JPH02125363A - Document retrieving device - Google Patents

Abstract

PURPOSE:To improve the retrieval efficiency by performing fuzzy retrieval and changing the keyword connection based on user's judgement or instruction. CONSTITUTION:A document accuracy calculating part 51 calculates a document accuracy indicating the degree, in which each document in a device satisfies a user's question statement (retrieval formula), based on the retrieval formula transferred from a request processing part 54 and sorts documents in the order of document accuracy and presents them to a user. The user selects a document to be more finely seen from a list of document titles and document accuracy of retrieval results and sees document contents. The user refers to contents of the list to judge whether the document which he sees at present satisfies his retrieval formula or not. Thus, the keyword connection is changed by learning based on user's judgement to reflect the judgement on retrieval results at the time of next retrieval.

Description

[Detailed description of the invention] [Industrial application field] The present invention relates to a document retrieval device, and particularly to a document retrieval device.

When searching for documents.

The present invention relates to a flexible and high-speed document retrieval device in which each document in a search result is ordered by an evaluation value having a larger value as the documents are closer to a user's search request, and furthermore, the ordering can be changed according to the user's judgment and instructions.

[Conventional technology]

In conventional document search devices, when registering a document, a registration operator selects and registers keywords that are considered appropriate, and when searching for documents, the search device user specifies keywords that are considered appropriate from a thesaurus. Is going. According to this method, high-speed search is possible.

[Problem to be solved by the invention]

The conventional device described above requires storage capacity for keyword registration, and there is a problem with the validity of the keywords selected by the registration operator. Furthermore, the classification and updating work using a thesaurus is complicated, and there are also problems with their validity.

Further, since conventional document search devices only check whether the search conditions specified by the user are satisfied, it is not possible to conduct an ambiguous search regarding the extent to which the search conditions are satisfied. Therefore.

The search must be performed again to find a document that almost, but not completely, satisfies the search conditions specified by the user. Another problem is that if there is no document that satisfies the search conditions specified by the user, it is extremely troublesome to search for the document closest to the search conditions.

As a method to solve these problems, a method of expressing the relationship between keywords numerically has been considered. However, no specific numerical expression method is provided. Furthermore, since it did not have a learning function, it was insufficient to be used as an actual document retrieval device.

The present invention overcomes these drawbacks of the prior art.

Utilizing the concept of dynamic keyword connection, there is no need for complicated classification work based on thesaurus etc. when registering documents, and when searching documents, it is possible to identify documents with high value such as #1 that are close to the user's search request. The present invention aims to provide a document retrieval device that allows users to flexibly search for documents that meet their requirements based on the degree of document accuracy. In particular, when the search question text is a logical sum of one keyword or multiple keywords, the present invention changes the keyword connection based on the user's judgment/instruction, thereby changing the user's judgment during the first search. An object of the present invention is to provide a document retrieval device equipped with a keyword connection learning method that is reflected in search results.

[Means to solve the problem]

In order to solve the above-mentioned problems, the present invention provides a keyword extraction means for extracting keywords from document information when a registered document and its document information are registered in a file, and a relationship between the registered document and the keyword. An inverted 7 isle creation means for creating an inverted file.

A keyword connection table processing means that creates a keyword connection table that describes the degree of association of related information between keywords, changes the relationship speed value of the related information that has already been recorded, and generates new related information; A document search device comprising a document selection means for selecting a document matching a search formula from the keyword connection table and the inverted file using a keyword.

The document selection means uses a value indicating the strength of the relationship between a specific keyword group and the keyword group of registered documents in each file as document accuracy, and calculates the document accuracy using a keyword connection table, an inverted file, and a predetermined method. The document selection means outputs documents as search results in order of document accuracy when a search formula is input, and when the document accuracy of each document is input. , change the keyword connection by a predetermined method.

