JPH041853A - Document retrieving device - Google Patents

Abstract

PURPOSE:To improve the document retrieving efficiency by providing an index extracting means which generates an index and a retrieving device which displays a document format stored in a document format storage means in a retrieving state and retrieves a document by reference to the index and based on the inputted retrieving conditions. CONSTITUTION:A document format storage means 9 stores the formats of the documents to be retrieved, and an index extraction means 10 extracts the data necessary for an index based on the format stored in the means 9 and generates the index. Then a retrieving means 12 displays the format stored in the means 9 in a retrieving state and retrieves a document by reference to the index and based on the inputted retrieving conditions. As a result, an index is automatically produced and at the same time the data extracted out of the document are accurately designated based on each format. Furthermore a retrieving operator can easily obtain the retrieving conditions in a blank filling way without knowing any special retrieving language. Thus the document retrieving efficiency is improved.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a document retrieval device that uses an index to retrieve a designated document from among a plurality of accumulated documents, and particularly relates to a document retrieval device that uses an index to retrieve a designated document from among a plurality of accumulated documents. The present invention relates to a document retrieval device.

[Prior Art] In order to construct a document database using a computer system, it is necessary to create an index that includes keywords that appropriately represent document contents depending on the purpose of use. In conventional document retrieval devices, indexes are generally created manually, where a person looks at the document itself, finds necessary keywords from the document based on the format, etc., and inputs them into a single table-format index. It was created by. Therefore, building a document database requires 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 input means such as a word processor for inputting a document, 2 is a document storage means such as a magnetic disk device for storing the input document, and 3 is a natural language processing function that automatically extracts keywords from the stored document. 4 is a keyword storage means for storing the extracted keywords; 5 is a search request input means such as a keyboard or display for inputting keywords serving as search conditions as a search request; 6 is a search request input means for inputting keywords to be used as search conditions; A search means refers to the keyword storage means 4 based on the keyword and searches for a designated document from among the documents stored in the document storage means 2, and 7 is a search result output means such as a display for outputting the search results. .

This system divides the input document into words based on grammar, then evaluates the divided words and removes unnecessary terms such as particles. Keywords are automatically extracted and set.

[Problems to be Solved by the Invention] Conventional document retrieval devices have been configured as described above, and as mentioned above, indexes generally have to be created manually, which requires a great deal of cost and time to construct a database. There was a problem that it took a while. In addition, devices that automatically extract keywords from documents using natural language processing functions have been put into practical use, but since all nouns, verbs, etc. are used as keywords, the extracted keywords may be ambiguous or inappropriate. There were problems such as it took time to extract and set keywords, and searches could not be performed efficiently.

This invention was made to solve the above-mentioned problems, and is capable of automatically extracting only the really necessary keywords, automatically extracting the detailed contents of one document as an accurate index, and The object of the present invention is to obtain a document retrieval device that can perform efficient retrieval.

[Means for Solving the Problems] A document search device according to the present invention includes a document format storage means for storing the format of a document to be searched, and an index from the document to be searched using the format stored in the document format storage means. an index extraction means that extracts data necessary for the search and generates an index, and displays the format stored in the document format storage means at the time of search, and extracts the document using the index based on the search conditions entered. and a search means for searching.

[Operations] In this invention, we focus on the fact that documents are generally written in a fixed format, store this document format, and use the document format to extract data necessary for indexing from the document to be searched. By doing this, the index will be automatically generated.

By using formats in this way, index creation can be done accurately because the data to be extracted from documents is specified by the format, and the document structure can be analyzed using the format, so it is possible to create an index in detail. Items can be used as indexes. Furthermore, since a format can be presented to the searcher during a search, the searcher can easily provide search conditions in a fill-in-the-blank format without knowing a special search language.

[Example] 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 a document retrieval device according to an embodiment, and the same or corresponding parts as in FIG. In the figure, 8 is a document format input means such as a word processor for inputting a format defined 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 a document storage means 2. Index extraction means 11 extracts data necessary for an index from documents stored in the document format storage means 9 using the format stored in the document format storage means 9 to generate an index; The index storage means 12, such as a magnetic disk device, displays the document format stored in the document format storage means 9 on the display of the search request input means 5 at the time of a search, and in response to a search entered from the keyboard. This is a search means for searching documents in the document storage means 2 based on conditions using the index in the index storage means 11, and the search results are outputted to the search result output means 7 such as a display. Note that the index extraction means 10 and the search means 12 are realized by a processor constituting a computer system and software running thereon.

Next, the operation will be explained.

As mentioned above, this application focuses on the fact that documents are generally written in a fixed format, and in particular, technical documents that are subject to document databases, such as research reports, standards, specifications, etc., are shown in Figure 2. As shown in the figure, 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 frame of ruled lines. In addition to the above-mentioned formats, formats handled in this application include tables in documents (for example, oO standard tables), and formats that do not need frames such as those in specifications.

First, the document itself is input by the document input means 1 and stored in the document storage means 2 as in the conventional case. Further, in this apparatus, the format of the document is inputted from the document format input means 8 and stored in the document format storage means 9. One type of document format is sufficient when handling only documents with the same format, but when handling documents with different formats, only that type is input, and that type is specified when extracting an index or searching. Generally, a document consists of a collection of paragraphs (logical units that can be considered as a group in a sense). Therefore, as shown in Figure 3, the data input as a format includes the document structure (frames, paragraph titles) that shows what kind of paragraphs the document consists of, and the position of the data to be extracted (shaded areas). ) and the correspondence relationship with the extracted data index (pointer indicated by an arrow). In the case of Figure 3, a paragraph is defined as the area enclosed by ruled lines.

