fixed format, and in particular, technical documents, such as research reports, standards, and specifications, that are subject to the document database are shown in FIG. As shown in, the cover of the first page and the table of contents of the second page have a common format in which each field indicating the type and title is divided by a ruled frame. In addition,
In addition to the above, the format used in the present application may be a table in a document (for example, a standard table of XX) or a frame such as a specification.

First, the document itself is the document input means 1 as in the conventional case.
Is input and stored in the document storage means 2. Further, in this apparatus, the format of the document is input from the document format input means 8 and stored in the document format storage means 9. This document format may be of one type when handling only documents of the same format, but only that type is input when handling documents of different formats, and that type is specified during index extraction and retrieval. In general, a document consists of a set of paragraphs (logical units that in a sense can be regarded as a block). Therefore, as shown in FIG. 3, the data input as the format includes a document structure (frame, paragraph title) showing what kind of paragraph the document is composed of,
The correspondence between the position of the data to be extracted (hatched portion) and the index of the extracted data (pointer indicated by an arrow) and the like. In the case of FIG. 3, the paragraph means the inside surrounded by ruled lines.

On the other hand, the index extracting means 10 each time a document is input from the document inputting means 1 and stored in the document storing means 2, the document structure analyzing process (step S1) and the extracted data determining process (step S1) shown in FIG. A series of processing of step S2) and index generation processing (step S3) is executed. FIG. 4 (b) shows the input, processing contents, and output at each processing step, and the data flow is shown in FIG. 4 (c). In the document structure analysis of the figure, an example of generating an index from the cover of a document surrounded by ruled lines has been taken, but in the document structure analysis, paragraphs are surrounded by ruled lines as shown in FIG. 5 (a). Alternatively, it is possible to cope with the paragrow having a variable length as shown in FIG. 5 (b).

FIG. 6 is a diagram for explaining the document structure analysis process in more detail. The document structure analysis process (step S1) is an inference unit, and the document data retrieved from the document storage means 2 is used as the minimum component. A document element analysis process for decomposing (step S11) and a paragraph correspondence process (step S11) for corresponding each obtained minimum constituent element to a paragraph of the format data retrieved from the document format storage means 9 (several alternatives are possible). S12) and the structural analysis processing (step S13) of selecting and outputting the correspondence having the highest possibility from the correspondence of the obtained paragraphs, and the following processing is performed respectively.

(1) Step S11 (Document Element Analysis) Here, the contents of the document data are decomposed into the minimum constituent elements, and the respective elements are sequentially numbered. The smallest numbered constituent elements are called element data. The minimum components mentioned here are: a) rows b) rows in a table c) areas other than documents, such as figures and graphs, and rows are character strings up to where there are line feed symbols or table ruled lines. Is.

(2) Step S12 (corresponding to paragraph) Here, the head of the element data is matched with the paragraph title (keyword for finding the paragraph) of each paragraph in the format data, and the element data to be the start of each paragraph is selected ( You can select more than one). These correspondences are called paragraph correspondence data. Specific examples are shown in FIGS. 7 (a) and 7 (b). The same figure (a),
From the two data shown in (b), five paragraph corresponding data as shown in FIG. 8 are obtained. The paragraph-corresponding data shown in the figure are data that contradict each other, but at this stage, both of them are held as candidates, and either of and is selected in the next step S13 (structural analysis).

(3) Step S13 (Structural analysis) The most appropriate combination is selected from the paragraph corresponding data obtained in step S12 using the production rule set, and the finally determined set of paragraph corresponding data is divided into paragraphs. Secure as data. As a result, the element data at the head of each paragraph is determined, and thus the set of element data forming each paragraph is also determined.

As for the production ruleset, the consistency of the paragraph order and the certainty of the paragraph can be considered. Below is an example of a production ruleset.

In the above paragraph-corresponding data (FIG. 8), 3 ^* is excluded from the candidates by rule set b) and 2 ^** is excluded from the rule set a). Then, the remaining associations are selected as the most appropriate combination.

When the document structure analysis process (step S1) is completed as described above, in the next extraction data determination process (step S2), from the element data set forming each paragraph obtained in step S13 (structure analysis), The data excluding the paragraph title of the paragraph is extracted as index data (Fig. 9 and Fig. 4 (c)).
reference).