[Effect]

The document retrieval device according to the present invention calculates the document certainty that represents the degree to which each document in the device satisfies the search formula for the user's search question sentence (hereinafter referred to as the search formula),
Documents are sorted in descending order of document accuracy and presented to the user. The user can select a document to view in more detail from a list of document titles and document certainty in the search results, and view the contents of one document. From the contents of the list, the user determines whether the document currently being viewed is suitable for his or her search formula. Keyword connections can be changed through learning based on such user judgments, and the judgments can be reflected in the search results at the next search.

〔Example〕

Next, an embodiment of a document retrieval device according to the present invention will be described with reference to the accompanying drawings.

The document search device of the present invention uses keyword connections that describe the relationship between keywords and each registered document. For each document, we introduce an evaluation value called document accuracy, which has a larger value as it is closer to the user's search request, and when searching for documents, users can flexibly select documents that meet their requirements based on the degree of document accuracy. select.

A keyword connection describes a relationship between keywords. For keyword connections, the magnitude of association is in the range 0 to l.

0 represents that there is no relationship between the keywords, a value of 0 or more represents that there is a relationship, and 1 represents that the magnitude of the relationship is maximum.

Keyword connections can also be viewed as a two-dimensional array. In that case, the relationship velocity between the i-th keyword and the j-th Keyard is expressed as Wi(K j ) or more simply Wij.

FIG. 1 shows the configuration of a document retrieval device according to the present invention.

When the keyword extraction unit 10 inputs the registered document 11,
Extract the keyword, keyword and registered document 1
1, a document information management section 20, a keyword connection table processing section 30. It is output to the inverted file creation section 40.

Document management section 20, keyword connection table processing section 30
．． The inverted file creation section 40 has files 21, 31, and 41, respectively. The document information management unit 20 stores keywords and bibliographic information in a file 21, and creates a database in a format that can be used during a search.

The keyword connection table management unit 20 creates a keyword connection table that describes necessary keywords and keyword connections (hereinafter referred to as related information), and stores it in the file 31. Furthermore, if there is a request from the request processing unit 54, the weight of the related information is changed.

The inverted file creation unit 40 describes the relationship between keywords and documents, and stores it in a file 41.

The document selection section 50 consists of sections 51 to 55.

The document accuracy calculation unit 51 calculates the document accuracy for each document based on the search formula transferred from the request processing unit 54 and using necessary information in the keyword connection table stored in the file 31. The calculation method will be explained in detail later.

The document sorting unit 52 arranges the documents stored in the file 21 in descending order of document accuracy calculated by the document accuracy calculation unit 51 and transfers them to the 1 display management unit 55.

The display management section 55 displays search results according to the user's instructions given from the request processing section 53.

The learning management unit 54 performs keyword connection learning according to the user's instructions given from the request processing unit S3. The learning management unit 54 calculates how much each weight of the keyword connection should be changed, and the actual value change is realized by giving an instruction to the keyword connection table processing unit 30. The learning method will be explained in detail later.

The request processing unit 53 transfers the received search expression to the document certainty calculation unit 51 in order to search for a document matching the search request. At the same time, if necessary, the keyword connection table processing unit 30 is instructed to change the weight of related information regarding the final keyword group.

The purpose of document search processing is to search and display appropriate documents according to user requests. The document selection process is performed by the document selection unit sO.

Search conditions are performed by specifying keywords and setting conditions for other bibliographic information. First, it is determined whether condition settings other than keywords for each document are satisfied. If the conditions are met, the document accuracy of the document is calculated; if the conditions are not met, the document accuracy of the document is not calculated and the value is set to 0.

Keywords are specified using keyword expressions that express the information the user wants to obtain. A keyword expression is a combination of one keyword or multiple keywords using OR (sum operation). A combination of multiple keywords is used when a single keyword expressing the information the user is trying to obtain does not exist in the document management device.

Next, a method of calculating each document accuracy for the keyword expression Query will be described (the actual calculation is performed by the document accuracy calculation unit in the document selection unit 50).

The algebraic sum e is used in the document accuracy calculation according to the present invention.

xFBy=x+y-xy=1 (1x) (1-y) Also, ΦXj=X1ΦX2e -eXn=1-77(1-Xj
)j.

