JP2885487B2 - Document information retrieval device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an in-document information retrieval apparatus for retrieving contents in a document file by inputting a sentence, a word string, or a word.

[Conventional technology]

As a device for searching the contents of a conventional document file,
2. Description of the Related Art There is known an apparatus that searches for a portion including a character string by inputting the character string.

[Problems to be solved by the invention]

However, in the conventional device, since a search cannot be performed unless the character strings completely match, for example, it is not possible to input a character string of “format” and search for a portion including “format” having the same meaning as the format. You couldn't search for parts containing "how to open a file" by typing "open file".

SUMMARY OF THE INVENTION An object of the present invention is to easily and quickly search for a part including a word having the same meaning and a part including a sentence having a high semantic similarity to an input sentence, even if the input sentence is not a matched character string. It is an object of the present invention to provide an in-document information search device having a function capable of performing such a function.

[Means for solving the problem]

A document structure extraction unit for extracting first information of chapters, headings and paragraphs in a document file, a registered sentence analysis unit for extracting notations and semantic categories, and an index table for generating an index table based on the notations and semantic categories A generating unit, a search sentence analyzing unit for extracting the notation and the semantic category of the words included in the search sentence, a similarity calculating unit for calculating the similarity, and a candidate display unit for displaying a high similarity. It has a configuration that includes it.

[Action]

The notation and meaning of the words extracted by the search sentence analysis unit by referring to the chapter, heading, and paragraph information in the document file extracted by the document structure extraction unit and the index table generated by the index table generation unit Based on the category, the similarity calculation unit calculates the similarity with the search sentence, and based on the similarity, displays the headline sentence or paragraph as a candidate whose contents should be displayed by the candidate display means. Searches for a content that contains a semantically similar part even if there is no character string that matches the column.

〔Example〕

FIG. 1 is a block diagram showing an embodiment of the present invention. Hereinafter, the operation of the embodiment of the present invention will be described with reference to FIG.

FIG. 2 is a diagram showing an example of a document file. The document file storage unit 1 is composed of a plurality of chapters, each chapter is composed of a headline sentence and a text, each text is composed of a plurality of paragraphs, and each paragraph is composed of a plurality of sentences. Document files such as are stored.

The document structure extraction unit 2 analyzes the document file 21 stored in the document file storage unit 1, extracts chapters, headlines, and paragraphs, and creates a document structure table indicating the positions and hierarchical relationships. A chapter, a headline, and a paragraph can be extracted by various methods. Here, one example will be described.

First, a character string indicating [chapter number] such as "1."
Find strings such as "Chapter". [Chapter number] can be extracted by searching for a character string that satisfies the following pattern.

[Chapter number] = [Numerical string] + "." OR [Numerical string] + "Chapter" (However, [Numerical string] starts from the beginning of the line) Next, the blank characters following [Chapter number] are removed A character string up to a line feed code is extracted as [headline sentence].

 [Paragraph] is extracted according to the following rules.

[Paragraph] = character string following [line feed code] up to [line feed code] (however, this [line feed code] excludes the line feed code at the end of the line) [Chapter number] to the line immediately before the next line describing [Chapter number].

With the above-described method, a chapter, a headline, and a paragraph are extracted from the document file, and a position in each document file is obtained to create a document structure table.

FIG. 3 is a diagram showing an example of a document structure table.
The document structure table 22 created by the document structure extraction unit 2 is stored in the document structure table unit 3. ID in Fig. 3
Is an identifier for managing chapters, headlines, and paragraphs. For example, A1 is the first chapter, A2 is the second chapter, and B1 is the headline sentence shown in FIG. , C1 is the first paragraph of the first chapter and corresponds to "the simplest method shown in FIG. 2". Further, from the table 22, it is possible to easily extract information that the headline sentence of the first chapter includes B1, and the paragraphs included in the chapter include C1, C2,..., C6. S and E represent the start position and the end position of a chapter, a headline, and a paragraph in the document file. By referring to this, a character string corresponding to each chapter, each headline, and each paragraph can be read from the document file storage unit 1.

The registration sentence analysis unit 6 sends each headline sentence and each paragraph to the morphological analysis unit 5. The morphological analysis unit 5 refers to the word dictionary unit 4 and extracts the notation and semantic category of each word included in the sentence to be analyzed.

