JP2002024280A - Device and method for document retrieval - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a document retrieving device which retrieves a document having contents close to a reference document from plural object documents to be retrieved. SOLUTION: This document retrieving device 10 is equipped with an evaluated value DB 43 stored with keywords included in the reference document and evaluated values showing how often the keywords are included uniquely in the reference document and obtains a total value by totalizing the evaluated values of the keywords included in the object documents to be retrieved according to the evaluated value DB 43 and divides the total value by the number of the keywords included in the documents to be retrieved to find the document evaluated values of the documents to be retrieved. The document evaluated values are compared with a previously set reference value and an object document having a document evaluated value larger than the reference value is extracted as a document having contents close to the reference document.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a document search apparatus and a document search method for searching a plurality of search target documents for documents having contents close to a reference document serving as a reference.

[0002]

2. Description of the Related Art Conventionally, there has been a keyword search as a method of searching for a desired document from a plurality of search target documents.
The keyword search is a method of searching for a document including an input keyword from a plurality of search target documents. According to this method, a user can search for a desired document from a plurality of search target documents by inputting a keyword predicted to be included in the desired document.

[0003]

However, the keyword search has the following problems. That is,
When expressions are not unified in a document to be searched, a search by a keyword may cause a search omission due to a problem of spelling of a notation. In addition, although a keyword is selected by a user, it is difficult to predict a keyword included in a desired document, and in some cases, a desired document cannot be searched by a keyword search.

Accordingly, the present invention has been made to solve the above-mentioned problems, and to provide a document retrieval apparatus and a document retrieval method for retrieving a document having a content close to a reference document as a reference from a plurality of retrieval target documents. Aim.

[0005]

SUMMARY OF THE INVENTION A document search apparatus according to the present invention is a document search apparatus for searching a plurality of search target documents to be searched for documents having contents close to a reference document as a reference. An evaluation value storage unit that stores a plurality of keywords included in the document and an evaluation value indicating the degree to which each keyword is uniquely included in the reference document; and all words included in the plurality of search target documents as search words. Search word extraction means to be extracted, search word storage means for storing the search word extracted by the search word extraction means in association with a search target document code for specifying the search target document to be extracted, and search word storage means The search word is compared with the keyword stored in the evaluation value storage means. If the search word matches the keyword, the keyword is added to the search word. Evaluation value totaling means for assigning an evaluation value of the search word, and totalizing the evaluation value given to the search word based on the search target document code to calculate a total value for each search target document; and Document evaluation value calculating means for calculating the document evaluation value of each search target document by dividing the total evaluation value of each search target document by the number of keywords included in the search target document; A search document extracting unit that compares a document evaluation value calculated by the unit with a preset reference value and extracts a search target document having a document evaluation value larger than the reference value.

As described above, the search word extracting means
By extracting all search words from the search target document and comparing the extracted search words with the keywords included in the reference document, all search words that match the keywords included in the reference document are searched from the search target documents. Can be extracted. Then, based on the evaluation values stored in the evaluation value storage unit, the evaluation values of the keywords included in each search target document are totaled, and the total value is divided by the number of keywords included in each search target document, The document evaluation value of each search target document is calculated. Then, by comparing the document evaluation value with a preset reference value, a search target document including a keyword unique to the reference document can be extracted. As a result, it is possible to search for a search target document having contents similar to the reference document.

In the above-described document search apparatus, it is preferable that the search word extracting means does not redundantly extract the same search word from one search target document.

[0008] In the above-mentioned document retrieval apparatus, the same keyword may not be stored in the evaluation value storage means in duplicate. By avoiding duplication of keywords stored in the evaluation value storage means in this way, it is possible to optimize the time required for comparison with the search word stored in the search word storage means.

In the above document search apparatus, the keywords are extracted as hiragana, punctuation marks, special symbols and spaces included in the reference document as delimiters, and the search words are extracted as hiragana, punctuation marks and special symbols included in the search target document. And spaces may be extracted as delimiters.

Hiragana, punctuation, special symbols, and spaces are included irrespective of the contents of the document. Therefore, by not extracting them as keywords or search words, the accuracy of document search can be improved. .

[0011] Further, the document search device includes a keyword extracting means for extracting all words as keywords from at least one reference document, a keyword storage means for storing the keywords extracted for each reference document by the keyword extracting means, Reference word extraction means for extracting all words as reference words from a plurality of reference documents to be referenced for determining the evaluation value of a keyword, and keyword and reference word extraction means extracted for each reference document by the keyword extraction means All the word storage means for storing the reference words extracted for each reference document, and for each keyword, the number of keywords included in the keyword storage means is divided by the number of reference documents to obtain the keyword appearance rate in the reference document. Means for calculating the keyword appearance rate in the reference document to be calculated; For each keyword, the keyword appearance rate in all documents is calculated by dividing the total number of keywords included in the all word storage means and the number of reference words identical to the keywords by the total number of reference documents and reference documents. A keyword appearance rate calculating means for all documents; and an evaluation value calculating means for calculating an evaluation value of the keyword by dividing the keyword appearance rate in the reference document by the keyword occurrence rate in all documents. The obtained evaluation value may be stored in the evaluation value storage means.

