JPH10149368A - Document retrieval device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a document retrieval device capable of improving a reproducing rate without reducing the adaptation rate of retrieval. SOLUTION: An explanatory word dictionary preparing part 3 prepares an explanatory word dictionary having a certain word and a set of words explaining the word concerned as a pair based on the contents of a Japanese- language dictionary 1 and a synonym dictionary 2 stored in a dictionary storage part 10 and stores the prepared dictionary in an explanatory word dictionary storing part 4. A development part 6 develops a retrieving word inputted by a retrieving word input part 5 to a synonym and a logical expression based on the contents of the dictionary 2 and the contents of the explanatory word dictionary stored in the storing part 4. A retrieval part 8 obtains all documents including any of the retrieving word inputted to the input part 5 and the synonym as the contents of the documents as retrieved results. When the retrieving word is developed by the development part 6, all documents having paragraphs satisfying the developed logical expression are obtained as retrieved results. These retrieved results are outputted to a display part 9.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an electronic document search apparatus.

[0002]

2. Description of the Related Art As a search method for an enormous amount of documents, a search method using a keyword is generally used. When an arbitrary keyword (search term) is input as a search condition into the search system, all documents including the search term in the document content are obtained as search results. The search by this method is called a full-text search. A method is also widely used in which a keyword for search is added to each document in advance, and a document to which a keyword matching the input search word is added is used as a search result. The keyword added to each document may be automatically extracted from the document, or may be manually determined.

In these search systems, a document containing a word completely matching the search word entered by the user or a word completely matching the keyword entered by the user is added as a keyword for search. Only those documents that are found can be obtained as search results. Therefore, in such a search system, since a complete match between the search word and the keyword is required, it cannot be said that all documents required by the user can be obtained comprehensively. That is, the retrieval system has a relatively high relevance but a low recall. Here, the relevance is the ratio of documents desired by the user to the total number of documents obtained as search results, and the recall is the number of documents obtained as search results to the total number of documents desired by the user Is the ratio of

In order to improve the recall while keeping the relevance at a high value, there is a method in which the concept of a synonym is used to develop a search word input by a user or a keyword to be added to a document. Used.
For example, if the search term entered by the user is a "dictionary", the synonym dictionary is used to expand the dictionary into "dictionaries", "letters", "letters", "encyclopedias", etc. Search for documents that contain any of the words. Or, if the keyword to be added to the document is "dictionary", "dictionary"
In addition, all of the expanded words are added as keywords, and when any of the words is input as a search word, these documents are searched. By doing so, the aim is to improve the recall.

Tetsuro Ito, "Information Search", Shokodo (198
As noted in 5), what is emphasized in this scheme is the interchangeability (interchangeability).
(synonyms). Two words are interchangeable when the meaning of the context does not change even if the words are replaced with other words in the context in which each word appears.

[0006] The above method makes it possible to improve the recall without lowering the precision. However, the vocabulary of words that are interchangeable with certain words is limited. For this reason, a method for further improving the recall has been proposed. For example, the method described in Japanese Patent Application Laid-Open No. 7-210568 is intended to improve the recall by expanding a search term input by a user using a similar word dictionary (thesaurus) and searching. It is.
For example, if "pulp" is input as a search term and the search fails, the search term is expanded into broader terms such as "raw material" using a thesaurus, and a search is performed using the expanded term. As described in this document, the words to be developed are not limited to similar words and broader terms, but may be lower words, antonyms, and opposite words.

Japanese Patent Laid-Open Publication No. Hei 6-274541 discloses an improvement in the recall rate by retrieving a search term input by a user into related terms based on the co-occurrence relation of the terms. It is described that it is intended. For example, when “mathematics” is input as a search condition, the word is expanded into related words (associative words) such as “equation” and “geometry” based on the co-occurrence relation, and a search is also performed using the expanded words.

However, in the above-described development using the thesaurus and co-occurrence relation, the development may be performed to a word semantically significantly different from the word input by the user (a word having no exchangeability). Many. For example, in the above example, "pulp" and "raw material" are semantically different words. Therefore, there are many documents not intended by the user in the search result, and although the recall is improved, the relevance is significantly reduced.

[0009]

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and it is an object of the present invention to provide a document retrieval apparatus in which the recall is improved without lowering the relevance of retrieval. Is what you do.

