JP2007323454A - Document classification device and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To automatically apply classification to input documents. <P>SOLUTION: This document classification device comprises an input means 1 for inputting a document to be classified, a document similarity calculation means 3 for calculating the similarity of the input document with a document aggregate to which classification is previously applied by extracting a keyword, a document extraction means 2 for extracting a specified number of documents most similar to the input document from the document aggregate to which classification is previously applied, a score calculation means 4 for calculating the score of the classification of the extracted specified number of documents with the same classification number added with the similarity, and a classification aggregate extraction means 5 for extracting the classification where the calculated score is larger than the specified value. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention collects a document similar to a document to be classified by a BM25 or SMART method that is known with high accuracy in a search, and classifies the document into a category having a high appearance frequency in the collected document group, and It is an invention related to a program. In particular, the present invention deals with a multi-class classification problem in which multiple classifications are assigned to a single document, and automatically determines which classification among the classifications with a high appearance frequency is the classification of the document. It has become a framework.

Conventionally, there has been an effective method in research on a multi-class classification problem using machine learning methods such as the support vector machine method and the maximum entropy method (see Non-Patent Document 1). However, there was no effective method for collecting similar documents and using them to handle the multi-class classification problem. In particular, patent classification has a problem that it is difficult to use machine learning methods such as support vector machine method and maximum entropy method because of the large number of documents.
Jun Hirahiro, Masahiko Haruno, Attribute Selection in Text Classification by Support Vector Machine, Journal of Information Processing Society, Vol.41, No.4, 2000, p.1113-1123.

  In the conventional classification method using the machine learning method, a document having a large number of documents and a plurality of classifications assigned to one document cannot be accurately classified.

  The present invention solves the above-mentioned problems, handles a multi-class classification problem in which a plurality of classifications are given to a single document, and classifies the classification of the document as to which classification has a high appearance frequency. The purpose is to automatically decide what to do.

  FIG. 1 is an explanatory diagram of a document classification apparatus according to the present invention. In FIG. 1, 1 is an input unit (input unit), 2 is a document extraction unit (document extraction unit), 3 is a document similarity calculation unit (document similarity calculation unit), 4 is a score calculation unit (score calculation unit), Reference numeral 5 denotes a classification set extraction unit (classification set extraction unit), and reference numeral 6 denotes an output unit (output unit).

  The present invention has the following means in order to solve the conventional problems.

  (1): an input unit 1 for inputting a document to be classified, a document similarity calculation unit 3 for extracting and calculating a similarity between the input document and a document set to which classification is given in advance, Document extraction means 2 for extracting a specified number of documents most similar to the input document from a document set to which classification is given in advance, and the specified number of documents extracted according to the number of the same classification in consideration of the similarity. Score calculating means 4 for calculating a score for classification, and classification set extracting means 5 for extracting a classification for which the calculated score is greater than a specified value. For this reason, classification can be automatically given to the input document.

  (2): an input unit 1 for inputting a document to be classified, a document similarity calculation unit 3 for extracting and calculating a similarity between the input document and a document set to which classification is given in advance, Document extracting means 2 for extracting a specified number of documents most similar to the input document from a set of previously assigned documents, and how many documents the classification of the specified number of extracted documents appears in, Score calculating means 4 for calculating the score of the classification, and classification set extracting means 5 for extracting the average number of classifications attached to the extracted specified number of documents in the order of classification in which the calculated score is large. For this reason, it is possible to automatically assign a classification to an input document simply by setting a designated number of documents to be extracted from a document set to which a classification is assigned in advance.

  (3): In the document classification apparatus according to (1) or (2), the technical viewpoint of the classification of the plurality of extracted documents is converted into a two-dimensional table, and a mark indicating where the technology classified in the table is located. Is provided. For this reason, it is possible to easily find out what is a technical point of view to which no classification is given (technology that has not been developed).

  (4): In the document classification device of (3), the technical viewpoints are rearranged, and the portions not marked are collected again. For this reason, holes without classification can be found more easily.

  (5): In the document classification apparatus according to (1) to (4), the document is a patent document. For this reason, a classification such as F-term can be automatically assigned to a patent document.