As described above, the evaluation value obtained by dividing the keyword appearance rate in the reference document by the keyword appearance rate in all documents including the reference document and the reference document indicates that the keyword is included in the reference document. A value indicating the degree of unique inclusion. That is, the appearance rate of keywords in the reference document indicates how frequently each keyword appears in the reference document, and the appearance rate of keywords in all documents indicates the degree of appearance of each keyword in all documents. Indicates whether it appears at a frequency of As a result, a keyword that frequently appears in the reference document and is not often found in all documents can be determined to be unique to the reference document. Conversely, a keyword that frequently appears in the reference document but frequently appears in all documents can be determined to be a word that frequently appears in a general document instead of being unique to the reference document. Therefore, the degree to which a keyword is uniquely included in the reference document can be quantified as (evaluation rate of keyword in reference document) / (keyword appearance rate in all documents) and used as an evaluation value.

[0013] Further, the document search device includes a keyword extracting means for extracting all words as keywords from at least one reference document, a keyword storage means for storing the keywords extracted for each reference document by the keyword extracting means, A reference word extracting means for extracting all words as reference words from a plurality of reference documents to be referred to for determining an evaluation value of a keyword, and a reference word extracted for each reference document by the reference word extracting means. A reference word storage unit, a reference document keyword appearance rate calculation unit for calculating a keyword appearance ratio in the reference document by dividing the number of keywords included in the keyword storage unit by the number of reference documents for each keyword; Keyword included in the reference word storage means for the keyword Means for calculating the keyword appearance rate in the reference document by dividing the number of reference words identical to the number of reference documents by the number of reference documents, and the keyword appearance rate in the reference document as the keyword appearance rate in the reference document. Evaluation value calculation means for calculating an evaluation value of the keyword by dividing by a number obtained by adding a predetermined constant, wherein the evaluation value calculated by the evaluation value calculation means is stored in the evaluation value storage means. Is also good.

As described above, the evaluation value obtained by dividing the keyword appearance rate in the reference document by the number obtained by adding a predetermined constant to the keyword appearance rate in the reference document is determined by the keyword unique to the reference document. A value indicating the degree of inclusion. That is, the appearance rate of keywords in the reference document is
It indicates how frequently each keyword appears in the reference document, and the appearance rate of the keyword in the reference document indicates how often each keyword appears in the reference document. Thus, a keyword that frequently appears in the reference document and is not often found in the reference document can be determined to be unique to the reference document. Conversely, a keyword that frequently appears in the reference document but frequently appears in the reference document can be determined to be a word that is not unique to the reference document but frequently appears in a general document. Therefore, the degree to which the keyword is uniquely included in the reference document can be quantified as the evaluation value by (keyword appearance rate in the reference document) / (keyword appearance rate in the reference document + constant). Further, since a predetermined constant is added to the keyword appearance rate in the reference document, the denominator is prevented from being 0 even when the keyword does not appear in the reference document, and the evaluation value can always be calculated.

In the above-mentioned document search apparatus, it is preferable that the keyword extracting means does not redundantly extract the same keyword from one reference document.

Further, in the above document search device, it is preferable that the reference word extracting means does not redundantly extract the same reference word from one reference document.

A document retrieval method according to the present invention is a document retrieval method for retrieving a document having a content close to a reference document serving as a reference from a plurality of documents to be retrieved. An evaluation value storing step of storing a keyword and an evaluation value indicating the degree to which each keyword is uniquely included in the reference document; and a search word extracting step of extracting all words included in a plurality of search target documents as search words And a search word storing step of storing the search word extracted in the search word extracting step in association with a search target document code for specifying the search target document of the extraction source; and the search word and the evaluation value stored in the search word storage step Match the keyword stored in the storage step and match the search word with the keyword An evaluation value aggregation step of assigning a keyword evaluation value to a search word, and calculating the aggregation value for each search target document by totalizing the evaluation values assigned to the search word based on the search target document code, and an evaluation value aggregation step A document evaluation value calculating step of calculating the document evaluation value of each search target document by dividing the total evaluation value of each search target document calculated in the above by the number of keywords included in the search target document; A search document extracting step of comparing the document evaluation value calculated in the evaluation value calculation step with a preset reference value, and extracting a search target document having a document evaluation value larger than the reference value. .

As described above, in the search word extracting step, all search words are extracted from the document to be searched, and the extracted search words are compared with the keywords included in the reference document, so that the search words are included in the reference document. All search words that match the keyword to be searched can be extracted from the search target document. Then, based on the evaluation values stored in the evaluation value storage step, the evaluation values of the keywords included in each search target document are totaled, and the total value is divided by the number of keywords included in each search target document, The document evaluation value of each search target document is calculated. Then, by comparing the document evaluation value with a preset reference value, a search target document including a keyword unique to the reference document can be extracted. As a result, it is possible to search for a search target document having contents similar to the reference document.

In the above-described document search method, it is preferable that the search word extracting step does not redundantly extract the same search word from one search target document.

Further, in the above document search method, it is preferable that the evaluation value storing step does not store the same keyword repeatedly. By avoiding storing the same keyword in the evaluation value storing step in this way, the time required for matching with the search word can be optimized.

In the above document search method, the keywords are extracted as hiragana, punctuation marks, special symbols and spaces included in the reference document as delimiters, and the search words are extracted as hiragana, punctuation marks and special symbols included in the search target document. And spaces are preferably extracted as delimiters.

Hiragana, punctuation, special symbols, and spaces are included irrespective of the contents of the document. Therefore, by not extracting these as keywords or search words, the accuracy of document search can be improved. .

