JPH10289246A - Similar document retrieving device and similar document retrieving method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a similar document retrieving device which efficiently retrieves document that is similar to a retrieval key document even when there are many documents. SOLUTION: This device inputs a retrieval key document from an input device 2, extracts document remarked context information that suggests the contents from each document in an external storage device 4 and the retrieval key document, extracts category remarked context information with a document group of different contents in the device 4 all together and calculates category similarity between the document remarked context information and the category remarked context information that is extracted from the document group of different contents in a lump. It selects a document group that calculates document similarity with the retrieval key document from a document database in accordance with the calculated category similarity and outputs identification information of a document that is retrieved based on each document similarity which is calculated by the document similarity calculating means as a retrieval result to a display device 3.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a similar document search device and a similar document search method.

[0002]

2. Description of the Related Art Conventionally, there has been proposed a similar document search apparatus for extracting a document similar to the contents of a document from a search target document database using the document as a search key. This similar document search device compares a word included in a document serving as a search key with a word included in each document stored in the search target document database, and The similarity with each document stored in the search target document database is calculated, and similar documents are extracted according to the degree of the similarity.

[0003] As a method of calculating the similarity, a document as a search key and the type, appearance frequency, appearance location, and the like of a word contained in each document stored in the search target document database are used. A calculation method using a vector method is employed.

[0004]

However, in the above-described prior art, similarity calculation between each document stored in the search target document database and a document serving as a search key is stored in the search target document database. Therefore, when the number of documents stored in the search target document database is large, the search processing time is inevitably increased.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problem. Even when the number of documents stored in the search target document database is large, the search processing time can be reduced and the search key can be efficiently saved. It is an object of the present invention to provide a similar document search device and a similar document search method capable of searching for a document similar to.

[0006]

According to the first aspect of the present invention,
In a similar document search apparatus for extracting a similar document from a document database in which a document group is stored for each content by using one document as a search key, an input means for inputting a search key document, each document in the document database and the search Document attention context information extraction means for extracting document attention context information suggesting the content from the key document, and category attention context information extraction means for extracting category attention context information by grouping the content-based documents in the document database Category similarity calculating means for calculating category similarity between document attention context information extracted from a search key document and category attention context information extracted from a group of documents for each content; and category similarity calculation means The document similarity between the document key database and the search key document according to the category similarity calculated by A similarity calculation document selection means for selecting a group of documents to be output, and a document group selected based on the document attention context information of the document group selected by the similarity calculation document selection means and the document attention context information of the search key document. Document similarity calculating means for calculating each document similarity, and output means for outputting, as a search result, identification information of a document searched based on each document similarity calculated by the document similarity calculating means. It is assumed that.

According to a second aspect of the present invention, there is provided a similar document search method for extracting a similar document from a document database in which a group of documents is stored for each content by using one document as a search key. Extracting document attention context information that suggests the contents from each document in the database and the search key document, extracting category attention context information as a group of content-specific documents in the document database,
Calculate the category similarity between the document attention context information extracted from the search key document and the category attention context information extracted from the group of documents classified by content, and search the document database according to the calculated category similarity. A document group for calculating the document similarity with the key document is selected, and each document similarity of the selected document group is determined based on the document attention context information of the selected document group and the document attention context information of the search key document. Calculate,
It is characterized in that identification information of a document searched based on each calculated document similarity is output as a search result.

According to a second aspect of the present invention, there is provided a similar document searching method using the configuration of the similar document searching apparatus according to the first aspect of the present invention.
Document attention context information extracting means extracts document attention context information suggesting the contents from each document in the document database and the search key document, and a category attention context information information extraction means extracts a document group for each content in the document database. The category attention context information is extracted as a group.

Further, the category similarity calculating means calculates
Calculate the category similarity between the document attention context information extracted from the search key document and the category attention context information extracted from the group of documents classified by content, and according to the category similarity calculated by the similarity calculation document selection means. A document group for calculating a document similarity with the search key document is selected from the document database, and identification information of the document searched based on each document similarity calculated by the output unit is output as a search result. is there.