  (6): an input unit 1 for inputting a document to be classified, a document similarity calculation unit 3 for calculating a similarity between the input document and a pre-classified document set by extracting a keyword, Document extraction means 2 for extracting a specified number of documents most similar to the input document from a document set to which classification is given in advance, and the specified number of documents extracted according to the number of the same classification in consideration of the similarity. The score calculation means 4 for calculating the classification score and the classification set extraction means 5 for extracting the classification for which the calculated score is greater than the specified value are programs for causing the computer to function. Therefore, by installing this program in a computer, it is possible to easily provide a document classification device that can automatically classify an input document.

  (7): An input unit 1 for inputting a document to be classified, a document similarity calculation unit 3 for calculating a similarity between the input document and a pre-classified document set by extracting a keyword, Document extracting means 2 for extracting a specified number of documents most similar to the input document from a set of previously assigned documents, and how many documents the classification of the specified number of extracted documents appears in, The computer functions as score calculation means 4 for calculating the score of the classification and classification set extraction means 5 for extracting the average number of classifications attached to the specified number of extracted documents in the order of classification in which the calculated score is large. Program. For this reason, by installing this program on a computer, it is possible to automatically assign a classification to an input document simply by setting a designated number of documents to be extracted from a document set to which classification has been assigned in advance. The device can be provided easily.

  The present invention has the following effects.

  (1): The similarity between the document input by the document similarity calculation unit and the document set to which the classification is given is calculated by extracting a keyword, and the document extraction unit calculates the similarity from the document set to which the classification is given in advance. Extract a specified number of documents most similar to the input document, calculate a classification score of the extracted specified number of documents according to the number of the same classification taking the similarity into account, and extract a classification set Since the category with the calculated score larger than the specified value is extracted by the means, the category can be automatically given to the input document.

  (2): The similarity between the document input by the document similarity calculation unit and the document set to which the classification is given is extracted and calculated, and the document extraction unit calculates the similarity from the document set to which the classification is given in advance. A specified number of documents that are most similar to the input document are extracted, and the classification score is calculated according to the number of documents in which the specified number of document classifications extracted by the score calculating means appear, and a classification set extracting means In order to extract the average number of classifications attached to the specified number of documents extracted in the order of classification in descending order of the score calculated in step 1, simply specify the number of documents to be extracted from the document set to which classification has been assigned. Classification can be automatically given to the selected document.

  (3): A technical point of view (development with no classification) is made because the technical point of view of the classification of the extracted documents is made into a two-dimensional table and a mark indicating where the technology classified in the table is located. You can easily find out what the technology is).

  (4): Since the technical viewpoints are rearranged and the portions not marked are collected again, holes without classification can be found more easily.

  (5): Since the document is a patent document, a classification such as F-term can be automatically given to the patent document.

  In the present invention, documents similar to the document to be classified are collected by the BM25 or SMART method known with high accuracy in the search, and the documents are classified into categories having a high appearance frequency in the document group. In particular, a multi-class classification problem in which multiple classifications are assigned to a single document is handled, and among the classifications with a high frequency of occurrence, up to which classification is used as the classification of the document It has become.

(1): Description of Document Classification Device FIG. 1 is an explanatory diagram of a document classification device. 1, the document classification apparatus includes an input unit (input unit) 1, a document extraction unit (document extraction unit) 2, a document similarity calculation unit (document similarity calculation unit) 3, and a score calculation unit (score calculation unit). 4. A classification set extraction unit (classification set extraction unit) 5 and an output unit (output unit) 6 are provided.

  The input unit 1 is input means for inputting a document such as a patent document. The document extracting unit 2 is a document extracting unit that extracts a document (k pieces) similar to a document to be classified. The document similarity calculation unit 3 is a document similarity calculation unit that calculates the similarity between documents. The score calculation unit 4 is score calculation means for calculating a classification score. The classification set extraction unit 5 is a classification set extraction unit that extracts a classification set of documents to be classified (with a score equal to or higher than a specified value) based on the classification score. The output unit 6 is output means for outputting the classification of the document to be classified.

(2): Description of Patent Document Classification Device Patent documents (patent documents) are classified by IPC, FI, F-term, and the like. In particular, the F-term distinguishes a certain technical scope (theme) from various technical viewpoints from many viewpoints. For example, there are many purposes, applications, structures, materials, manufacturing methods, processing methods, control means, etc. It is based on a term list that distinguishes technologies from the technical point of view. For this reason, a plurality of F-terms (patent classifications) are usually assigned to one patent document. Hereinafter, a description will be given of a case where a patent document is used as a document.