Further, in the above-mentioned document search method, a keyword extracting step of extracting all words from at least one reference document as keywords, and a keyword storing the keywords extracted for each reference document in the keyword extracting step in a keyword storage means. A storage step, a reference word extraction step of extracting all words as reference words from a plurality of reference documents to be referred to for determining an evaluation value of the keyword, and a keyword extracted for each reference document in the keyword extraction step. An all-word storage step of storing the reference words extracted for each reference document in the reference word extraction step in the all-word storage means, and for each keyword, dividing the number of keywords included in the keyword storage means by the number of reference documents In the reference document A keyword appearance rate calculation step in the reference document for calculating the word appearance rate, and for each keyword, the number of keywords included in the all word storage means and the sum of the number of reference words identical to the keywords are determined by the number of reference documents and reference documents. And calculating the keyword appearance rate in all documents by dividing by the total of the keywords, and calculating the keyword evaluation value by dividing the keyword appearance rate in the reference document by the keyword appearance rate in all documents. An evaluation value calculation step may be further provided, and the evaluation value calculated in the evaluation value calculation step may be stored in the evaluation value storage step.

As described above, the evaluation value obtained by dividing the keyword appearance rate in the reference document by the keyword appearance rate in all the documents including the reference document and the reference document is calculated as follows. A value indicating the degree of unique inclusion. That is, the appearance rate of keywords in the reference document indicates how frequently each keyword appears in the reference document, and the appearance rate of keywords in all documents indicates the degree of appearance of each keyword in all documents. Indicates whether it appears at a frequency of As a result, a keyword that frequently appears in the reference document and is not often found in all documents can be determined to be unique to the reference document. Conversely, a keyword that frequently appears in the reference document but frequently appears in all documents can be determined to be a word that frequently appears in a general document instead of being unique to the reference document. Therefore, the degree to which a keyword is uniquely included in the reference document can be quantified as (evaluation rate of keyword in reference document) / (keyword appearance rate in all documents) and used as an evaluation value.

Further, in the above-mentioned document search method, a keyword extracting step of extracting all words from at least one reference document as keywords, and a keyword storing the keywords extracted for each reference document in the keyword extracting step in a keyword storage means. A storage step, a reference word extraction step of extracting all words as reference words from a plurality of reference documents to be referenced for determining the evaluation value of the keyword, and a reference extracted for each reference document in the reference word extraction step A reference word storage step of storing words in the reference word storage means, and for each keyword in the reference document calculating the keyword appearance rate in the reference document by dividing the number of keywords included in the keyword storage means by the number of reference documents Keyword appearance rate calculation step , For each keyword,
Calculating the keyword appearance rate in the reference document by dividing the number of reference words identical to the keywords contained in the reference word storage means by the number of reference documents; An evaluation value calculating step of calculating an evaluation value of the keyword by dividing by a number obtained by adding a predetermined constant to the keyword appearance rate in the reference document, and storing the evaluation value calculated in the evaluation value calculating step in an evaluation value storing step May be stored.

As described above, the evaluation value obtained by dividing the keyword occurrence rate in the reference document by the number obtained by adding a predetermined constant to the keyword occurrence rate in the reference document is determined by the keyword unique to the reference document. A value indicating the degree of inclusion. That is, the appearance rate of keywords in the reference document is
It indicates how frequently each keyword appears in the reference document, and the appearance rate of the keyword in the reference document indicates how often each keyword appears in the reference document. Thus, a keyword that frequently appears in the reference document and is not often found in the reference document can be determined to be unique to the reference document. Conversely, a keyword that frequently appears in the reference document but frequently appears in the reference document can be determined to be a word that is not unique to the reference document but frequently appears in a general document. Therefore, the degree to which the keyword is uniquely included in the reference document can be quantified as the evaluation value by (keyword appearance rate in the reference document) / (keyword appearance rate in the reference document + constant). Further, since a predetermined constant is added to the keyword appearance rate in the reference document, the denominator is prevented from being 0 even when the keyword does not appear in the reference document, and the evaluation value can always be calculated.

In the above-mentioned document search method, it is preferable that the keyword extracting step does not duplicately extract the same keyword from one reference document.

In the above-described document search method, it is preferable that the reference word extracting step does not redundantly extract the same reference word from one reference document.

[0029]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of a document retrieval apparatus according to the present invention will be described below in detail with reference to the drawings.
In the description of the drawings, the same elements will be denoted by the same reference symbols, without redundant description.

First, the document retrieval apparatus 10 according to the present embodiment
The document retrieval device 10 is a device that retrieves a document having contents close to a reference document serving as a reference from a plurality of documents to be retrieved.

FIG. 1 is a block diagram showing the configuration of the document search device 10. The document search apparatus 10 includes an input unit 20 for inputting a document, a word extracting unit 30 for extracting words from the document, and various databases 40 for storing the words extracted by the word extracting unit 30.
Further, the document search device 10 includes an evaluation value totalizing unit 54, a document evaluation value calculating unit 55, and a search document extracting unit 56 for searching a document having a content close to the reference document from a plurality of search target documents. Have. Further, the document search device 10 includes a keyword appearance rate calculation means 51 in the reference document, a keyword appearance rate calculation means 52 in all documents, and an evaluation value calculation means 53 for calculating the evaluation value.

