JPH09101951A - Document retrieving device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To save labor for adding an index by selecting a document at one part of all the documents, calculating and extracting the degree of similarity between the selected document and any strongly related document when narrowing the documents retrieved by a keyword into the target document. SOLUTION: A document retrieving part 4 performs the retrieval of documents from a document data base 3 while using the keyword. Besides, the high-order and low-order bocaburaries of the retrieval keyword are extracted while using a thesaurus table 2, the retrieval of documents is performed, and the results are held in a retrieved result holding part 5. Next, a user interface 1 selects the document of the requested field for narrowing down the retrieved results. Afterwards, a word extracting part 6 extracts words contained in the respective documents in the retrieved result holding part. A similarity degree calculating part 7 outputs the degree of similarity between the selected document and the documents in the retrieved result holding part 5 by mutually comparing the words extracted by the extracting part 6. A document selecting part 8 selects the respective documents out of the holding part 5 from the highest rank of similarity degree, narrows them down and displays the result for a user.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document retrieval apparatus for retrieving a document database using a designated character string, and more particularly to a function for retrieving retrieved documents in more detail.

[0002]

2. Description of the Related Art As a conventional document retrieval apparatus, Japanese Patent Laid-Open No.
No. -10062 is disclosed. In the conventional document database search, a method is used in which a character string is specified at the time of search and a document having the specified character string as an index is output as a search result. Further, in the above-mentioned document, by using the dictionary data that holds the upper-lower relationship in which the character string specified at the time of search is registered in advance, the character string related to the character string specified at the time of retrieval is extracted, It is also possible to search using a character string. As a result, it is possible to perform a search even when the keyword specified at the time of the search is different from the index provided in advance in the document database that is the search target.

[0003]

Generally, in the conventional document database search, in order to search a target document only by designating a keyword, a valid search keyword must be selected. Further, when such a keyword cannot be given at once, it is necessary to use a method of adding a keyword to a document once searched with a certain keyword. There is a problem that the searcher himself must determine what keywords are effective in order to narrow down target documents from such a document database.

For example, first, a document database is searched with a keyword such as "car", and as a result, documents of topics related to cars such as "car development", "car accident", and "car trade problem" are found. It is assumed that it is obtained as a search result. At this time, in the first search stage, in order to obtain information such as what topic was searched as the entire searched document, it was necessary to search the entire document. In order to obtain a topical document such as "development of car", which is the purpose of the search, it is necessary to select an appropriate keyword from the keywords assigned to those documents.

In order for the above search to be effective,
It is necessary to add an index for searching to each document in a large number of document databases to be searched, but a lot of effort is required to add an index to such a large number of documents. There is also the problem that

[0006]

In order to solve the above-mentioned problems, the present invention provides a document search apparatus for searching a document according to a user's purpose from a plurality of documents searched using a designated character string. A means for temporarily holding a plurality of retrieved documents, a means for selecting a document intended by the user from the temporarily held documents, and a document similar to the document selected by the user for the temporary It has a means to search from the sentences stored in.

[0007]

1 is a block diagram of a document retrieval apparatus showing an example of an embodiment of the present invention. 1 is a user interface for inputting a search keyword and selecting an appropriate document from search results, 2 is a thesaurus table that holds dictionary data of upper-lower relationships between vocabularies, and 3 is a document that holds documents to be searched The database 4 is a document search unit that searches for documents in the document database 3 using the search keyword input in the user interface 1 and the thesaurus table 2. Reference numeral 5 is the document search unit 4
A search result holding unit that holds some documents that are the results of the search result; a word extraction unit 6 that extracts words included in each document of the search result documents held in the search result holding unit 5; Reference numeral 7 denotes a similarity calculation unit that calculates the similarity between the document selected by the user interface 1 and each search result document, and 8 indicates the similarity calculation between the document selected by the user interface 1 and each search result document. A document selection unit that narrows down documents using the similarity information calculated by the unit 7.

Next, the operation of the above-described document search device will be described. First, the user inputs a search keyword for searching a document in the document database 3 on the user interface 1. The document search unit 4 searches for a document from the document database 3 using the search keyword input through the user interface 1. Further, the vocabulary that is higher or lower than the search keyword is extracted using the thesaurus table 2, and the documents are searched using these vocabulary as well. Then, the document as the search result is held in the search result holding unit 5.