FIG. 2 is an explanatory diagram of the patent document classification apparatus. In FIG. 2, the patent document classification apparatus includes an input unit (input unit) 1, a KDOC extraction unit (KDOC extraction unit) 2, a document similarity calculation unit (document similarity calculation unit) 3, and a score (Score _M1 (x) ) A calculation unit (score calculation unit) 4, an F-term x set extraction unit (F-term x set extraction unit) 5, and an output unit (output unit) 6 are provided.

The input unit 1 is input means for inputting a patent document. The KDOC extraction unit 2 is a KDOC extraction unit that extracts patent documents (k) similar to the patent documents to be classified. Here, KDOC is k extracted patent documents. The document similarity calculation unit 3 is a document similarity calculation unit that calculates the similarity between patent documents. The score (Score _M1 (x)) calculating unit 4 is score calculating means for calculating a patent classification score (Score _M1 (x)). The F-term x set extraction unit 5 is classification set extraction means for extracting a set of F-term x of patent documents to be classified based on patent classification scores. The output unit 6 is an output means for outputting a set of F-term x of patent documents to be classified.

(3): Explanation of Patent Document Classification Processing FIG. 3 is a flowchart of patent document classification processing. Hereinafter, description will be given according to the processing S1 to S5 of FIG.

  S1: A patent document to be classified is input to the input unit 1.

  S2: The KDOC extraction unit 2 extracts k patent documents (KDOC) similar to the inputted patent document to be classified. Here, the document similarity calculation unit 3 obtains the similarity between the inputted patent document to be classified and the patent document set (in a storage means such as a database) given as learning data. The patent document set given as learning data is a document set to which a correct F-term classification is assigned. The ruby-ir toolkit was used to extract k patent documents. k is a value determined by experiments.

S3: The score (Score _M1 (x)) calculating unit 4 calculates a patent classification score (Score _M1 (x)).

  S4: The F-term x set extraction unit 5 extracts a set of F-term x (scores greater than or equal to a specified value) of patent documents to be classified based on patent classification scores.

  S5: The output unit 6 outputs a set of F-term x of patent documents to be classified.

  FIG. 4 is a process flowchart for obtaining the similarity between the input patent document and the selected patent document. Hereinafter, a description will be given according to the processes S11 to S12 of FIG.

  S11: The document similarity calculation unit 3 extracts keywords from the input patent document. As this keyword, nouns were extracted using morphological analysis technology.

S12: The document similarity calculation unit 3 next selects a patent document including at least one of the above keywords from all the patent documents having a given input theme (theme may not be given) in the learning data. Take out and calculate the Sim _SMART of the taken out patent document. This Sim _SMART is used as the similarity between each patent document in the learning data.

(4): Explanation of F-term x set extraction There are four methods for extracting the F-term x set as follows.

a) Description of Method 1 First, the patent classification device (KDOC extraction unit 2) first sets k patent documents most similar to the input as a set of patent documents (correct F-term classification assigned as learning data). From the document set). These k patent documents will be referred to as KDOC. The ruby-ir toolkit was used to retrieve the documents. k is a value determined by experiments.

(Reference for ruby-ir toolkit)
ruby-ir-eng, "Masao Utiyama", "Information Retrieval Module for Ruby", 2005,
("Www2.nict.go.jp/jt/a132/members/mutiyama/software")
Next, the patent classification device (score calculation unit 4) calculates the score (Score _M1 (x)) of F-term x by sorting KDOCs according to the following equation (1).

here,
role (x, i) = 1 (if i-th document has F-term x classification)
= 0 (otherwise)
Here, score _doc (i) is the value of the similarity of the document in which the similarity between the input document and the selected document is the i-th largest, and k _r is a constant determined by experiment. You can also simplify score _doc (i) as follows:

score _doc (i) = 1001-i
The patent classification device (classification set extraction unit 5) finally extracts a set of F-term x that satisfies the following expression (2).