[0010]

According to a first aspect of the present invention, there is provided a document search apparatus, comprising: a document storage means for storing a plurality of documents; and at least a search word for searching a desired document from the document storage means. Search word input means for inputting a word, a synonym dictionary describing word synonymous relations, and a dictionary storage means storing the contents of a word dictionary describing word meanings, and the synonym dictionary in the dictionary storage means Dictionary creation means for creating an explanatory word dictionary that describes each word and a plurality of word pairs that explain the word based on the description content of the word dictionary;
Expansion means for expanding the search word input by the search word input means into a set of words expressing similar meanings to the search word using an explanatory word dictionary created by the dictionary creation means;
And a search unit for obtaining a set of documents from the document storage unit based on the set of words expanded by the expansion unit.

According to a second aspect of the present invention, in the document retrieval apparatus, a keyword acquiring means for acquiring a keyword corresponding to each of a plurality of documents, a synonym dictionary describing the synonymous relation of the words, and a word Dictionary storage means for storing the contents of a word dictionary in which the meaning of the word is described, and a plurality of words for explaining each word based on the description contents of the synonym dictionary and the word dictionary in the dictionary storage means. A dictionary creating means for creating an explanatory word dictionary describing the pair, and a set of keywords expressing similar meanings to the keywords obtained by the keyword obtaining means using the explanatory word dictionary created by the dictionary creating means. Expanding means for expanding, and for each document, a set of keywords obtained from the keyword obtaining means and a set of keywords obtained from the expanding means are stored as a pair with the document. Document storage means, search word input means for inputting at least a search word for searching for a desired document from the document storage means, and the search word input by the search word input means and stored in the document storage means And a retrieval unit for comparing a set of keywords or a set of keywords to obtain a set of documents from the document storage unit.

According to a third aspect of the present invention, in the document search device according to the first or second aspect, the set expanded by the expansion means is a logical expression including a word or a keyword. It is.

[0013]

FIG. 1 is a block diagram showing a first embodiment of a document search apparatus according to the present invention. In the figure, 1 is a Japanese language dictionary, 2 is a synonym dictionary, 3 is an explanatory word dictionary creation unit, 4 is an explanatory word dictionary storage unit, 5 is a search word input unit, 6 is a development unit, 7 is a document storage unit, 8 is A search unit, 9 is a display unit, and 10 is a dictionary storage unit. The dictionary storage unit 10 stores the contents of the Japanese language dictionary 1 and the synonym dictionary 2. The national language dictionary 1 is the text content of the digitized Japanese language dictionary itself, and provides a description equivalent to a headword using a word other than the headword for each headword. The description is assumed to be expressed in a natural language here. Of course, the content of the Japanese language dictionary 1 may be a Japanese language dictionary, or a dictionary specialized in a specialized field such as a science and technology term dictionary or an economic term dictionary. Alternatively, some dictionaries may be used in combination. The synonym dictionary 2 is a list of a set of words that can be exchanged, and is stored in the form of digitized text.

The explanatory word dictionary creating section 3 creates an explanatory word dictionary based on the contents of the Japanese language dictionary 1 and the contents of the synonym dictionary 2 stored in the dictionary storage section 10. Explanatory word dictionary storage unit 4
Stores the contents of the explanatory word dictionary created by the explanatory word dictionary creating unit 3. Here, the explanatory word dictionary is a dictionary having a pair of a certain word and a set of words explaining the word. FIG. 2 is an explanatory diagram of an explanatory word dictionary according to the present invention, and FIG.
Is an explanatory diagram of a conventional similar word dictionary. A conventional similar word dictionary (thesaurus) is, as shown in FIGS. 3A and 3B,
By defining the correspondence between two terms of a word, a word tree structure or a network structure is constructed as a whole. On the other hand, the explanatory word dictionary in the present invention is
As shown in FIG. 2, the correspondence between a word and a set of words is defined. In the explanatory word dictionary, there is no correspondence between words and two terms. For example, as shown in FIG.
"Similar word" has a synonymous relationship only with the word set of "word, meaning, same", and includes two terms such as "similar word" and "word", "similar word" and "meaning", and "word" and "meaning". No correspondence between terms is specified.