This solves the conventional problem that the search processing time increases as the number of documents stored in the document database increases, and the amount of calculation of the document similarity can be reduced regardless of the number of documents. This greatly reduces the efficiency of similar document search by the user.

According to a third aspect of the present invention, there is provided a similar document search apparatus for extracting a similar document from a document database in which a group of documents is stored by content using one document as a search key. Category similarity distribution setting means for setting a category similarity distribution of a search target document; and document attention context information extraction for extracting document attention context information suggesting the content from each document in the document database and the search key document. Means, a category attention context information information extracting means for extracting a category attention context information as a group of documents according to content in the document database, a document attention context information extracted from the search key document, and a content A category similarity calculating means for calculating a category similarity with category attention context information extracted from a document group; Calculating a document similarity belonging to a category including a document similar to a search key document from the document database based on the category similarity calculated by the degree calculating means and the category similarity distribution set by the category similarity distribution setting means. Similarity calculation document selecting means for selecting a group of documents for use, and a document group selected based on the document attention context information of the document group selected by the similarity calculation document selection means and the document attention context information of the search key document. Document similarity calculating means for calculating each document similarity; and output means for outputting, as a search result, identification information of a document searched based on each document similarity calculated by the document similarity calculating means. Is what you do.

According to a fourth aspect of the present invention, there is provided a similar document search method for extracting a similar document from a document database in which a group of documents is stored for each content by using one document as a search key. A category similarity distribution of a target document is set, document attention context information indicating the contents is extracted from each document in the document database and the search key document, and a document group according to the contents in the document database is grouped as a group. Attention context information is extracted, and the category similarity between the document attention context information extracted from the search key document and the category attention context information extracted from the group of documents according to the content is calculated. Based on the set category similarity distribution, documents belonging to a category including a document that is similar to a search key document from the document database The document group for degree calculation is selected, and the document similarity of the selected document group is calculated based on the document attention context information of the selected document group and the document attention context information of the search key document, and the calculated document similarity is calculated. It is characterized in that the identification information of the document searched based on the degree is output as a search result.

A similar document search method according to a fourth aspect of the present invention using the configuration of the similar document search apparatus according to the third aspect of the invention basically has the same operation as the first and second aspects of the invention. In addition to the above, by setting the category similarity distribution of the search target document in advance, it is possible to arbitrarily select the category including the document closest to the search key document and prevent the search document from being omitted. It exerts the effect that the search efficiency can be greatly improved.

According to a fifth aspect of the present invention, there is provided a similar document search apparatus for extracting a similar document from a document database in which a document group is stored for each content by using one document as a search key. A category similarity distribution setting means for setting a category similarity distribution of a search target document; a document number distribution setting means for setting a document number distribution of a search target document; and a document in a document database and the search key document. Document attention context information extraction means for extracting document attention context information suggesting the contents, category attention context information information extraction means for extracting category attention context information by grouping documents according to content in a document database, and retrieval The category of the document attention context information extracted from the key document and the category attention context information extracted from a group of documents classified by content Category similarity calculating means for calculating similarity; and a search key from the document database based on the category similarity calculated by the category similarity calculating means and the category similarity distribution set by the category similarity distribution setting means. A similarity of selecting a document group for calculating a document similarity belonging to a category including a document similar to the document and narrowing down the number of document groups according to the document number distribution of the search target document set by the document number distribution setting means. Document calculation means for calculating each document similarity based on the document attention context information of the document group narrowed down by the similarity calculation document selection means and the document attention context information of the search key document; Output means for outputting, as a search result, identification information of a document searched based on each document similarity calculated by the document similarity calculation means. It is an feature.