FIG. 2 is an explanatory diagram showing an example of the search results stored in the search result storage unit 5. In FIG.
Each of documents 11 to 15 is an example of a document obtained as a result of retrieval using the keyword “car”. Next, in order to narrow down the search results, the user selects a document in the field requested by the user from the search results using the user interface 1. This selection is made by displaying a part of the retrieved document to the user.

This selection may be one document or more. However, in order to select a plurality of documents, it is necessary to display many documents, so normally one document is selected. Shall be done. Next, the word extraction unit 6 extracts the words contained in each document of the search result holding unit.

FIG. 3 is an explanatory diagram showing an example of words extracted from the document shown in FIG. In FIG. 3, 21 word strings are extracted from the document 11 in FIG. 2, 22 word strings are extracted from the document 12 in FIG. 2, and 23 word strings are extracted from the document 13 in FIG. 2, the 24 word strings are extracted from the document 14 of FIG. 2, and the 25 word strings are shown in FIG.
An example of what is extracted from the document 15 of FIG. In FIG. 3, words are separated by "/".

Next, the similarity calculation unit 7 compares the words extracted by the word extraction unit 6 with the similarity between the document selected by the user interface 1 and each document in the search result holding unit 5. Done in. As an example of document similarity calculation, a method of counting how many same words as the word in the document specified by the user exist in each retrieved document and treating the numerical value as the document similarity is considered. To be Also, as a semantic relationship between words, K.
W. Church et al .. “Using Statistics in Lexical Anal
ysis ”, Lexicalacquisition: Exploiting on-line resou
rces to build a lexicon. (Zernik Uri (ed.)), London, L
awrence Erlbaum Associates, 1991, pp.115-164, which compares the values of mutual information, calculates between words in each document, and calculates the similarity between documents by summing them. Conceivable.

FIG. 4 is an explanatory diagram showing the similarity and the result of calculation of the similarity with each document in the search result holding portion when the document 11 in FIG. 2 is selected by the user. The calculation of the degree of similarity here uses the former method of the above-mentioned methods of counting how many same words as the word in the document designated by the user exist in each retrieved document.

In FIG. 4, 31 indicates the document 12 of FIG. 2, and 35 indicates the same number of words as the selected document in the document. 32 shows the document 13 of FIG. 2 and 36 shows the same number of words as the selected document in the document. 33 is shown in FIG.
Document 14 and 37 indicates the same number of words as the selected document in the document. 34 indicates the document 15 in FIG.
38 indicates the same number of words as the selected document in the document.

Finally, the document selecting section 8 uses the similarity information of the documents calculated by the similarity calculating section 7 to select each document in the search result holding section 5 in the descending order of similarity. The result is narrowed down and displayed to the user through the user interface 1. For example, when the result of FIG. 4 is used, it is determined that the degree of similarity is high in the order of 33, 31, 32, 34, and for example, when the top 10% is displayed, the document 33 is displayed.

[0016]

As described above, according to the present invention, when the documents temporarily searched by the keyword are narrowed down to the target documents, some of the documents are selected. A document having a strong relationship with the selected document is extracted by calculating the document similarity. With this, in order to narrow down the target documents,
It is not necessary for the searcher to determine which keyword is valid, and it is possible to narrow down the documents to be searched without knowing the keyword.

Further, it is not necessary to add a large number of search indexes to a large number of documents in the document database to be searched, and the labor for adding indexes can be reduced.

[Brief description of the drawings]

FIG. 1 is a block diagram of a document search device showing an example of an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing an example of a search result.

FIG. 3 is an explanatory diagram showing an example of words extracted from a document.

FIG. 4 is an explanatory diagram showing a calculation result of similarity.

[Explanation of symbols]

 1 User Interface 2 Thesaurus Table 3 Document Database 4 Document Search Section 5 Search Result Holding Section 6 Word Extraction Section 7 Similarity Calculation Section 8 Document Selection Section

Claims (2)

[Claims] 1. A document retrieval apparatus for retrieving a document according to a user's purpose from a plurality of documents retrieved using a designated character string, and means for temporarily holding the retrieved plurality of documents. It has means for selecting a document intended by the user from the temporarily stored documents, and means for searching a document similar to the document selected by the user from the temporarily stored sentences. Document retrieval device. 2. The document search device according to claim 1, wherein a means for extracting a word in the temporarily held document and a similarity between the word information and the document selected by the user are obtained. A document retrieving apparatus comprising means and means for retrieving a document similar to the document selected by the user using the similarity information.