As described above, in the expansion of a search word or an additional keyword based on the correspondence between two terms, expansion to a word having no synonymous relationship is performed except for expansion to a word having interchangeability. , The precision is reduced. By expanding a search word or an additional keyword into a set of words, it is possible to maintain a synonymous relationship that cannot be expressed by a single word, and it is possible to improve the recall without lowering the precision. The first embodiment shows a case where a search word is expanded into a set of words, and a second embodiment described later shows a case where an additional keyword is expanded into a set of words.

The expansion using the explanatory word dictionary must not simply decompose a word and expand it into a set of a plurality of words. For example, instead of simply decomposing “climbing” into “plane” (plate) and “climbing” and associating them, they are associated with a set of words such as “baseball, pitcher, participation”. Only by expanding to a set having semantically distant words as constituent elements, it is possible to realize a significant improvement in recall.

Returning to FIG. 1, the search word input unit 5 is a user interface that allows a user to input a word (search word) or a search formula including the word (search word) as search conditions for performing a search. is there. The development unit 6 stores the dictionary storage unit 1
0 and the contents of the explanatory word dictionary stored in the explanatory word dictionary storage means 4,
The search term input to the search term input unit 5 is expanded into a synonym and a logical expression. The search word, the search word and the logical expression obtained as the expansion result are notified to the search unit 8.

The document storage means 7 stores a plurality of documents to be searched in the form of digitized text. The search unit 8 specifies a document serving as a search result from the document storage unit 7 based on the search word obtained from the expansion unit 6, a word having the same meaning as the search word, and a logical expression. The display unit 9 is a user interface for displaying a document specified by the search unit 8 to a user as a search result.

FIG. 4 is a flowchart showing an example of the processing procedure of the explanatory word dictionary creating section 3. Explanation word dictionary creation unit 3
Is basically a process of associating each headword in the Japanese language dictionary 1 with a set of independent words appearing in the description of the headword as a description vocabulary. In particular, the Japanese language dictionary provides a description equivalent to a headword for each headword using words other than the headword. Therefore, this data can be said to be the most suitable data for obtaining a set of words having synonymous relationships (explanatory vocabulary). It is assumed that M in the drawing is a real constant that satisfies 0 <M <1.

First, in S21, a headword and its description are extracted from the Japanese dictionary 1, and in S22, the head headword of the Japanese dictionary 1 is set to A. In S23, it is determined whether or not the explanatory sentence corresponding to the headword A has already been subjected to morphological analysis processing. If the explanatory sentence has been subjected to morphological analysis processing, it has already been acquired as an explanatory vocabulary in the processing of another headword. Therefore, the process proceeds to S30 and proceeds to the processing of the next headword. If the morphological analysis processing has not been performed yet, the morphological analysis processing is performed on the explanatory sentence corresponding to the headword A in S24 to extract an independent word.

In S25, the synonym dictionary 2 describes the word synonymous with the headword A.
It is determined whether or not there is a word that is a headword of. If such a word exists, in S26, the morpheme is added to the description corresponding to all the headwords having the same meaning as the headword A (these are referred to as headwords A1, A2, ..., An). Analyze and extract independent words. Then, in S27, the independent words of the headword A extracted in S24 and the headwords A1, A extracted in S26.
2,. . . , An, among the independent words in the description corresponding to
All of the independent words appearing at least (n + 1) × M times are headwords A and A1, A2,. . . , An, respectively. However, if a synonym exists in the extracted independent word, it is counted as the same word. In this case, the word used as the explanatory vocabulary may be any one of the synonyms.

At S25, a word synonymous with the headword A does not exist in the synonym dictionary 2, or the synonym dictionary 2
If it does not exist as a headword in the national language dictionary 1, the set of independent words obtained in S24 is used as an explanatory vocabulary for the headword A in S28. However, when a synonym exists in the extracted independent word, any one of the synonyms may be set as one of the explanatory vocabulary.

In S29, the headword A is entered in the Japanese language dictionary 1.
It is determined whether or not the next headword exists, and if so, the process returns to S23 with the next headword as a new headword A in S30, and the process is repeated up to the last headword. .

Hereinafter, an example of the operation of the first embodiment of the present invention will be described using a specific example. FIG. 5 is an explanatory diagram of an example of the contents of the synonym dictionary, and FIG. 6 is an explanatory diagram of an example of a morphological analysis result of the contents of the Japanese language dictionary. Here, as shown in FIG. 5, the synonym dictionary 2 describes four words of “synonyms”, “similar words”, “synonyms”, and “synonyms” as synonyms. It is assumed that there is a description that "meaning" and "significance", "word" and "word", and "equivalent" and "equal" and "same" are synonyms. Further, it is assumed that the national language dictionary 1 describes, as headwords, “synonyms”, “similar words”, “synonyms”, and “synonyms”, and explanatory sentences corresponding to the respective headwords. FIG. 6 shows the independent words extracted from the explanatory sentences corresponding to these headwords.
It is assumed that M = 0.8.