{X | Score _M1 (x) ≧ k _p × max _y Score _M1 (y)} (2)
However, k _p is a constant determined by experiment. This is the classification required by the set of F-term x extracted.

Explanation of usage example of Method 1 (F-term1, F-term2, etc. below are F-terms used in each document)
Document A Similarity with input document 100 F-term1
Document B Similarity with input document 90 F-term1 F-term2
Document C Similarity with input document 80 F-term1
Document D Similarity with input document 70 F-term3
And kr = 0.99,
The score of F-term1 is 100 + 90 * 0.99 + 80 * 0.99 ^ 2 = 267.5
F-term2 score is 90 * 0.99 = 89.1
F-term3 score is 70 * 0.99 ^ 3 = 67.9
It becomes.

  If kp = 0.9, the classification of scores with a score of 240.8 or higher, which is 0.9 times the top score of 267.5, is extracted. In this case, only F-term1 satisfies it, so only F-term1 is taken out as an answer.

b) Description of Method 2 First, the document classification device extracts KDOC as in Method 1. The document classifier then counts how many documents F-term x appears in KDOC. When this number is written as F _KDOC (x), the document classification device finally takes out a set of F-term x satisfying the following expression (3).

{X | F _KDOC (x) ≧ k _u × k},
However, k _u is a constant determined by experiment. However, when k _u = 0.5, this method is identical to the original k-neighbor method.

c) Description of Method 3 First, the document classification device extracts KDOC as in Method 1. The document classifier then calculates F _KDOC (x). The document classification device finally takes out k _f F-terms in descending order of the value of F _KDOC (x), and determines them as the classification to be obtained. Here, k _f is a constant determined by experiment.

d) Description of Method 4 First, the document classification device extracts KDOC as in Method 1. The document classifier then calculates F _KDOC (x). Document classification apparatus, finally, removed k _a number of F-term in the order of decreasing F _KDOC (x), and classified to obtain this. However, k _a is the average of the number of F-term categories used for each document in KDOC.

  In order to confirm the effectiveness of each of the above methods, the following baseline method was used in the experiment.

(1) Baseline 1
The document classification device first extracts all F-term classifications from all patent documents having a given theme classification in the learning data.

The document classification device takes k _b F-terms at random and sets them as the classifications to be obtained.
Where k _b is the average of the number of F-term classifications applied to patent documents with a given theme classification.

(2) Baseline 2
First, the document classification device extracts all F-term classifications from all patent documents having a given theme classification in the learning data, and arranges them in descending order of the number of documents in which the classification appears. Document classification apparatus, in descending order of the number of documents classified appeared removed k _b number of F-term and classified for obtaining it. Where k _b is the average of the number of F-term classifications applied to patent documents with a given theme classification.

(3) Original k-nearest neighbor method (cited reference)
Fukunaga, 1972; Okamoto and Yugami, 1997; Yang and Liu, 1999; Duda et al., 2001; Guo et al., 2004
The original k-nearest neighbor method is used for each F-term classification. First, the document classification device retrieves KDOC as in Method 1. Document classification apparatus for each F-term classification, the number of articles having the classification in KDOC and (NUM _+), number of articles with no classification _- Request (NUM). The document classification device takes out an F-term classification in which the value of NUM ₊ is NUM _- or more, and uses it as a classification to obtain. This method is equivalent to the following explanation.

First, the document classification device retrieves KDOC as in Method 1. The document classifier then calculates F _KDOC (x). The document classification device finally takes out a set of F-term x that satisfies the following expression.
{X | F _KDOC (x) ≧ 0.5 × k}

(5): Description of Calculation of Similarity Between Documents The following four methods can be used to calculate the similarity between each patent document in the learning data and the input patent document.

a) Description of SMART First, the document classification device extracts a keyword from an input patent document. As keywords, nouns were extracted using morphological analysis technology. Next, a document including at least one keyword is extracted from all patent documents having a given input theme in the learning data. The document classification device (document similarity calculation unit 3) uses the following equation (3) to calculate Sim _SMART of each extracted document. Sim _SMART is used as the similarity between the input document and each patent document in the learning data.