The input means 20 includes a reference document input means 21 for inputting a reference document, a reference document input means 22 for inputting a reference document, and a search target document input means 23 for inputting a search target document. And here,
The reference document is a document that is referred to when setting an evaluation value for a keyword included in the reference document. In the document search device 10 according to the present embodiment, the reference document input unit 2
1. Reference document input means 22, search target document input means 23
However, these are different from each other only in a document to be input and have the same function of inputting a document. Therefore, these input means may be integrated into one. Specific examples of the input unit include a scanner that reads a document written on paper and a disk drive that reads a document stored in a file.

The word extracting unit 30 extracts a word from the reference document as a keyword.
And a reference word extracting unit 32 for extracting a word from the reference document as a reference word, and a search word extracting unit 33 for extracting a word from the search target document as a search word. In the document search device 10 according to the present embodiment, the keyword extracting unit 31 and the reference word extracting unit 3
2. A search word extraction unit 33 is provided, but these are different only in the document from which the word is extracted, and have the same function of extracting a word from the document. It may be integrated into one. Each of the word extracting means has a function of extracting words in a document by using hiragana, punctuation marks, special symbols, and spaces as delimiters. In addition, all word extraction means do not extract duplicate words from one document,
Words cut out from a document are collated with words already cut out from the same document, and have a function of extracting only words that do not match.

Next, each database 40 will be described. The document search device 10 includes a keyword database (hereinafter, referred to as “keyword DB”) 41, an all-word database (hereinafter, referred to as “all-word DB”) 42, and an evaluation value database (hereinafter, referred to as “evaluation value DB”) 43.
And a search word database (hereinafter, “search word D
B ”) 44).

The keyword DB 41 stores keywords extracted from the reference document. The keyword is stored in association with a reference document code that specifies the reference document from which the keyword is extracted. FIG. 4 shows an example of the keyword DB 41.

The all-word DB 42 stores the keywords extracted from the reference document and the reference words extracted from the reference document. The keyword and the reference word are stored in association with a reference document code that specifies the reference document from which each is extracted and a reference document code that specifies the reference document. FIG. 5 shows an example of the all-word DB 42.

The evaluation value DB 43 stores evaluation values indicating the degree to which keywords contained in the reference document are unique to the reference document. FIG. 7 shows an example of the evaluation value DB 43.

The search word DB 44 stores search words extracted from documents to be searched. The search word is stored in association with a search target document code that specifies a search target document as an extraction source. FIG. 8 shows an example of the search word DB 44.

Next, means for searching for a desired document from the search target document will be described.

The evaluation value totalizing means 54 has a function of adding and totaling the evaluation values of all keywords included in one search target document. Then, the tally value obtained by tallying is input to the document evaluation value calculation means 55.

The document evaluation value calculation means 55 has a function of calculating the document evaluation value by dividing the total value calculated by the evaluation value totalization means 54 by the number of keywords included in the search target document. Then, the calculated document evaluation value is input to the search document extracting means 56.

The search document extracting unit 56 extracts a desired document from a plurality of search target documents by comparing the document evaluation value calculated by the document evaluation value calculation unit 55 with a preset reference value. Having. Specifically, when the document evaluation value is larger than the reference value, a document having the document evaluation value is extracted.

Next, means for calculating the evaluation value of the keyword will be described.

Means of calculating keyword appearance rate in reference document 51
Has a function of calculating an appearance rate at which a common keyword appears in each of a plurality of reference documents. When there are M reference documents and a common keyword exists for A of them, the keyword appearance rate in the reference document is calculated by A / M. The document search device 10 according to the present embodiment searches for a keyword stored in the keyword DB 41, calculates how many identical keywords exist, and divides the calculated number of keywords by the number of reference documents. , Calculate the keyword appearance rate in the reference document. When the number of reference documents is one, keywords are extracted without duplication. Therefore, the keyword appearance rate in the reference document is 1 for each keyword.

Means for calculating keyword appearance rate in all documents 52
Has a function of calculating an appearance rate at which a keyword common to all documents including the reference document and the reference document appears. When there are M reference documents and N reference documents, and there are B keywords common to them, the keyword appearance rate in all documents is calculated as B / (M + N). The document search device 10 according to the present embodiment searches for keywords and reference words stored in the entire document DB, and calculates how many identical keywords and the same reference words exist. Here, the “reference word” is a name conveniently added to a word extracted from the reference document, and thus “the same reference word as the keyword” means a keyword included in the reference document. The keyword appearance rate in all documents is calculated by dividing the calculated number of keywords by the number of all documents.

The evaluation value calculating means 53 has a function of calculating the evaluation value of the keyword by dividing the keyword appearance rate in the reference document by the keyword appearance rate in all documents.

Next, the operation of the document search apparatus 10 of the present embodiment will be described, and a document search method according to the embodiment will be described.

Here, a case will be described as an example in which a patent publication having contents close to the reference patent publication is searched from a plurality of patent publications to be searched. FIG. 2 is a flowchart showing the operation of the document search device 10.

As shown in FIG. 2, the document search by the document search apparatus 10 includes an evaluation value calculation step (S10 to S18) for calculating an evaluation value of a keyword included in the reference document, and a document search based on the evaluation value. Document search stage for searching (S20)
To S26). If the evaluation value of the keyword has been calculated in advance, the document is searched for (S10) without going through the evaluation value calculation steps (S10 to S18).
20 to S26).