In the explanatory word dictionary creating unit 3, the processing shown in FIG. 4 is performed, and when "synonym" is selected as the headword A, the morpheme is added to the explanatory sentence corresponding to the headword "synonym" in S24. An analysis process is performed to extract independent words such as “meaning”, “equivalent”, “word”, and “synonym” as shown in the first line of FIG. In S25, referring to the contents of the synonym dictionary 2 shown in FIG. 5, "similar words", "synonyms", and "synonyms" are obtained as words synonymous with the headword "synonyms". Since these words are the headwords of the Japanese language dictionary 1 as shown in FIG. 6, the explanation sentence corresponding to these words is subjected to morphological analysis processing in S26, and the independent words as shown in the second to fourth lines in FIG. Get. Then, by collecting independent words that appear (3 + 1) × 0.8 = 3.2 times or more from these, “words” are obtained. At this time, from the contents of the synonym dictionary 2 shown in FIG. 5, since “meaning” and “significance” are synonymous, the number of appearances is added to four, and either “meaning” or “significance” is obtained. . Similarly,
Since "equal", "equal", and "same" are synonymous, adding the number of appearances results in four times, "equal", "equal",
Either "same" is obtained. By such processing, the explanatory vocabulary for each of the terms “synonym”, “similar word”, “synonym”, and “synonym” is “word”, “meaning”,
Or "words", "meaning", "equals" or "words"
"Significance", "same", etc.

In the expansion unit 6, based on the contents of the synonym dictionary 2 and the contents of the explanatory word dictionary created as described above, the search words input to the search word input unit 5 are converted into synonyms and logical expressions. Expand to Here, it is assumed that the search word input to the search word input unit 5 is W. The description vocabulary of the search word W is W1, W
, Wn, and the synonym of the explanatory vocabulary W1 is W1
, W12, ..., the synonyms of the explanatory vocabulary W2 are W21,
W22,...,...
1, Wn2,..., The search word W is (W1 or W11 or W12 or...)
and (W2or W21 or W22 or
・ ・ ・) And ・ ・ ・ and (Wn or W
n1 or Wn2 or ...). If there is no explanatory vocabulary for W, no expansion is performed.

In the above example, when "similar word" is input as a search word, for example, "word", "meaning", "same", etc. are obtained as the description vocabulary of "similar word". From the contents of the synonym dictionary 2 as shown in FIG. 5, for example, (word or word) and (meaning or meaning)
and (equal or same or equal).

[0028] The search unit 8 converts any one of a document including the search word W input to the search word input unit 5 as a document content and a synonym of the search word W input to the search word input unit 5 into the document content. Are all search results. Furthermore, if the expansion of the search word has been performed by the expansion unit 6, all documents having a predetermined unit of text that satisfies the expanded logical expression are set as search results. The predetermined unit may be, for example, a sentence unit, a paragraph unit, a chapter unit, or the like. For example,
When a paragraph is a unit, the input search word W is
At (W11 or W12 or ...) a
nd (W21 or W22 or ...) and
... and (Wn1 or Wn2 or ...
···, when expanded to W11, W12, ...
, W21, W22,..., And Wn1, Wn2,. I do. When a search expression including a search word is input from the search word input unit 5, such a search is performed for each search word, and a given logical operation may be performed to obtain a search result. .

In the above specific example, the expansion unit 6 (word or word) and (meaning or meaning) and (equivalent or same or) from the search word "similar word"
Equal). The search unit 8 searches for a document including the search term “similar word” and its synonyms “synonym”, “synonym”, and “synonym”, and includes a paragraph satisfying the expanded logical expression. Search for documents. For example, a document including a paragraph "This section considers the classification of words. In 1.1, classification was performed on condition that they have the same meaning ...." is obtained by retrieval. Although this document does not include words such as "similar words" and "synonyms", it can be said that this document is substantially written for "similar words". In this way, the recall can be improved.