The document accuracy is a numerical value representing the strength of the relationship between each document and the search formula specified by the user. Since the keyword connection learning method of the present invention is intended for the case where the search expression is the logical sum of one or more keywords, the calculation method in that case will be explained. For each keyword, we calculate the algebraic sum of the set of keywords included in the keyword connection weight search formula. next,
A set K of keywords included in the resulting m-th document
Find the algebraic sum regarding L (m) and use it as the document accuracy.

Here, 1 is an element of the combination of keywords K L (m) for the m-th document of interest, and j is an element of the set of keywords QUERY included in the search expression.

Furthermore, by using the relationship of equation (A2), it can be modified as follows.

Once the document accuracy calculation unit 51 of the document selection unit 50 calculates the document accuracy, next, the document sorting unit 5 of the document selection unit SO
In step 2, documents are sorted in descending order of document accuracy. and,
The results are sent to the display management section 55 and displayed to the user.

A keyword connection to which the keyword connection learning method according to the present invention can be applied is one in which each weight takes a value within the range of θ to 1, and the weight is fixed to 1 when i = j.

The learning method is as follows. An evaluation function E is set in advance to evaluate how close the search results are to the user's judgment, that is, to quantify the difference between the document accuracy of a document and the user's judgment. The evaluation function in this learning method will be specifically shown later. In the learning of keyword connections, the weight of the keyword Lanexi 1 is changed so as to reduce the value of the evaluation function E just defined.

Wi j 4-Wi j+αΔWij where α is a positive constant and is a learning coefficient that determines the learning speed. The amount of change ΔWij is determined based on the steepest descent method. That is, how the evaluation function E is given in the two learning methods 1 and 2, and how the change amount Nij of keyword connections is given based on it will be explained in detail below.

(Learning method 1) In learning method 1, the user determines whether a certain document is appropriate or inappropriate for a search expression. The evaluation function E is the square of the difference between the file accuracy of a certain document and the user's judgment (1 is given to a document that is appropriate, and O is given to a document that is inappropriate). is given as follows.

Here, FCm is the document certainty of the m-th document.

In the following, the calculation formula for the change amount and lfwij is different depending on (1) when the document is determined to be appropriate and (2) when the document is determined to be inappropriate, so the calculation formula will be explained in detail in two cases.

■ When the document is judged to be appropriate First, the partial differential of the evaluation function E with respect to the weight Wij is (A4)
Here, k is an element of the set of keywords K L (m) for the m-th document of interest, and 1 is the set of keywords included in the search formula QUER? Take the elements of

However, combinations in which (k*t) becomes (i, j) are excluded.

In Equation (6), it can be rewritten as in the case of Wij Fengl, and the amount of calculation of partial differentials can be reduced. Wi
When j = 1, the amount of change is positive, and on the other hand, the weight of the keyword connexin must be in the range from 0 to 1, so the value cannot actually be changed. Therefore, it becomes.

■ When a document is judged to be inappropriate First, which index E1 should be given a partial differential by the weight WiJ?The partial differential of FCm by Wij should be given by equation (6). However, Wij =
In the case of 1, you must calculate without omitting it this time.

therefore, becomes.

Now, as described above, AVij can be found in each case of ■ and ■. The actual weight needs to be changed only when AVij is not 0, and the weight value is fixed to l when i = j, so the flowchart for learning keyword connections is as shown in Figure 2 and the third factor. Become.

According to this learning method, keyword connections can be changed based on the user's judgment and instructions, and those judgments and instructions can be reflected in the search results at the time of the next search. Furthermore, it can be searched at high speed, making it suitable for interactive processing. According to this method, keyword connections can be improved little by little during daily searches.

(Learning method 2) In learning method 2, the user determines whether all documents are appropriate or inappropriate for the search expression, document by document.

The evaluation function E is FCm, which quantifies the file accuracy of a certain document and the user's judgment. The sum of the squares of the differences over all documents is given as follows.