According to a sixth aspect of the present invention, there is provided a similar document search method for extracting a similar document from a document database in which a group of documents is stored for each content by using one document as a search key. A category similarity distribution of the target document is set, a document number distribution of the search target document is set, and document attention context information indicating the content is extracted from each document in the document database and the search key document. Category similarity context information extracted from the search key document as a group of document groups according to the content of the document, and the category similarity between the category attention context information extracted from the group of documents according to the content and the category similarity context information Based on the calculated category similarity and the set category similarity distribution, a sentence similar to the search key document from the document database. The document group for calculating the document similarity belonging to the category that includes is selected and narrowed down to the number of documents according to the set document number distribution of the search target documents. Document similarity is calculated based on document attention context information of the document, and the identification information of the document searched based on each document similarity calculated by the document similarity calculation means is output as a search result. Things.

The similar document search method according to the sixth aspect of the present invention using the configuration of the similar document search apparatus according to the fifth aspect of the invention basically has the same operation as the third and fourth aspects of the invention. In addition to the above, the maximum number of documents for calculating the similarity between the search key and each document is determined based on the number of documents stored in the document database or the number of designated documents, and the similarity is calculated based on the maximum number of documents. Can be narrowed down and the number of documents for which the document similarity calculation is performed can be narrowed down, and the similar document search efficiency of the user can be greatly improved.

[0017]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing a configuration of a similar document search apparatus according to the present embodiment. The similar document search device according to the present embodiment includes a control device 1 including a CPU and a memory, an input device 2 such as a keyboard, a display device 3 for displaying search results of similar documents, and a document to be searched. An external storage device 4 for storing data and the like.

FIG. 2 is a block diagram showing a detailed configuration example of the control device 1, the input device 2, and the display device 3.

The control device 1 comprises a control unit and a memory unit.

The control unit performs various controls and processes, and includes a main processing unit 200, an initialization unit 201, and an input unit 20.
2. Output unit 203, category-specific document number setting unit 204, similarity distribution setting unit 205, document number distribution setting unit 206, document database Note Z-th context information extraction unit 207, search key document attention context information extraction unit 208, category Attention context information information extraction unit 209, category similarity calculation unit 210, similarity calculation document selection unit 211, category selection unit 212, category document selection unit 213, document similarity calculation unit 214, category document setting unit 215, search result output A part 216 and the like are provided.

The memory section includes a category-specific document number buffer section 220, a similarity distribution setting buffer section 221, a document number distribution setting buffer section 222, a document database focused context information buffer section 223, a search key document focused context information buffer section 224, Category attention context information buffer unit 225, category similarity buffer unit 226, similarity calculation document buffer unit 227, selected category buffer unit 228, selected document buffer unit 229, document similarity buffer unit 230, category document buffer unit 231, work buffer And the like.

Here, the initialization section 201 initializes each buffer section. The input unit 202 outputs information of a search key document and other various information input by the user from the input device 2 to the work buffer unit 240.

The output unit 203 outputs the contents of the search key document and various settings made by the input unit 202 to the display device 3 for display.

The number-of-categories document number setting unit 204 sets the number of documents for each category in the document database which is classified into categories according to the contents of the documents. The number of documents for each category set here is stored in the category-specific document number buffer unit 220.

A similarity distribution setting unit 205 calculates a similarity between a group of documents belonging to each category and a search key document from a similarity value between the search key document and each category in each document in the document database. Make settings to determine.

In this case, the setting is made so as to calculate the similarity between a group of documents belonging to a category higher than the similarity value specified by the user and the search key document, or the similarity value distribution of all the categories. A category is set so that similarity calculation is performed up to a document group based on statistical information (deviation value or the like). The content set here is the similarity distribution setting buffer unit 221.
Is stored in

The document number distribution setting unit 206 uses the similarity distribution setting unit 205 to determine the similarity between a document group belonging to each category and the search key document based on the similarity value between the search key document and each category in the document database. The category up to the calculation of the degree is determined, and the maximum value of the number of documents for which the degree of similarity to the search key document is calculated is set accordingly.