It should be noted that the explanation word dictionary can be created by the explanation word dictionary creating section 3 for all headwords of the national language dictionary 1 in advance as described above. Then, it is also possible to configure so that an explanatory word dictionary is created for a necessary word based on the search word.

FIG. 7 is a block diagram showing a second embodiment of the document search apparatus of the present invention. In the figure, the same parts as those in FIG. Reference numeral 11 denotes a keyword acquisition unit. In the above-described first embodiment, a case has been described in which a search word is expanded into a set of words. In the second embodiment, a keyword added to a document is expanded into a set of words. Shows the case.

The keyword acquiring section 11 presents the document stored in the document storing section 7 to the user, and receives a keyword to be added from the user. Alternatively, the document is analyzed and keywords are automatically extracted. Also in this case, the user may be able to add or delete a keyword. The expansion unit 6 is obtained from the keyword acquisition unit 11 based on the contents of the synonym dictionary 2 stored in the dictionary storage unit 10 and the contents of the explanatory word dictionary stored in the explanatory word dictionary storage unit 4. Expand keywords into synonyms and logical expressions. The keyword obtained from the keyword obtaining unit 11 and the synonyms and logical expressions obtained as a result of the development are stored in the document storage unit 7 in pairs with the corresponding documents. Document storage unit 7
Stores a plurality of documents to be searched in the form of digitized text. At this time, a keyword obtained from the expansion unit 6, its synonym, and a logical expression are added to each document. The search unit 8 compares a search word or a logical expression in the search expression obtained from the search word input unit 5 with a keyword added to each document stored in the document storage unit 7 and its synonym and logical expression. Then, a document serving as a search result is specified.

FIG. 8 is an explanatory diagram of an example of the data structure of the document storage unit in the second embodiment of the document search device of the present invention. In the second embodiment, the keyword acquired by the keyword acquiring unit 11 is expanded by the expanding unit 6 into a logical expression as described in the first embodiment, and the keyword is stored in a pair with the corresponding document. Stored in the unit 7. Therefore,
Each document has (W11 or W12 or ...) and
(W21 or W22 or ...) and
.. And (Wn1 or Wn2 or...) Are also added.

In the example shown in FIG.
“Similar words” or “synonyms”, “synonyms”, and “synonyms” are “(word or word) and (meaning or
Meaning) and (equivalent or same or equal) "
And stored as one of the keywords. Similarly, in Document 3, “special material” is changed to “(excellent o
r excellence or excellence or excellence) and (talent or ingenuity or intellect).

The search unit 8 includes a word or a logical expression in the search expression input from the search word input unit 5 and a keyword (synonym of keyword) added to each document stored in the document storage unit 7. Are compared with each other by a normal search method, and all corresponding documents are set as search results. When a logical expression is included in the search expression, the logical expression is expanded in accordance with De Morgan's law, and a set of and connections of words is connected or connected, that is, (w11 and w12 and.・ ・) Or
(W21 andw22 and ...) or
・ ・ Or (wm1 and wm2and ・ ・
・)

A set of m and connected words of the converted logical expression (w11 and w12 and ...) (w21 and w22 and ...) (wm1 and wm2 and ...) Paying attention to any of these, when there is a document A satisfying the following conditions, the document A is set as a search result.
Now, one of the logical expressions added to the document A in the document storage unit 7 is (W11 or W12 or ...) and
(W21 or W22 or ...) and
.. And (Wn1 or Wn2 or...), And focus on the set of the i-th word (wi1 and wi2 and...) In the converted m logical expressions. N sets of (W11, W12,...) (W21, W22,...) (Wn1, Wn2,...) In the logical expression added to the document A , Wi2,... Exist as elements, and if wi1, wi2,.
If there are words that are not included in any of the n sets in the logical expression added to the document A, words that are not included in the n sets are included in the document When the keyword added to A is equal to the keyword or its synonym, document A is regarded as one of the search results.

As a specific example, when the data stored in the document storage unit 7 has the contents shown in FIG.
When the search word "word" is input, the search is performed according to a normal search method because it is not a logical expression, and "document 2" is a search result. The logical expression of the document A shown in FIG.
r word) and (meaning or meaning) and (equivalent or same or equal) "also include the search word" word ", but" meaning "connected with and
Alternatively, since “significance”, “equal”, “same” or “equal” does not exist in the search condition, document 1 is not searched.