And the final index generation process (step S3)
In step S2 (extracted data determination process), the index is generated in the index storage means 11 by inputting the data extracted in the field of the table storing index data (see FIG. 4 (c)). .

Next, the search will be described. For example, when the index data as shown in FIG. 10 (a) is prepared by the automatic extraction described above, the searcher writes “'Yamaguchi' wrote, There is a word "software design." I want to see the main part of the "research report"", the search means 12 corresponds to the search request input screen as shown in FIG. 10 (b). The frame of the format to be displayed is displayed, and the searcher simply inputs the necessary items in this format displayed by the system, and the search condition is given to the search means 12. It should be noted that * in the figure is a well-known wild card, and when given, a text search is performed. As a result, the search means 12 performs a normal search using the index, and as a search result, the designated place of the document satisfying the search request is output as shown in FIG. Is output in multiple windows.

[Effects of the Invention] As described above, according to the present invention, the document storage means for storing the document input via the document input means and the format of the document input via the document format input means are stored. The document format storage means and the document data stored in the document storage means are decomposed into some element data, and based on the paragraph style in the format data stored in the document format storage means Paragraph data candidates are extracted, paragraph data candidates that conform to the rules for document formats such as consistency of paragraph order and certainty as paragraphs are determined as paragraph data, and the determined paragraph data are combined to index the document. Displaying the format data stored in the document format storage means at the time of retrieval,
On the other hand, since a search means for searching a document using the above-mentioned index based on the input search condition is provided,
An index can be created automatically, and the data to be extracted from the document is specified by the format, so it can be performed accurately, and since the document structure can be analyzed using the format, detailed items of the document content can be indexed. In addition, a highly reliable index can be obtained by the most appropriate combination of paragraphs, and the reliability at the time of search is also high. Further, since the format can be presented to the searcher in the search, the searcher can easily give the search condition to the fill-in-the-blank formula without knowing a special search language.

[Brief description of drawings]

FIG. 1 is a block diagram showing the overall configuration of a document retrieval apparatus according to an embodiment of the present invention, FIG. 2 is a diagram for explaining a document format, and FIG. 3 is a diagram for explaining data input as a format. 4 and FIG. 4 are diagrams for explaining the operation of the index extracting means, FIG. 5 is a diagram showing another example of a document in which index extraction is possible, and FIG. 6 is a detailed explanation of the document structure analysis process. FIG. 7, FIG. 7 shows an example of paragraph correspondence, FIG. 8 shows an example of paragraph correspondence data, FIG. 9 shows an example of extracted data, and FIG. FIG. 11 is a block diagram showing a configuration of a conventional example for the purpose of explanation. 1 is a document input means, 2 is a document storage means, 5 is a search request input means, 7 is a search result output means, 8 is a document format input means, 9 is a document format storage means, 10 is an index extraction means, 11 is an index storage Means, 12 is a search means. In the drawings, the same reference numerals indicate the same or corresponding parts.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Takao Hirata 5-1-1, Ofuna, Kamakura-shi, Kanagawa Sanryo Electric Co., Ltd. Information Electronics Research Laboratory (56) Reference JP-A-2-72468 (JP, A) JP 63-226765 (JP, A) Proc. Of the 39th National Conference of Information Processing Society of Japan (Late 1989) 953

Claims (1)

(57) [Claims] 1. A document retrieval device for retrieving a document designated by using an index from a plurality of stored documents, and a document storage means for storing the document inputted via the document input means. A document format storage means for storing the format of the document input via the document format input means, and the document data stored in the document storage means are separated into some element data and stored in the document format storage means. Paragraph data candidates are extracted from the above element data based on the paragraph title in the formatted data, and paragraph data candidates conforming to the document format rules such as consistency of paragraph order and certainty as paragraphs are selected. Index extraction means for determining the paragraph data and combining the determined paragraph data to generate a document index; Document search characterized by comprising a search means for displaying the format data stored in the document format storage means at the time of searching, and for searching the document using the index based on the search condition input thereto. apparatus.