Even if the similarity distribution setting unit 205 sets a target category for calculating the similarity between a document group belonging to each category and a search key document, the number of documents belonging to that category remains the document number distribution setting unit 206 When the value exceeds the value set by the document number distribution setting unit 206, the category having a low similarity to the search key document is set to a document group belonging to each category so that the number of documents is equal to or less than the value set by the document number distribution setting unit 206. Is excluded from the target categories up to calculation of similarity between the search key document and the search key document. The content set here is stored in the document number distribution setting buffer unit 222.

Document Database Attention Context Information Extraction Unit 20
7 extracts, from each document stored in the document database, context information that can be noticed in representing the content of the document.

The context information in this case includes word type, number of appearances, appearance position, co-occurrence information, and the like.

The context information data is created for each document and stored in the document database attention context information buffer unit 223.

Search key document attention context information extraction unit 208
Extracts, from a search key document input from the input unit 202, context information that can be noticed in representing the contents of the document.

The context information in this case includes word type, number of appearances, appearance position, co-occurrence information, and the like.

The context information data is stored in the search key document attention context information buffer unit 224.

Category attention context information information extraction unit 209
Retrieves the context information extracted by the document database focused context information extracting unit 207 from the document database focused context information buffer unit 223, and also converts the document ID-category information created by the category document setting unit 215 into the category document buffer unit. The context information of each document of the same category is collected and stored in the category attention context information buffer unit 225.

The category similarity calculation unit 210 includes a search key document attention context information buffer unit 224 created by the search key document attention context information extraction unit 208 and a document database attention context information created by the category attention context information information extraction unit 209. The similarity between the search key document and each category is calculated from the document information buffer unit 223. The calculated category similarity value is stored in the category similarity buffer unit 226.
Is stored in

The similarity calculation document selection unit 211 extracts the document ID belonging to the category stored in the selection category buffer unit 228 by referring to the category document buffer unit 231, and extracts the document ID from the similarity calculation document buffer unit 227. Document ID to be selected and selected document buffer unit 22
9 is stored.

The category selection unit 212 calls the category-based similarity value calculated by the category similarity calculation unit 210 from the category similarity buffer unit 226, and specifies the similarity distribution setting buffer unit 221 specified by the similarity distribution setting unit 205. Are stored in the selected category buffer unit 228.

The category document selection unit 213 stores the number of documents for calculating the similarity between the search key document and each document stored in the document database in the category document buffer unit 231. Part 22
The number of documents belonging to the category stored in 8 is referred to the category-specific document number buffer unit 220 to calculate the number of documents for which the similarity is calculated.

If the number of documents does not match the condition of the number of documents stored in the selected document number buffer unit 229, the number of documents stored in the selected category buffer unit 228 is determined to match. Delete the category. The category to be deleted is the category similarity buffer unit 2
26, the categories having the lowest similarity are sequentially deleted.

The document similarity calculation section 214 converts the context information corresponding to the document ID stored in the selected document buffer section 229 into the document database attention context information buffer section 223.
From the context information of each document ID and the context information of the search key document stored in the search key document attention context information buffer unit 224 to obtain a similarity. The similarity with the document is stored in the document similarity buffer unit 230.

The category document setting section 215 is a document I that is uniquely determined for each document file in the document database.
D and a category are set, and the category document buffer unit 23 is set.
1 is stored.

The search result output unit 216 refers to each document similarity value stored in the document similarity buffer unit 230 and outputs the documents with the highest similarity to the display device 3 in order.

Next, a procedure for creating a document database by the present embodiment will be described with reference to FIG. 3, and a similar document search procedure will be described with reference to FIG.

First, the procedure for creating a document database will be described with reference to FIG.