  In this expression, T means the set of keywords that appear in both the input patent document and the retrieved patent document, tf means the number of times the keyword t appears in the retrieved document, and avtf is extracted Qtf means the number of occurrences of keyword t in the input document, utf means the number of different keywords in the extracted document, and pivot means learning Means the average of the number of different keywords per document in all documents in the data, N means the total number of patent documents with a given input theme classification in the training data, and n is the number of documents in which the keyword t appears Means number.

  SMART is one of the keyword weighting methods for information retrieval (cited reference; Singhal et al., 1996; Singhal, 1997).

b) Description of BM25 The document classification device first extracts keywords from the input patent document. As keywords, nouns were extracted using morphological analysis technology. Next, a document including at least one of the keywords is extracted from all patent documents having a given input theme classification in the learning data. The document classification device (document similarity calculation unit 3) uses the following equation (6) to calculate Sim _BM25 of each extracted document. Sim _BM25 is used as the similarity between the input document and each patent document in the learning data.

T, tf, qtf, N, n are the same as those in SMART. dl is the length of the retrieved article, avdl is the average of the length of the article in all documents, k ₁ , k _{3 and} b are constants determined by experiments. As default values of ruby-ir toolkit, values of k ₁ = 1, k ₃ = 1000, b = 1 were used. Instead of log {(N-n + 0.5) / (n + 0.5)} in the original formula of BM25, log (N / n) was used. This is because the original formula outputs a negative score. It was confirmed that the formula corrected in the experiment gave higher accuracy.

  BM25 is one of the keyword weighting methods for information retrieval (cited reference; Robertson et al., 1994).

c) Description of Tfidf First, the document classification device extracts keywords from the input patent document. As keywords, nouns were extracted using morphological analysis technology. Next, a document including at least one of the keywords is extracted from all documents having a given input theme classification in the learning data. The document classification device (document similarity calculation unit 3) uses the following equation (9) to calculate Sim _Tfidf of each extracted document. Sim _Tfidf is used as the similarity between the input document and each document in the training data.

この式で、T 、tf、N 、 nは、 SMARTのものと同一である。

In this equation, T 1, tf, N and n are the same as those of SMART.

d) Description of Overlap First, the document classification device extracts keywords from the input patent document. As keywords, nouns were extracted using morphological analysis technology. Next, a document including at least one of the keywords is extracted from all documents having a given input theme classification in the learning data. The document classification device (document similarity calculation unit 3) uses the following equation (10) to calculate the Sim _Overlap of each extracted document. Sim _Overlap is used as the similarity between the input document and each document in the training data.

この式で、T は、 SMARTのものと同一である。

In this equation, T is the same as that of SMART.

(6): Explanation of Experimental Results FIG. 5 is an explanatory diagram of experimental results. In FIG. 5, keywords are extracted from the summary part and the claim part of the patent document. Dry run data was used to determine the parameters for each method. Experimental results with Formal run data show the performance of the method. In the table of FIG. 5, * is given to the method with the highest performance, and-means that it is inferior to the method of * with a significant difference of 0.01. This significant difference test utilizes a two-sided t-test. The experimental results show that the method using SMART and Method 1 is the best.

(7): Description of Document Classification Contest FIG. 6 is an explanatory diagram of formal run at NTCIR-5 Patent Workshop. In Fig. 6, NTCIR-5 Patent Workshop is a document classification contest. Three teams, including our team, participated. Our system (document classification device) has an overwhelming accuracy difference from other teams, indicating the superiority of the system. Our system 1 is a method using BM25 and method 1, and the detailed implementation is different from the comparative experiment of the above method.

  In the above contest, when the theme classification of a patent document is given, the F-term classification of the input Japanese patent document is requested, and F-measure is used for evaluation. F-measure is the reciprocal of the average of the reciprocal of recall and the reciprocal of precision. The recall ratio is the ratio of correct outputs among correct answers, and the relevance ratio is the ratio of correct outputs among all outputs. Expressed as an expression:

(8): Explanation of Discovery of New Patent Possibility As described above, the present invention is an invention related to document classification. Documents similar to the document to be classified are collected by the BM25 and SMART methods that are known with high accuracy in the search, and the documents are classified into categories having a high appearance frequency in the document group. In particular, a multi-class classification problem in which multiple classifications are assigned to a single document is handled, and among the classifications with a high frequency of occurrence, up to which classification is used as the classification of the document It has become.