For example, a logical expression "(searcher or computer) and excellent an"
When "d talent" is input, first, an equivalent logical expression (searcher and excellent and talent) or
(Computer and excellence and talent). About the set of m and connected words of the converted logical expression (searcher and excellent and talent) (computer and excellent and talent), first, paying attention to the word set (searcher, excellent and talent), it is added to the document. Assuming that any of the words “searcher”, “excellent”, and “talent” exists as an element in all of the n sets in the logical expression, the logical expression “(excellent or honor or Excellence or excellence) and (talent or ingenuity or intellect). That is, among the logical expressions in Document 3, (excellent o
r excellence or excellence or excellence), there is “excellence” in the attention (searcher, excellence, talent),
And (Talent or Talent or Talent) means "Talent"
Exists. However, since this logical expression does not include “searcher”, it is determined whether or not the keyword added to the document 3 is equal to the keyword or a synonym thereof. However, since "searcher" is not equal to the keyword of document 3 or its synonym, document 3 is not considered as a search document at this time.

Focusing on the next word set (computer, excellence, talent), the logical expression of the document 3 also applies in this case. Although the logical expression of this document 3 does not include “computer”, since this word has been registered as a keyword of document 3, document 3 satisfies the condition. Therefore, document 3 is obtained as a search result.

[0040]

As is apparent from the above description, according to the present invention, a search word input by a user or a keyword to be added to a document is converted into a set of words (logical expression) by using an explanatory word dictionary. The effect is that the recall can be improved without lowering the relevance of the search.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing a first embodiment of a document search device of the present invention.

FIG. 2 is an explanatory diagram of an explanatory word dictionary according to the present invention.

FIG. 3 is an explanatory diagram of a conventional similar word dictionary.

FIG. 4 is a flowchart illustrating an example of a processing procedure of an explanatory word dictionary creating unit 3.

FIG. 5 is an explanatory diagram of an example of the contents of a synonym dictionary.

FIG. 6 is an explanatory diagram of an example of a morphological analysis result of the contents of a Japanese language dictionary.

FIG. 7 is a configuration diagram showing a second embodiment of the document search device of the present invention.

FIG. 8 is an explanatory diagram illustrating an example of a data structure of a document storage unit according to the second embodiment of the document search device of the present invention.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Japanese language dictionary, 2 ... Synonym dictionary, 3 ... Explanatory word dictionary preparation part, 4 ... Explanatory word dictionary storage part, 5 ... Search term input part, 6 ... Development part, 7 ... Document storage part, 8 ... Search part, 9 display unit, 10
... Dictionary storage unit, 11 ... Keyword acquisition unit.

Claims (3)

[Claims] A document storage unit for storing a plurality of documents;
A search word input unit for inputting at least a search word for searching for a desired document from the document storage unit; a synonym dictionary describing synonymous relations of words; and a word dictionary describing word meanings of words are stored. Dictionary storing means, and dictionary creating means for creating an explanatory word dictionary that describes pairs of each word and a plurality of words that explain the word based on the description contents of the synonym dictionary and the word dictionary in the dictionary storing means Expanding means for expanding the search word input by the search word input means into a set of words expressing similar meanings to the search word using an explanatory word dictionary created by the dictionary creating means; A document search apparatus, comprising: a search unit that obtains a set of documents from the document storage unit based on a set of words expanded by the expansion unit. 2. A keyword acquiring means for obtaining a keyword corresponding to each of a plurality of documents, a synonym dictionary describing a synonym of the word and a word dictionary describing the meaning of the word. Dictionary creation means for creating a dictionary storage means for creating and a description word dictionary describing pairs of each word and a plurality of words explaining the word based on the description contents of the synonym dictionary and the word dictionary in the dictionary storage means Means, and expanding means for expanding into a set of keywords expressing similar meanings to the keyword obtained by the keyword obtaining means using the explanatory word dictionary created by the dictionary creating means,
A document storage unit for storing, for each document, a set of keywords obtained from the keyword obtaining unit and the keywords obtained from the expanding unit in a pair with the document; and searching for a desired document from the document storage unit. Search word input means for inputting at least a search word, and comparing the search word input by the search word input means with a keyword or a set of keywords stored in the document storage means, A document search device comprising a search means for obtaining a set of documents. 3. The document search apparatus according to claim 1, wherein the set expanded by the expansion unit is a logical expression including a word or a keyword.