First, the evaluation value calculation stage (S10 to S18)
Will be described. First, the keyword extracting means 31
Extracts a keyword from the reference document (S10). FIG. 3 is a diagram illustrating an example of the reference document. When this reference document is read from the beginning of the sentence using hiragana, punctuation marks, special symbols, and spaces as delimiters, words such as "task", "light-shielding colored layer", "formation", etc. are sequentially extracted as keywords. Here, the word “light-shielding colored layer” is also present in the second line of the reference document shown in FIG. 3, but since “light-shielding colored layer” has already been extracted as a keyword, it is not redundantly extracted. . In this way, all words are extracted as keywords from the reference document, and the extracted keywords are stored in the keyword DB 41 in association with the reference document code specifying the reference document (S1).
0). FIG. 4 is a diagram illustrating an example of the keyword DB 41. Here, keywords are extracted from the three reference documents B001 to B003.

Next, the reference word extracting means 32 extracts a reference word from the reference document (S12). In the extraction of the reference word, hiragana, punctuation, special symbols, and spaces are extracted as delimiters in the same manner as the extraction of the keyword.
Then, the extracted reference word is stored in the all-word DB 42 in association with the reference document code specifying the reference document (S12). The keyword extracted by the keyword extracting means 31 is also stored in the all-word DB 42. FIG. 5 is a diagram illustrating an example of the all-word DB 42.
Here, reference words are extracted from the 97 reference documents R001 to R097.

Next, the reference document keyword appearance rate calculating means 51 calculates the keyword appearance rate in the reference document based on the keyword DB 41 (S14). A specific example will be described with reference to FIG. Referring to the keyword DB 41 shown in FIG. 4, the keyword “assignment” is included in any of the reference documents B001 to B003.
The number of keywords in the reference document is three. Since the number of reference documents is three, the keyword appearance rate in the reference document is 1. Similarly, as shown in FIG. 6, the number of keywords in the reference document and the keyword appearance rate in the reference document are calculated. Here, the case where the number of reference documents is three is described. However, when the number of reference documents is one, both the number of keywords in the reference document and the number of reference documents are one.
Both the number of keywords in the reference document and the keyword appearance rate in the reference document are always 1 both.

Further, the keyword occurrence rate calculation means 52 in all documents calculates the keyword occurrence rate in all documents including the reference document and the reference document by the same method as the keyword occurrence rate in reference document (S16). . That is, the number of keywords included in all documents is changed to the total number of documents (in this case, 10
0), the overall document keyword appearance rate as shown in FIG. 6 is calculated.

Next, the evaluation value calculating means 53 divides the keyword appearance rate in the reference document by the keyword appearance rate in all the documents (keyword appearance rate in the reference document / keyword appearance rate in all documents) to obtain each keyword. Is calculated (S18). FIG. 7 shows the calculated evaluation values.
This evaluation value is a value indicating to what extent the keyword is uniquely included in the reference document. More specifically, the keyword appearance rate in the reference document, which is the numerator of the evaluation value calculation formula, indicates the proportion of documents in which the keyword is included in the reference document. Can be said to depend on the specificity of the keyword. For example, if a document related to a semiconductor is selected as a reference document, terms relating to the semiconductor such as “semiconductor” and “P-type layer” are considered to be included in any reference document, and an important key in document search. Is the word However, since very common terms such as “problem” and “conventional” are also included in the reference document, if the keyword appearance rate in the reference document is used as it is as the evaluation value,
Document search cannot be performed with high accuracy. On the other hand, the keyword appearance rate in all documents, which is the denominator of the evaluation value calculation formula, represents the proportion of documents that include the keyword in all documents, and the magnitude of the keyword appearance rate in all documents is It depends on the generality of the keyword. That is, keywords such as “problem” and “conventional” having a high keyword appearance rate in all documents are keywords used in gazettes in any technical field. Therefore,
By dividing by the keyword appearance rate in all documents so that the evaluation value of such a general keyword does not increase, the evaluation value that balances the uniqueness and generality of the keyword included in the reference document is obtained. .

Next, the document search step (S20 to S26) will be described.

First, the search word extracting means 33 extracts a search word from the search target document, and stores the extracted search word in the search word DB 44 in association with the search target document code for specifying the extraction source search target document (S20). ). FIG. 8 is a diagram showing the search word DB 44.

Next, the search words stored in the search word DB 44 are compared with the keywords stored in the evaluation value DB 43, and the evaluation values of the keywords are totaled for each search target document specified by the search target document code. Yes (S2
2). That is, a search word included in the search word DB 44 is compared with a keyword, an evaluation value of the keyword is assigned to a search word that matches the keyword, and the assigned evaluation value is added to each search target document. To calculate the total value. FIG. 9 shows an example of calculating the total value of the evaluation values.

Subsequently, the evaluation value calculation means 53 calculates the number of keywords included in each search target document. And
The document evaluation value of each search target document is calculated by dividing the total value calculated by the evaluation value totaling means 54 by the number of keywords (S24). For example, as shown in FIG. 9, the search target document code F001 includes 53 keywords, and the total value of the evaluation values is 38.2.
7, the document evaluation value is 0.72.

Next, the document extracting means compares the document evaluation value with a preset reference value, and extracts a document having a document evaluation value larger than the reference value as a search document (S26). As a result, a search target document having contents close to the reference document is searched from among the plurality of search target documents. In FIG. 9, since “2” is set as the reference value, two documents of the search target document codes F007 and F008 are extracted as documents having contents close to the reference document.