  The patent document has a code for classifying a plurality of patents. The code is generally given manually, but if the present invention is used, the code can be given even to some extent automatically, and there is an effect of reducing the manual work. Further, if patent data can be automatically classified, the following effects can be obtained. A patent document has a classification code for classifying patents from various viewpoints called F-terms. By using these codes, it is possible to know which problem is handled by each patent and in what way. If this information is organized for each patent and the data in the table of Fig. 7 is created (automatically), there are patents that deal with which method and which method, and patents that deal with which method and which method. You can see if there is.

  FIG. 7 is an explanatory diagram of the discovery of the possibility of a new patent. For example, in the table of discovery of the possibility of a new patent in FIG. It is. Circles indicate those with an F-term. For example, the circle at the upper left is the one with the F-term (Method 1, Problem 1) of Patent 1.

  Where an ellipse is shown, there is no patent, and it is possible to consider a new patent that deals with problems 3 to 6 in methods 3 to 5, and may find a new patent. The present invention can also be used in systems that assist in discovering such patent possibilities. In FIG. 7, those with an F term are simply indicated by circles, but frequency information may be added. For example, different displays such as a circle, a double circle, and a triple circle can be made depending on the number of patents with the same F term.

(Explanation of rearrangement of technical points of view)
After issuing the table of FIG. 7, it is possible to rearrange the problems 1, 2, 3,... 1, 2, 3,.

・ Method A
The dimension of the problem vector is the number of methods, the dimension of the method vector is the number of problems, and each vector element must be “1” if there is a method in question or a patent for the problem of the method. For example, each problem vector and each method vector are arranged based on the proximity of the vectors.

  The close problems are arranged as close as possible to each other, the distant problems are arranged as far as possible, the close methods are arranged as close as possible, and the distant methods are arranged as far as possible from each other.

・ Method B
Method A is a more detailed method. First, the closest problems are brought together. Attach the problem closest to one of the problems at the end of the attached list. In addition, attach the problem closest to one of the problems at the end of the attached list.

  Repeat this. Sort the list in order. In the same way, the list is created and rearranged.

・ Method C
It uses dual scaling.

(Reference)
Taichiro Ueda, Masao Karita, Kazue Honda ", Practical Workshop Excel Extensive Multivariate Analysis", "Hidekazu System", 2003,
Dual scaling has the ability to place both problems and methods similar to one two-dimensional figure by eigenvalue computation. The input to the dual scaling method is a two-dimensional table of problems and methods, where each problem 1, 2, 3,... And method 1, 2, 3,. If there is a patent handled in, enter “0” if it is not “1”. If such a table is input, dual scaling can arrange both problems and methods similar to one two-dimensional figure by eigenvalue calculation.

  This two-dimensional diagram uses two axes, an axis based on the first eigenvalue and an axis based on the second eigenvalue, but here only uses an axis based on the first eigenvalue. In the table of the original problem and method, using each value on the axis based on the first eigenvalue in each of the problems 1, 2, 3,. .., And methods 1, 2, 3,... Are rearranged in the order of their values. Then, it can be rearranged so as to have more “1” s near the diagonal of the table.

  In the description of the methods A, B, and C, the value of an element in a vector and the value of each element in a table of inputs to a dual scaling method are “1” unless there is a patent that deals with the problem in that method. On the contrary, if there is a patent dealing with the problem by that method, it may be “1” if it is not “0”.

  For example, in the method C, if there is a patent that deals with the problem by the method, if the table is rearranged as “1” if it is not “0”, the result is as shown in FIG.

  FIG. 8 is an explanatory diagram of rearrangement from the technical viewpoint. In FIG. 8, it can be seen that blank spaces (indicating that there is no prior patent) are concentrated near the diagonal line of the table. This method is useful for searching for new patent possibilities where there is no large prior patent. Here, the above methods A, B, and C are shown, but a method of rearranging by another method may be used. For example, create multiple tables to be rearranged at random, and blanks appear continuously to form a large rectangle in the table, and among the tables created by judging the size of the rectangle, the size of the rectangle is the most. A method of rearranging it may be used, assuming that a larger one is better.

  In fact, by using the method of the present invention, it is the best of the three participating organizations in the F-term classification subtask of the evaluation type workshop NTCIR5 patent document classification task held at the National Institute of Informatics in 2005. The accuracy was achieved.