First, the initialization section 201 is started up, and the memory section is cleared (step S101). Then, the category document setting unit 215 starts up, and sets a document ID and a category for the registered document in the document database (step S102). The document ID is for uniquely determining a document in the document database, and there is no duplication. The category is for grouping the documents in the document database for each document content, and the number of categories is determined by the number of classifications of the document content in the document database. One category is assigned to one document. The document ID-category document name (file name) is linked and stored in the category document buffer unit 231 as shown in FIG.

Next, the category-specific document number setting section 204 is activated, the number of categories-specific documents in the document database is set (step S103), and stored in the category-specific document number buffer section 220.

Then, the document database attention context information extraction unit 207 starts up, creates attention context information representing the contents of the document from each document in the document database stored in the external storage device 4, and generates the document database attention context information. The data is stored in the buffer unit 223 (step S104). In the attention context information, in the case of co-occurrence information, for example, in the case of "document relating to the announcement of a new product", "new product ... announcement" has an important word as word information in representing the contents of the document.

This is performed for each document in the document database.
As shown in FIG. 8, it is stored in the document database attention context information buffer unit 223.

Subsequently, the category attention context information extraction unit 20
9 starts, refers to the category document buffer unit 231 and the document database attention context information buffer unit 223, collects the attention context information of the documents belonging to the category for each category, and stores it in the category attention context information buffer unit 225 (step). S105). As shown in FIG. 10, the category attention context information is linked and stored in the category attention context information buffer unit 225. This completes the document database creation procedure.

Next, a similar document search procedure will be described with reference to FIG.

First, the initialization unit 201 is activated to clear the memory unit (step S201). Then, the user selects whether to input a search key document, execute a search, or set a search (step S202).

When a search key document is selected in step S202, the input unit 202 is activated, and the input device 2 inputs the search key document shown in FIG.
A search key document as shown in FIG. 7 is input (step S7).
205), the data of the search key document is stored in the work buffer
0 is stored.

Then, the search key document attention context information extraction unit 208 is activated, calls the search key document data from the work buffer unit 240, and creates attention context information representing the contents of the document. As shown in FIG. It is stored in the key document attention context information buffer unit 224 (step S206).
Then, the process returns to step S102.

When the search setting is selected in step S202, the similarity distribution setting unit 205 is activated, and based on the calculation result of the similarity between the search key document and the category, a search is performed for a document belonging to any category. Sets whether to calculate the category similarity with the key document. For example, as shown in FIG. 6, a reference value (for example, a similarity of 0.5) is set when the category similarity between the search key document and each category is to be narrowed down to only a category having a certain reference value or more (step S20).
3). Alternatively, statistics of the category similarity between the search key document and each category may be obtained and set based on the standard deviation or the like. The setting content of the similarity distribution is stored in the similarity distribution setting buffer unit 221 as shown in FIG.

Next, the document number distribution setting unit 206 is activated,
The number of documents for which the document similarity with the search key document is to be calculated (for example, 4000 or less) in case the number of documents cannot be reduced below the expected value even by narrowing down the category similarity between the search key document and each category. ) Is set (step S204).

Data of the set number of documents is stored in the document number distribution setting buffer section 222 as shown in FIG. Then, the process returns to step S102.

If search execution is selected in step S202, the category similarity calculation unit 210 starts up, and
The category similarity between the search key document and each category is calculated with reference to the category attention context information buffer unit 225 and the search key document attention context information buffer unit 224 (step S).
207), and is stored in the category similarity buffer unit 226 as shown in FIG.

Then, the category selection section 212 is activated,
It is determined whether or not the similarity distribution is set in the similarity distribution setting buffer unit 221 (step S208).
If the similarity distribution is set in the similarity distribution setting buffer unit 221 in step S208, the category of the category similarity that matches the condition stored in the similarity distribution setting buffer unit 221 (for example, 2, 4, or 4). Etc.),
As shown in FIG. 13, the data is stored in the selection category buffer unit 228 (step S209).