In fact, when providing fCm, the user only needs to tell the search device what is appropriate. Further, m takes an element in the set ALLDOCO of all documents.

First, the partial differential of the evaluation function E with respect to the weight Wij and the partial differential of FCm with respect to Wij are given by the above equation (6). (
According to equation 6), if the keywords in the m-th document do not include the first keyword, the partial differential is 0, so in equation (2), m takes the element of the set of all documents. A set of documents containing the first keyword DO
You can do this by taking the elements of C(l.

Wi j 'x; In the case of 1, equation (7) can be rewritten.

When Wij=1, it must be calculated without omission. Therefore, , : becomes.

Seat 1j is determined as described above. Actual weight changes only need to be made when #lj is not 0, and l
In the case of =j, the weight value is fixed to l, so the flowchart of keyword connection learning is as shown in FIGS. 4 to 6.

In the case of this learning method as well, keyword connections can be changed based on the user's judgment and instructions, and those judgments and instructions can be reflected in the search results at the time of the next search. Processing using this method is relatively slow, so it is suitable for patch processing and is used when initializing keyword connections.

〔Effect of the invention〕

According to the present invention, vague searches are possible, keyword connections can be changed based on the user's judgments and instructions, and the judgments and instructions can be reflected in the search results at the time of the next search. As a result, documents that the user has needed in the past are ranked higher, making the search more efficient.

Conversely, documents that the user has deemed unnecessary in the past are ranked lower, making the search more efficient. Furthermore, keywords that are not used by the user can be removed as unnecessary, and the storage device can be used efficiently.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of a document search device according to the present invention; FIGS. 2 and 3 are flowcharts showing a learning method performed in the document search device according to the present invention; FIGS. 4 and 5; FIG. 6 is a flowchart showing another learning method performed using the document search device according to the present invention. Explanation of symbols of main parts 1G...Keyword extraction section 11...Registered document 12...Bibliographic information 20...Document information management section 21.31.41...File 30...Keyword connection table processing Part 40...
Inverted file creation section 50...Document selection section 51...Document accuracy calculation section 52...Document sorting section 53...Request processing section 54...Learning management section 55...Display management section O ...Keyboard 0...Display

Claims (1)

[Claims] 1. Keyword extracting means for extracting keywords from the document information when a registered document and its document information are registered in a file; and an inverted keyword indicating a relationship between the registered document and the keyword. an inverted file creation means that creates a file; and a keyword connection table that describes the degree of association of related information between the keywords, changes the value of the degree of association of related information that has already been recorded, and generates new related information. A document retrieval device comprising: a keyword connection table processing means for generating a keyword connection table; and a document selection means for selecting a document matching a search formula from the keyword connection table and the inverted file according to an input keyword, the document selection means comprising: A document whose document accuracy is calculated using the keyword connection table, the inverted file, and a predetermined method, with document accuracy being a value indicating the strength of association between a specific keyword group and a keyword group of registered documents in each file. The document selection means includes a probability calculation means, when the search formula is input, the document selection means outputs documents as search results in order of document probability, and when the document probability of each document is input. , a document retrieval device characterized in that a keyword connection is changed by a predetermined method. 2. Keyword extraction means for extracting keywords from the document information when a registered document and its document information are registered in a file; and an inverted file for creating an inverted file indicating the relationship between the registered document and the keywords. a keyword connection table that creates a keyword connection table that describes the degree of association of related information between the keywords, changes the value of the degree of association of related information that has already been recorded, and generates new related information; In a document retrieval device comprising a processing means and a document selection means for selecting a document matching a search condition from the keyword connection table and the inverted file according to an input keyword, the document selection means selects a document matching a specific keyword group and each document. A value indicating the strength of association between a keyword group of a registered document in a file is defined as document accuracy, and the document accuracy calculation means calculates document accuracy using the keyword connection table, the inverted file, and a predetermined method. , wherein the document selection means outputs documents as search results in order of document certainty when the search formula and document group are input, and further changes keyword connections according to a predetermined method. Search device.