Next, the document retrieval apparatus 10 according to the present embodiment
The effect of the document search method will be described.

The document retrieval apparatus 10 according to the present embodiment has an evaluation value DB 43 storing an evaluation value indicating how much a keyword included in a reference document is unique to the reference document.
, And a document is searched based on the evaluation value, so that a document having a content close to the reference document can be extracted from a plurality of search target documents.

The document retrieval device 10 according to the present embodiment
Is the evaluation value obtained by dividing the keyword appearance rate in the reference document, which is the rate at which the keyword appears in the reference document, by the keyword appearance rate in the entire document, which is the rate at which the keyword appears in all the documents combining the reference document and the reference document Is calculated, the evaluation value indicates the degree to which each keyword is uniquely included in the reference document, and a highly accurate document search can be performed by using this evaluation value.

The document search method according to the present embodiment
An evaluation value indicating the degree to which the keywords contained in the reference document are unique to the reference document is calculated for each keyword, and the document is searched based on the evaluation value. Can be extracted from a plurality of search target documents.

Next, a document search device 60 according to a second embodiment of the present invention will be described. The document search device 60 according to the second embodiment includes an input unit 20 for inputting a document, a word extraction unit 30 for extracting words from the document, and various databases 40 for storing the words extracted by the word extraction unit 30. An evaluation value summation unit 54 and a document evaluation value calculation unit 55 for searching for a document having a content close to the reference document from a plurality of search target documents.
And a search document extracting unit 56, the basic configuration is the same as that of the document search apparatus 10 according to the first embodiment.

Next, elements of the document search device 60 according to the second embodiment, which have different functions from those of the document search device 10 according to the first embodiment, will be described.

The reference word database (hereinafter, referred to as “reference word DB”) 46 stores the reference words extracted from the reference documents by the reference word extraction means 32 in association with the reference documents from which the reference words are extracted. FIG. 11 is a diagram illustrating an example of the reference word DB 46. As shown in FIG. 11, unlike the all-word DB 42, the reference word DB 46 does not store keywords extracted from the reference document.

The reference document keyword appearance rate calculating means 61 has a function of calculating an appearance rate at which a keyword common to each reference document appears. If there are N reference documents and B has a common keyword, the keyword appearance rate in all documents is calculated as B / N.

The evaluation value calculating means 62 has a function of calculating the evaluation value of each keyword by dividing the keyword appearance rate in the reference document by a value obtained by adding a predetermined constant to the keyword appearance rate in the reference document. .

Next, the operation of the document search device 60 according to the second embodiment will be described, and a document search method according to the second embodiment will be described.

FIG. 12 is a flowchart showing the operation of the document search device 60.

First, a keyword is extracted from the reference document and stored in the keyword DB 41 (S30). Next, a reference word is extracted from the reference document and stored in the reference word DB (S32). Then, the reference document keyword appearance rate calculation unit 51 calculates the keyword appearance rate in the reference document based on the keyword DB 41 (S34). Then, the reference document keyword appearance rate calculation means 61 calculates the keyword appearance rate in the reference document based on the reference word DB 46 (S36). FIG. 13 shows an example of calculating the keyword appearance rate in the reference document and the keyword appearance rate in the reference document. Here, the keyword “polydome switch” is included in the reference document but not included in the reference document, so that the keyword appearance rate in the reference document is 0.

Next, the evaluation value calculating means 62 divides the keyword appearance rate in the reference document by a value obtained by adding a predetermined constant (here, 0.01) to the keyword appearance rate in the reference document to obtain each keyword. Was calculated, and FIG.
The evaluation value calculated as shown in FIG. 4 is stored in the evaluation value DB 43 (S38). Here, the reason why the predetermined constant is added is, for example, the keyword “polydome switch” (FIG. 1).
Even if the keyword appearance rate in the reference document becomes 0 as in (3), the evaluation value can be calculated by dividing by a constant. Note that the value of the predetermined constant is preferably set according to the balance with other evaluation values.

Next, in the same manner as described in the first embodiment, based on the evaluation value stored in the evaluation value DB 43, a document having contents close to the reference document is searched from the search target document (S20 to S20). S26).

As described above, by searching for a document based on the evaluation value by dividing the keyword appearance rate in the reference document by the value obtained by adding a predetermined constant to the keyword appearance rate in the reference document, a highly accurate document can be obtained. Search can be performed.

Although the embodiments of the present invention have been described in detail, the present invention is not limited to the above embodiments.

In the above embodiment, when words are extracted from a reference document or a search target document, the words extracted from one document are not duplicated, but the same word is extracted from one document. It is good. This allows
Different weights can be assigned to words that are used repeatedly and words that are not.

In the above embodiment, the search target document and the reference document are described as being different from each other. However, the search target document and the reference document may be the same.

A general word (for example, the keyword “problem” in the above embodiment) that is known in advance to be included in almost all search target documents
May not be evaluated as an evaluation value. That is, the system further includes a general word database storing words included in almost all search target documents, the evaluation value totalizing means totals evaluation values except for words stored in the general word database, and the document evaluation value calculating means Alternatively, of the keywords included in each search target document, the total value may be divided by the number of keywords other than the words stored in the general word database. With such a configuration, the document evaluation value is not diluted by a general word, so that the accuracy of document search can be further improved.