Next, when the category document selection section 213 is activated and the data of the document number distribution is set in the document number distribution setting buffer section 222 (step S210), the selection category buffer section 228 and the document number by category are set. The buffer unit 220 is called, and the selection category buffer unit 228 is called.
If the total number of documents belonging to each category stored in the category does not match the condition stored in the document number distribution setting buffer 222, the category similarity buffer 226 is called and stored in the selected category buffer 228. The categories having the lowest similarity are sequentially deleted from the selection category buffer unit 228, and the document number distribution setting buffer unit 2 is deleted.
The number of documents meeting the condition stored in the storage 22 is set (step S211).

If no similarity distribution is set in the similarity distribution setting buffer unit 221 in step S208, the category selecting unit 212 is activated, and the category with the highest category similarity is selected in the category similarity buffer unit 226. Then (step S212), it is stored in the selection category buffer unit 228.

Next, the similarity calculation document selection unit 211 is activated, and the document ID corresponding to the category stored in the selection category buffer unit 228 is stored in the category document buffer unit 2.
The selected document buffer unit 229 shown in FIG.
(Step S213).

Next, the document similarity calculating section 214 is activated,
Each document ID stored in the selected document buffer unit 229
Is retrieved from the document database focused context information buffer unit 223, and the document similarity with the search key document focused context information buffer unit 224 is calculated (step S214), and the result is stored in the document similarity buffer. As shown in FIG. 15, “document ID-
And stored as “similarity”.

Next, the document similarity between the documents of all the document IDs stored in the selected document buffer unit 229 and the retrieval key document is calculated, and the document similarity for each document ID is stored in the document similarity buffer unit. 230 (step S21).
4).

Then, the search result output unit 216 starts up,
As shown in FIG. 18, a document having a high similarity stored in the document similarity buffer 230 is set as a result of the search processing, as shown in FIG.
Output from the output unit 203 to the output device 3 (step S2
15).

Finally, it is determined whether or not to continue the similarity search process (step S216). If the process is to be continued, the process proceeds to step S202, and if the process is to be ended, all search processes are ended.

The present invention is not limited to the above embodiment.

[0069]

According to the first and second aspects of the present invention, the amount of calculation of the document similarity is greatly reduced regardless of the number of documents, and the user's similar document search efficiency is greatly improved. And a similar document search method using the similar document search device.

According to the third and fourth aspects of the present invention, the same effects as those of the first and second aspects of the invention can be obtained.
A similar document search apparatus capable of arbitrarily selecting a category including a document closest to a search key document to prevent omission of search and greatly improving similar document search efficiency, and a similar document using the similar document search apparatus Can provide a search method.

According to the fifth and sixth aspects of the invention, the same effects as those of the third and fourth aspects of the invention can be obtained.
It is possible to provide a similar document search device and a similar document search method using the similar document search device, which can greatly improve similar document search efficiency by narrowing down and searching the number of documents for which the document similarity calculation is performed.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a schematic configuration of an apparatus according to an embodiment of the present invention.

FIG. 2 is a block diagram of a control unit of the apparatus according to the embodiment of the present invention.

FIG. 3 is a flowchart showing a procedure for creating a document database according to the embodiment.

FIG. 4 is a flowchart showing a similar document search procedure according to the embodiment.

FIG. 5 is a diagram illustrating an example of storing a document count buffer unit for each category according to the present embodiment;

FIG. 6 is a diagram illustrating a storage example of a similarity distribution setting buffer unit according to the present embodiment;

FIG. 7 is a diagram illustrating an example of storage of a document number distribution setting buffer unit according to the present embodiment;

FIG. 8 is a diagram illustrating an example of storage of a document database attention context information buffer unit according to the present embodiment.

FIG. 9 is a diagram illustrating a storage example of a search key document attention context information buffer unit according to the present embodiment;

FIG. 10 is a diagram illustrating an example of storage of a category attention context information buffer unit according to the present embodiment.