On the other hand, every time a document is input from the document input means 1 and stored in the document storage means 2, the index extraction means 10 performs the document structure analysis process (step Sl) shown in FIG. (Step S2) and index generation processing (Step S3) are executed. Fourth
Figure ('b) shows the input, processing contents, and output at each processing step, and the flow of data is shown in Figure 4 (fc). Note that in the document structure analysis shown in the same figure, an example was taken in which an index is generated from the cover of a document surrounded by ruled lines, but document structure analysis is also performed when paragraphs are surrounded by ruled lines, as shown in Figure 5 (a). Even if the length of the paragraph is variable as shown in FIG. 5(b), it can be handled.

FIG. 6 is a diagram for explaining the above-mentioned document structure analysis process in more detail. a document element analysis process (step 511) that decomposes each of the obtained minimum constituent elements into paragraphs of the format data taken out from the document format storage means 9 (several alternatives are possible); Step 512) and a structure analysis process (Step 513) in which the most likely correspondence is selected and output from among the obtained paragraph correspondences, and the following processes are performed respectively.

(1) Step 511 (Document Element Analysis) Here, the contents of the document data are decomposed into minimum constituent elements, and each element is numbered in order. The minimum numbered components are called element data. The minimum constituent elements referred to here are a) rows b) rows in tables C) areas other than 9 documents such as figures and graphs, and a line is a string of characters up to the line feed symbol or table border It is.

(2) Step 512 (paragraph correspondence) Here, the beginning 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 that becomes the beginning of each paragraph is selected ( (Multiple selections possible) These correspondences are called paragraph correspondence data. A specific example is shown in Figure 7 (
a), (bl). Five paragraph-corresponding data as shown in Fig. 8 are obtained from the two data shown in (al, (bl) in the same figure.Paragraph-corresponding data ■ and ■ shown in the figure Although these are mutually contradictory data, both are held as candidates at this stage, and in the next step 513 (structural analysis), either ■ or ■ is selected.

(3) Step 513 (Structural analysis) Select the most appropriate combination from the paragraph-corresponding data obtained in step S12 above using the production rule set, and divide the finally determined set of paragraph-corresponding data into paragraphs. The amount will be secured as data. As a result, the element data at the beginning of each paragraph is determined, and therefore the set of element data that constitutes each paragraph is also determined.

As a production rule set, the consistency of the order of paragraphs, the certainty of paragraphs, etc. can be considered. An example of a production ruleset is shown below.

- Consistency of paragraph order a) In paragraph correspondence data, correspondences that disturb the ascending order of paragraph IDs are excluded from candidates.

- Probability as a paragraph b) If the paragraph title is a title and the corresponding minimum constituent element is 40 characters or more, it is excluded from the candidates.

In the above paragraph correspondence data (FIG. 8), 31 is excluded from the candidates according to rule set b) and 2° is excluded from the candidates according to rule set a). The remaining correspondences are then selected as the most appropriate combinations.

When the document structure analysis process (step 81) is completed as described above, the next extracted data determination process (step S2)
Then, from the element data set constituting each paragraph obtained in step 513 (structure analysis), excluding the paragraph title of that paragraph, is extracted as index data (see Figure 9 and Figure 4 above). t
(see el).

Then, the final index generation process (step S3)
In this step, an index is generated on the index storage means 11 by inputting the data extracted in step S2 (extraction data determination process) into a field of a table storing index data (see FIG. 4 (C1)).

Next, to explain the time of searching, for example, Fig. 10 fa
If the index data shown in 1 is prepared by the above-mentioned automatic extraction, the searcher will search for ``Summary of 1 research report written by ``Yamaguchi''. When making a search request such as "I want to see the part of Search conditions are provided to the search means 12 by simply entering the necessary items in this format.

Note that the book in the figure is a well-known wildcard, and when this is given, a text search is performed. This results in
The search means 12 performs a normal search using an index, and as a search result, the specified location of the document that satisfies the search request is output as shown in FIG. be done.

[Effects of the Invention] As described above, according to the present invention, a document format storage means for storing the format of a document to be searched, and a format necessary for indexing from the document to be searched using the format stored in the document format storage means. an index extraction means for extracting data and generating an index; and displaying the format stored in the document format storage means at the time of search, and searching for a document using the index based on the search condition input thereto; Since it is equipped with a search means to automatically create an index, and the data to be extracted from a document is specified by a format, it can be done accurately. Also, since the document structure can be analyzed using the format, it is possible to automatically create an index. Detailed contents can be used as an index. Furthermore, since a format can be presented to the searcher during a search, the searcher can easily provide search conditions in a fill-in-the-blank format without knowing a special search language.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the overall configuration of a document retrieval device 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 is a diagram for explaining the operation of the index extraction means, FIG. 5 is a diagram showing another example of a document from which index extraction is possible, and FIG. 6 is a diagram for explaining the document structure analysis process in more detail. FIG. 7 is a diagram showing an example of bar graph correspondence, FIG. 8 is a diagram showing an example of paragraph correspondence data, FIG. 9 is a diagram showing an example of extracted data, and FIG. 10 is a diagram showing the operation of the search means. FIG. 11, which is a diagram for explanation, is a block diagram showing the configuration of a conventional example. 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, and 11 is an index storage Means, 12 is a search means. In addition, in the figures, the same reference numerals indicate the same or equivalent parts.

Claims (1)

[Claims] A document retrieval device that retrieves a specified document from a plurality of stored documents using an index, comprising: a document format storage unit that stores a format of a document to be retrieved; and a document to be retrieved. index extracting means for extracting data necessary for an index from the format using the format stored in the document format storage means and generating an index; displaying the format stored in the document format storage means at the time of retrieval; and a search means for searching for a document using the index based on search conditions input to the document.