[0079]

According to the present invention, there is provided an evaluation value storage means for storing an evaluation value indicating the degree to which a keyword is uniquely included in a reference document, and the evaluation value stored in the evaluation value storage means and a document to be searched. Since the document evaluation value of each search target document is calculated based on the keywords included in the search target document, it is possible to search the search target document for documents having contents similar to the reference document.

Further, according to the present invention, the keyword appearance rate in the reference document, which is the rate at which the keyword appears in the reference document, is used to calculate the keyword appearance rate, which is the rate at which the keyword appears in all documents including the reference document and the reference document. Since the evaluation value of the keyword included in the reference document is calculated by dividing by the keyword appearance rate, the degree to which the keyword is uniquely included in the reference document can be used as the evaluation value.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a document search device according to an embodiment.

FIG. 2 is a flowchart illustrating an operation of the document search device.

FIG. 3 is a diagram illustrating an example of a reference document.

FIG. 4 is a diagram illustrating an example of a keyword DB.

FIG. 5 is a diagram illustrating an example of an all-word DB.

FIG. 6 is a diagram illustrating an example of calculating a keyword appearance rate.

FIG. 7 is a diagram illustrating an example of an evaluation value DB.

FIG. 8 is a diagram illustrating an example of a search word DB.

FIG. 9 is a diagram illustrating an example of calculating a document evaluation value.

FIG. 10 is a block diagram illustrating a document search device according to a second embodiment.

FIG. 11 is a diagram illustrating an example of a reference word DB.

FIG. 12 is a flowchart illustrating an operation of the document search device according to the second embodiment.

FIG. 13 is a diagram illustrating an example of calculating a keyword appearance rate.

FIG. 14 is a diagram illustrating an example of an evaluation value DB.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 10 ... Document search apparatus, 20 ... Input means, 21 ... Reference document input means, 22 ... Reference document input means, 23 ... Search target document input means, 30 ... Word extraction means , 31 ... keyword extraction means, 32 ... reference word extraction means, 33
..Search word extracting means, 40 ... database, 41 ...
Keyword database, 42 ... All word database, 43 ... Evaluation value database, 44 ... Search word database, 51 ... Keyword appearance rate calculating means in reference document, 52 ... Keyword appearance rate in all documents Calculation means, 5
3 ... Evaluation value calculation means, 54 ... Evaluation value aggregation means, 55 ...
Document evaluation value calculation means, 56... Search document extraction means.

Claims (18)