FIG. 11 is a diagram illustrating an example of storage of a category similarity buffer unit according to the present embodiment.

FIG. 12 is a diagram illustrating an example of storing a similarity calculation document buffer unit according to the present embodiment;

FIG. 13 is a diagram illustrating an example of storage of a selected category buffer unit according to the present embodiment.

FIG. 14 is a diagram illustrating an example of storage of a selected document buffer unit according to the present embodiment.

FIG. 15 is a diagram illustrating a storage example of a document similarity buffer unit according to the present embodiment.

FIG. 16 is a diagram illustrating an example of storage of a category document buffer unit according to the present embodiment.

FIG. 17 is a diagram illustrating an input example of a search key document according to the present embodiment.

FIG. 18 is a diagram illustrating an output example of a search result according to the present embodiment.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 control device 2 input device 3 display device 4 external storage device 200 main processing unit 201 initialization unit 202 input unit 203 output unit 204 category-specific document number setting unit 205 similarity distribution setting unit 206 document number distribution setting unit 207 document database attention Context information extraction unit 208 Search key document attention context information extraction unit 209 Category attention context information information extraction unit 210 Category similarity calculation unit 211 Similarity calculation document selection unit 212 Category selection unit 213 Category document selection unit 214 Document similarity calculation unit 215 Category document setting section 216 Search result output section 220 Document number buffer section by category 221 Similarity distribution setting buffer section 222 Document number distribution setting buffer section 223 Document database attention context information buffer section 224 Search key document attention context information buffer section 225 Gori attention contextual information buffer unit 226 category similarity buffer unit 227 similarity calculation document buffer unit 228 selected category buffer unit 229 selected document buffer unit 230 document similarity buffer unit 231 category document buffer 240 work buffer unit

 ──────────────────────────────────────────────────続 き Continuing from the front page (72) Inventor Yukio Nakamoto 1381-1, Shinmachi, Ome-shi, Tokyo Toshiba Computer Engineering Co., Ltd. (72) Takuya Nishina 1381-1, Shinmachi, Ome-shi, Tokyo Toshiba Computer (72) Inventor Naohide Kubota 1381 Shinmachi, Ome-shi, Tokyo Toshiba Computer Engineering Co., Ltd.

Claims (6)