FIG. 4 is a diagram showing an example of a registered sentence analysis result. FIG. 4 shows an example of the result of analyzing the headline sentence “output with format”. In this example, the word notation "format",
The semantic category [FMT] for "attached", "output", "format", and the semantic category [OUT] for "output" are extracted. The semantic category 24 is information used to link synonyms and related words, and is defined in the word dictionary unit 4 for independent words. However, words such as “a” and “do” that do not express a specific meaning alone are not added.

The registered sentence analysis unit 6 uses the word notation obtained in this way.
23 and the semantic category 24 are sent to the index table creating section 7 together with the ID of the headline and paragraph to which they belong.

The index table creation unit 7 uses the notation 23 and the semantic category 24 as an index based on the notation 23 and the semantic category 24, and extracts an index sentence including a word having the notation or the semantic category and a paragraph including the word. Generate Table 25.

FIG. 5 is a diagram showing an example of an index table using a semantic category as an index. With this index table 25, headlines B2 and B4 and paragraphs C8 and C18 including words having the semantic category [FMT] can be easily extracted.

Index table 25 created in this way
Are stored in the index table storage unit 8.

The document file 21, the document structure table 22, and the index table 25 accumulated as described above are used in search processing described below.

The search sentence input unit 9 allows the user to input a sentence, word string, or word, and sends the input character string to the search sentence analysis unit 10 as a search sentence. Note that the search sentence input unit 9 may be configured so that a search sentence is input by designating with a mouse or the like from among those already displayed.

The search sentence analysis unit 10 sends the search sentence 26 obtained by the search sentence input unit 9 to the morphological analysis unit 5, and extracts the word notation 23 and the semantic category 24 included in the search sentence 26.

FIG. 6 is a diagram showing an example of a search sentence analysis result when “write by format specification” is input as a search sentence. In this example, "format", "specify", "ni", "depend", "write" as word notation 23
Is extracted and the semantic category 24 is [FMT], [SITE
I] and [OUT] are extracted.

The similarity calculator 11 first calculates the similarity of each headline sentence and each paragraph with the search sentence. Various methods can be used for calculating the degree of similarity.
This can be realized by the method described in -111626. Here, a method of calculating the similarity according to how many semantic categories 24 match will be described. For example, if the semantic category 24 is n
If there is a match, 10n points shall be given. Using the semantic category extracted from the search sentence as an index, the ID of the headline sentence or paragraph is extracted by referring to the index table storage unit 8, and the similarity to each ID is calculated. When the index table 25 shown in FIG. 5 is stored, the headline sentences B2, B4, paragraphs C8, C18 are converted to [OUT] by the semantic category [FMT] extracted from the search sentence "write by format specification" 26. Headline B by
2. Paragraph C8 is extracted, and the similarity is calculated as shown in FIG. For example, paragraph C8 is given 20 points because it is extracted by the semantic categories [FMT] and [OUT].

Next, the similarity of each chapter is calculated based on these similarities. Various methods can be applied to the method of calculating the degree of similarity of each chapter. Here, a method based on the following calculation criteria will be described.

Calculation criteria: Similarity of chapter = (Similarity of headline sentence) + (Maximum similarity of paragraph included in that chapter) By referring to the information stored in the document structure table 3, heading sentence B2, paragraph C8, in the second chapter A2,
The headline sentence B4 and the paragraph C18 are found to be included in the fourth chapter A4, and the similarity between A2 and A4 is calculated as follows.

Similarity of A2 = (Similarity of B2) + (Similarity of C8) = 20 + 20 = 40 Similarity of A4 = (Similarity of B4) + (Similarity of C18) = 10 + 10 = 20 Thus, the headline sentence By calculating the similarity between chapters taking into account both the similarity between words and paragraphs, a chapter in which both the headline sentence and the paragraph contain the word related to the search sentence can be compared to a chapter in which only one of them contains the word related to the search sentence. ， Can be searched in high candidate order.

The candidate display unit 12 refers to the document structure table unit 3 for the headline sentence belonging to the chapter and the sentence belonging to the paragraph in descending order of the similarity of each chapter and in the descending order of similarity of each paragraph in the same chapter. Then, it is read from the document file storage unit 1 and displayed.