[Claims] 1. A document search apparatus for searching a plurality of search target documents for documents having contents close to a reference document serving as a reference, comprising: a plurality of keywords included in the reference document; An evaluation value storage unit that stores an evaluation value indicating a degree that a keyword is uniquely included in the reference document; a search word extraction unit that extracts all words included in the plurality of search target documents as search words; A search word storage unit that stores the search word extracted by the search word extraction unit in association with a search target document code that specifies the search target document as an extraction source; and the search word stored in the search word storage unit. The keyword is compared with the keyword stored in the evaluation value storage means, and when the search word matches the keyword, the search is performed. An evaluation value aggregation unit that assigns the evaluation value of the keyword to a word, and aggregates the evaluation value assigned to the search word based on the search target document code to calculate a total value for each search target document; Dividing the total value of the evaluation values in each of the search target documents calculated by the evaluation value totaling means by the number of the keywords included in the search target document, and calculating the document evaluation value of each of the search target documents A document evaluation value calculating unit that calculates the document evaluation value calculated by the document evaluation value calculating unit and a predetermined reference value, and the search target having the document evaluation value that is larger than the reference value A document retrieval apparatus, comprising: retrieval document extraction means for extracting a document. 2. The document search apparatus according to claim 1, wherein the search word extraction unit does not redundantly extract the same search word from one search target document. 3. The document search apparatus according to claim 1, wherein the same keyword is not stored in the evaluation value storage unit in duplicate. 4. The document search according to claim 1, wherein the keyword is extracted by using hiragana, punctuation, special symbols, and spaces included in the reference document as delimiters. apparatus. 5. The method according to claim 1, wherein the search word is extracted by using hiragana, punctuation marks, special symbols, and spaces included in the search target document as delimiters.
A document search device according to any one of the preceding claims. 6. A keyword extracting unit for extracting all words as keywords from at least one of the reference documents; a keyword storage unit for storing the keywords extracted for each of the reference documents by the keyword extracting unit; Reference word extraction means for extracting all words as reference words from a plurality of reference documents to be referred to for determining the evaluation value, and the keyword extracted for each of the reference documents by the keyword extraction means; and All word storage means for storing the reference words extracted for each of the reference documents by the reference word extraction means; and for each of the keywords, the number of the keywords included in the keyword storage means is divided by the number of the reference documents. To calculate the keyword appearance rate in the reference document. Means for calculating a keyword appearance rate in a reference document; and for each of the keywords, the sum of the number of the keywords included in the all word storage means and the number of the reference words identical to the keywords is calculated for the reference document and the reference document. A keyword occurrence rate calculating means for calculating a keyword occurrence rate in all documents by dividing by a sum of the numbers; a keyword occurrence rate in the reference document being divided by the keyword occurrence rate in all documents; 2. An evaluation value calculating means for calculating an evaluation value, further comprising: storing the evaluation value calculated by the evaluation value calculating means in the evaluation value storing means.
The document search device according to any one of claims 1 to 5. 7. A keyword extracting means for extracting all words as keywords from at least one reference document; a keyword storage means for storing the keywords extracted for each of the reference documents by the keyword extracting means; Reference word extraction means for extracting all words as reference words from a plurality of reference documents to be referred to for determining the evaluation value, and the reference words extracted for each of the reference documents by the reference word extraction means Reference word storage means for storing, for each of the keywords, the number of the keywords included in the keyword storage means divided by the number of the reference documents to calculate the keyword appearance rate in the reference document. Rate calculating means, and for each of the keywords, A reference document keyword appearance rate calculating means for calculating a keyword appearance rate in a reference document by dividing the number of the same reference words as the keywords contained in the reference word storage means by the number of the reference documents; And an evaluation value calculating means for calculating the evaluation value of the keyword by dividing the keyword appearance rate by the number obtained by adding a predetermined constant to the keyword appearance rate in the reference document. 2. The calculated evaluation value is stored in the evaluation value storage means.
The document search device according to any one of claims 1 to 5. 8. The document search device according to claim 6, wherein the keyword extracting unit does not redundantly extract the same keyword from one reference document. 9. The document search apparatus according to claim 6, wherein the reference word extracting unit does not redundantly extract the same reference word from one reference document. . 10. A document search method for searching a document having contents close to a reference document serving as a reference from a plurality of search target documents serving as a search target, comprising: a plurality of keywords included in the reference document; An evaluation value storing step of storing an evaluation value indicating a degree that a keyword is uniquely included in the reference document; a search word extracting step of extracting all words included in the plurality of search target documents as search words; A search word storage step of storing the search word extracted in the search word extraction step in association with a search target document code for specifying the search target document of the extraction source, and the search word stored in the search word storage step Collating the keyword stored in the evaluation value storing step with the search word; When the keyword matches, the evaluation value of the keyword is assigned to the search word, and the evaluation value assigned to the search word is tabulated based on the search target document code, and the evaluation value is calculated for each search target document. An evaluation value aggregation step of calculating an aggregation value, and dividing the aggregation value of the evaluation value in each of the search target documents calculated in the evaluation value aggregation step by the number of the keywords included in the search target document. A document evaluation value calculation step of calculating a document evaluation value of the search target document, and comparing the document evaluation value calculated in the document evaluation value calculation step with a preset reference value, and comparing the document evaluation value with the reference value. A document retrieval step for extracting the document to be retrieved having the document evaluation value. 11. The document search method according to claim 10, wherein the search word extracting step does not redundantly extract the same search word from one search target document. 12. The document search method according to claim 10, wherein the evaluation value storing step does not store the same keyword redundantly. 13. The method according to claim 10, wherein the keywords are extracted by using hiragana, punctuation, special symbols, and spaces included in the reference document as delimiters.
3. The document search method according to any one of 2. 14. The search word according to claim 10, wherein hiragana, punctuation, special symbols, and spaces included in the search target document are extracted as delimiters.
14. The document search method according to any one of items 13 to 13. 15. A keyword extracting step of extracting all words as keywords from at least one of the reference documents, and a keyword storing step of storing the keywords extracted for each of the reference documents in the keyword extracting step in a keyword storage unit. A reference word extracting step of extracting all words as reference words from a plurality of reference documents to be referred to for determining the evaluation value of the keyword; and a keyword extracted for each of the reference documents in the keyword extracting step. An all-word storage step of storing the keyword and the reference word extracted for each reference document in the reference word extraction step in an all-word storage means; and for each of the keywords, the keyword included in the keyword storage means Is divided by the number of the reference documents to calculate a keyword appearance rate in the reference document, and a keyword appearance rate in the reference document is calculated. For each of the keywords, the number of keywords included in the all word storage means And calculating a keyword occurrence rate in all documents by dividing a total number of reference words identical to the keyword by a total number of reference documents and the number of reference documents; An evaluation value calculating step of calculating the evaluation value of the keyword by dividing the keyword appearance rate in the document by the keyword occurrence rate in the entire document. The evaluation value calculated in the evaluation value calculating step is further provided. 15. The document search method according to claim 10, wherein the document is stored in the evaluation value storing step. 16. A keyword extracting step of extracting all words as keywords from at least one of the reference documents, and a keyword storing step of storing the keywords extracted for each of the reference documents in the keyword extracting step in a keyword storage unit. A reference word extraction step of extracting all words as reference words from a plurality of reference documents to be referred to for determining the evaluation value of the keyword; and A reference word storing step of storing the reference word in the reference word storage means, and for each of the keywords, dividing the number of the keywords included in the keyword storage means by the number of the reference documents, and generating a keyword in the reference document. Reference statement for calculating the rate Calculating the keyword appearance rate in the reference document by dividing the number of reference words identical to the keyword included in the reference word storage means by the number of reference documents for each of the keywords Calculating the keyword appearance rate in the reference document, and dividing the keyword appearance rate in the reference document by a number obtained by adding a predetermined constant to the keyword appearance rate in the reference document to calculate the evaluation value of the keyword The document search method according to any one of claims 10 to 14, further comprising a calculating step, wherein the evaluation value calculated in the evaluation value calculating step is stored in the evaluation value storing step. . 17. The document search method according to claim 15, wherein the keyword extracting step does not duplicately extract the same keyword from one reference document. 18. The document search according to claim 15, wherein the reference word extracting step does not duplicately extract the same reference word from one reference document. Method.