[Claims] 1. A similar document search apparatus for extracting a similar document from a document database in which a document group is stored for each content by using one document as a search key, an input means for inputting a search key document, Document attention context information extraction means for extracting document attention context information suggesting the contents from each document and the search key document, and category attention for extracting category attention context information as a group of content-specific documents in a document database Context information information extracting means; category similarity calculating means for calculating category similarity between document attention context information extracted from a search key document and category attention context information extracted from a group of documents classified by content; According to the category similarity calculated by the category similarity calculating means, a search key document and A similarity calculation document selection means for selecting a group of documents for which the document similarity is to be calculated, and document attention context information of a document group selected by the similarity calculation document selection means and document attention context information of a search key document. Document similarity calculating means for calculating each document similarity of the selected document group; output means for outputting, as a search result, identification information of a document searched based on each document similarity calculated by the document similarity calculating means; A similar document search device comprising: 2. A similar document search method for extracting a similar document from a document database in which a document group is stored for each content by using one document as a search key, wherein a search key document is created, and each document in the document database is created. Document attention context information that suggests the contents is extracted from the search key document, category attention context information is extracted as a group of content-specific documents in the document database, and the document attention context information extracted from the search key document and Calculates the category similarity with the category attention context information extracted from the group of documents according to the content, and calculates the document similarity between the document database and the search key document according to the calculated category similarity Each document of the document group selected based on the document attention context information of the selected document group and the document attention context information of the search key document Calculating a similarity score, outputting each document similarity calculated as a result searches the identification information of documents retrieved based, similar document search method according to claim. 3. A similar document search apparatus for extracting a similar document from a document database in which a group of documents is stored for each content by using one document as a search key, wherein: an input means for inputting a search key document; Category similarity distribution setting means for setting a category similarity distribution; document attention context information extracting means for extracting document attention context information suggesting the contents from each document in the document database and the search key document; Category attention context information information extraction means for extracting category attention context information as a group of documents according to content, document attention context information extracted from a search key document, and a category extracted from a group of content-based documents A category similarity calculating means for calculating a category similarity with the attention context information, and the category similarity calculating means Based on the calculated category similarity and the category similarity distribution set by the category similarity distribution setting means, a document group for calculating a document similarity belonging to a category including a document similar to a search key document from the document database. A similarity calculation document selecting means for selecting a document, and each document similarity of a document group selected based on the document attention context information of the document group selected by the similarity calculation document selection means and the document attention context information of the search key document A similar document characterized by comprising: document similarity calculating means for calculating the document similarity; and output means for outputting, as a search result, identification information of a document searched based on each document similarity calculated by the document similarity calculating means. Search device. 4. A similar document search method for extracting a similar document from a document database in which a document group is stored for each content by using one document as a search key, wherein a search key document is created, and a category similarity of the search target document is created. A distribution is set, and document attention context information suggesting the contents is extracted from each document in the document database and the search key document, and a category attention context information is extracted as a group of documents according to the contents in the document database. Calculates the category similarity between the document attention context information extracted from the search key document and the category attention context information extracted from a group of documents classified by content, and calculates the calculated category similarity and the set category similarity distribution A document group for calculating a document similarity belonging to a category including a document that is similar to a search key document from the document database based on The document similarity of the selected document group is calculated based on the document attention context information of the selected document group and the document attention context information of the search key document, and the document retrieved based on the calculated document similarity Outputting a search result of the identification information of the similar document. 5. A similar document search apparatus for extracting a similar document from a document database in which a document group is stored for each content by using one document as a search key, wherein: an input means for inputting a search key document; Category similarity distribution setting means for setting a category similarity distribution; document number distribution setting means for setting a document number distribution of documents to be searched; documents suggesting the contents from each document in the document database and the search key document Document attention context information extraction means for extracting attention context information; category attention context information extraction means for extracting category attention context information by grouping documents according to content in a document database; and documents extracted from search key documents A method for calculating the category similarity between the attention context information and the category attention context information extracted from a group of documents for each content. A document similar to the retrieval key document from the document database based on the category similarity calculating means and the category similarity calculated by the category similarity calculating means and the category similarity distribution set by the category similarity distribution setting means; Means for selecting a document group for calculating the document similarity belonging to the category including the number of documents, and narrowing down the number of documents according to the document number distribution of the search target document set by the document number distribution setting means. Document similarity calculating means for calculating each document similarity based on the document attention context information of the document group narrowed down by the similarity calculation document selection means and the document attention context information of the search key document; Output means for outputting, as a search result, identification information of a document searched based on each document similarity calculated by the calculation means. Similar document search device. 6. A similar document search method for extracting a similar document from a document database in which a group of documents is stored by content using one document as a search key, wherein a search key document is created, and a category similarity of the search target document is created. Setting a distribution, setting a document number distribution of a search target document, extracting document attention context information suggesting the content from each document in the document database and the search key document, and a document group for each content in the document database And collects the category attention context information, and calculates the category similarity between the document attention context information extracted from the search key document and the category attention context information extracted from the group of documents classified by content. Based on the category similarity and the set category similarity distribution, a category including a document that is similar to the retrieval key document from the document database is included. The document group for calculating the document similarity belonging to the selected document group is selected, and the number of documents is reduced to the number of documents according to the set document number distribution of the search target document. Calculating each document similarity based on the attention context information, and outputting, as a search result, identification information of a document searched based on each document similarity calculated by the document similarity calculation means; retrieval method.