FIG. 8 is a diagram showing an example of candidate display. In this example, chapter A2 is first and chapter A4 is second. In chapter A2, paragraph C8 takes first place. Here, other paragraphs are not shown, but if there are other extracted paragraphs, the paragraphs are displayed in descending order of similarity.

The document content display unit 13 allows the user to specify the headline or paragraph displayed on the candidate display unit 12, and displays the contents of the chapter containing the headline specified by the user or the contents before and after the paragraph specified by the user. , With reference to the document structure table section 3 and read out from the document file storage section 1 and displayed.

The processing control unit 14 activates the search sentence input unit 9, activates the candidate display unit 12, activates the document content display unit 13 in response to a user's instruction.
Activate and scroll the displayed contents. As a result, the user can instruct processing such as displaying the candidate again or searching for another part by looking at the displayed content.

[Effects of the Invention] As described above, according to the present invention, a headline or paragraph is extracted from a document file, and an index table is created in which semantic attributes and expressions are used as indexes. To quickly calculate the similarity between a search sentence, a headline sentence, and a paragraph, calculate the similarity of each chapter based on this, and display the search results in order of similarity, use a character string in a document file. Even if it is not a character string that matches, a portion where a word related to the input search sentence exists can be searched at high speed. For this reason, the user can easily and quickly find information that he or she wants to use in a new word.

Further, according to the present invention, since the main body is configured to be searchable in paragraph units, for example, a search is made for a part of "declare an array as described above and initialize as follows." You can search by the statement "initialize array". In this way, a search is possible even when a search sentence is related to two or more sentences.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is a diagram showing an example of a document file, FIG. 3 is a diagram showing an example of a document structure table, and FIG. FIG. 5 shows one example of an index table, FIG. 6 shows one example of a search sentence analysis result, FIG. 7 shows one example of similarity calculation, FIG. The figure shows an example of candidate display. DESCRIPTION OF SYMBOLS 1 ... Document file storage part, 2 ... Document structure extraction part, 3 ... Document structure table part, 4 ... Word dictionary part, 5 ... Morphological analysis part, 6 ... Registered sentence analysis part, 7 ... Index table creation part, 8 ... Index Table storage unit, 9 search sentence input unit, 10 search sentence analysis unit, 11 similarity calculation unit, 12 candidate display unit, 13 document content display unit, 14 processing control unit.

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-2-27478 (JP, A) Matsuo, Oyama, Nakagawa "User Input Support for Japanese Dialogue Processing" IPSJ 38th (Showa 64) First semester) National Convention Lectures, (I) p. 400-401 (Hei 1-3-15) Shibano, "SGML and full-text database" IPSJ Research Report (89-FI, 14-2) Vol. 89, No. 66, 1989 (Hei 1-7-27) (58) Fields investigated (Int. Cl. ⁶ , DB name) G06F 17/27-17/30 JICST Science and Technology Reference File

Claims (1)

(57) [Claims] A document file storage unit for storing a document file; a document structure extraction unit for extracting a chapter, a headline sentence, and a paragraph in the document file, and creating a document structure table indicating a position and a hierarchical relationship. A document structure table for storing the document structure table; a word dictionary for defining a word notation and a semantic category for each word; and words forming an input sentence by referring to the word dictionary. And a morphological analysis unit for extracting the notation of the word and the semantic category of the word, sending each headline sentence and each sentence in the document file to the morphological analysis unit, and writing the headline sentence and the word included in each sentence. A registered sentence analysis unit for extracting a semantic category; and a notation and a semantic category as an index based on the notation and the semantic category extracted by the registered sentence analysis unit.
An index table generating unit for generating an index table capable of extracting a headline sentence including the word having the notation or the semantic category and a paragraph including the word, and an index for storing the index table generated by the index table generating unit A table storage unit, a search sentence input unit for inputting a search sentence, a search sentence analysis unit for analyzing the search sentence and extracting the notation and meaning category of a word included in the search sentence, and the search sentence analysis unit Calculating the similarity between each headline sentence and each paragraph and the search sentence by referring to the index table storage unit based on the information obtained in step (1), and calculating the similarity based on the similarity. A calculation unit; a candidate display unit that reads out the high similarity obtained by the similarity calculation unit from the document file storage unit and displays the readout result; Article in the information retrieval apparatus characterized by having.