(9): Explanation of program installation Input unit (input unit) 1, document extraction unit (document extraction unit) 2, KDOC extraction unit (KDOC extraction unit) 2, document similarity calculation unit (document similarity calculation unit) 3, Score calculation unit (score calculation means) 4, score (Score _M1 (x)) calculation unit 4, classification set extraction unit (classification set extraction means) 5, F-term x set extraction unit (F-term x set extraction) The means 5, the output unit (output unit) 6, and the like can be configured by programs, and are executed by the main control unit (CPU) and stored in the main memory. This program is processed by a general computer (information processing apparatus). This computer is composed of hardware such as an input device as input means such as a main control unit, main memory, file device, display device, and keyboard.

  The program of the present invention is installed on this computer. In this installation, these programs are stored in a portable recording (storage) medium such as a floppy disk or a magneto-optical disk, and a drive device for accessing the recording medium provided in the computer is used. Alternatively, it is installed in a file device provided in the computer via a network such as a LAN. Then, the program steps necessary for processing are read from the file device into the main memory and executed by the main control unit.

It is explanatory drawing of the document classification device of this invention. It is explanatory drawing of the patent document classification | category apparatus of this invention. It is a classification process flowchart of the patent document of this invention. It is a processing flowchart which calculates | requires the similarity between the input patent document of this invention, and the selected patent document. It is explanatory drawing of the experimental result of this invention. It is explanatory drawing of Formal run in NTCIR-5 Patent Workshop of this invention. It is an explanatory diagram of the discovery of the possibility of a new patent of the present invention. It is explanatory drawing of the rearrangement of the technical viewpoint of this invention.

Explanation of symbols

1 Input section (input means)
2 Document extraction unit (document extraction means)
3 Document similarity calculation unit (document similarity calculation means)
4 score calculation part (score calculation means)
5 Classification set extraction unit (classification set extraction means)
6 Output section (output means)

Claims (7)

An input means for inputting a document to be classified;
A document similarity calculation means for extracting a keyword and calculating a similarity between the input document and a document set to which classification is given in advance;
A document extracting means for extracting a specified number of documents most similar to the input document from the pre-classified document set;
Score calculating means for calculating a score of the classification of the specified number of extracted documents according to the number of the same classification in consideration of the similarity;
A document classification apparatus comprising: a classification set extraction unit that extracts a classification in which the calculated score is greater than a specified value. An input means for inputting a document to be classified;
A document similarity calculation means for extracting a keyword and calculating a similarity between the input document and a document set to which classification is given in advance;
A document extracting means for extracting a specified number of documents most similar to the input document from the pre-classified document set;
Score calculating means for calculating a score of the classification according to how many classifications of the extracted designated number of documents appear;
A document classification apparatus comprising: a classification set extraction unit configured to extract an average number of classifications attached to the specified number of extracted documents in order of classification in which the calculated score is large.   3. The document classification apparatus according to claim 1, wherein a technical point of view of classification of the plurality of extracted documents is a two-dimensional table, and a mark indicating where the technology classified in the table is located is provided.   4. The document classification apparatus according to claim 3, wherein the technical points of view are rearranged, and the portions not marked are collected again.   The document classification apparatus according to claim 1, wherein the document is a patent document. An input means for inputting a document to be classified;
A document similarity calculation means for extracting a keyword and calculating a similarity between the input document and a document set to which classification is given in advance;
A document extracting means for extracting a specified number of documents most similar to the input document from the pre-classified document set;
Score calculating means for calculating a score of the classification of the specified number of extracted documents according to the number of the same classification in consideration of the similarity;
A program for causing a computer to function as classification set extraction means for extracting a classification in which the calculated score is greater than a specified value. An input means for inputting a document to be classified;
A document similarity calculation means for extracting a keyword and calculating a similarity between the input document and a document set to which classification is given in advance;
A document extracting means for extracting a specified number of documents most similar to the input document from the pre-classified document set;
Score calculating means for calculating a score of the classification according to how many classifications of the extracted designated number of documents appear;
A program for causing a computer to function as a classification set extraction unit that extracts an average number of classifications attached to the specified number of extracted documents in order of classification in which the calculated score is large.

Images