JPH11338867A - Document summarizing method and device and storage medium storing document summarizing program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To summarize even a document that includes the new information without preparing any heuristics by extracting a part where an object document is best featured from the object document based on the feature information and outputting the extracted part as a summary that features the object document. SOLUTION: In a document summarizing method which extracts the information featuring an object document that is read out of an input storage means storing the document data and then summarizes the document, an object document to be summarized are compared with a standard document set, i.e., a total document set that can be processed and the feature information featuring the object document is calculated out of the object document (S1). A part that features best the object document is extracted from the object document based on the feature information (S2). Then the extracted part is outputted as a summary that features the object document (S3). Thus, it's possible to properly summarize the object document even when the contents of a document assumed at the time of its usage start are different from those of the object document or when the object document includes the new information.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document summarizing method and apparatus, and a storage medium storing a document summarizing program. In particular, the present invention relates to a method for extracting document-characterizing portions used in document information processing from a document. The present invention relates to a document summarizing method and apparatus for summarizing a document, and a storage medium storing a document summarizing program.

[0002]

2. Description of the Related Art In recent years, the Internet has rapidly spread. Further, the data recording device has been increased in capacity and reduced in price. Along with this, a large amount of various information has been easily available via a network. Also, WW
With the spread of W, many users mutually generate and use information. However, as the amount of information that can be used has been dramatically increased, which is referred to as an information flood, it has become difficult to find useful information from among such information and to select it.

[0003] It is difficult to browse all such a large amount of information, search for useful information, and select it. Therefore,
In order to efficiently use appropriate information, it is necessary to extract characteristic information from a large amount of information and make necessary and sufficient information selectively available. Currently, an information search technique is used as a means for selectively using information. However, the amount of information that can be used via a network is enormous, and its contents are also diverse. For this reason, a document including many similar information is selected as a result of one search. In many cases, a summary characterizing each document is attached to each search result document and presented to a user to assist in selecting a document containing appropriate information from the search results.

[0004] It is difficult for a large number of documents to be given a proper and appropriate summary to each document by hand. Also, when work is performed manually, the summary information given is affected by the subjectivity and knowledge possessed by the workers. Can not be kept. Therefore, in order to selectively use information, a document summarization technique for automatically creating and adding an appropriate summarization to document contents is required.

[0005] As a typical document summarizing technique, there is a summarizing technique using heuristics. This heuristic-based summarization technique determines the importance of a sentence by combining position information in the document, words appearing in titles and headings, and the presence or absence of clue words,
This is a technique for extracting sentences determined to be important and combining the important sentences to create a summary.

This technique requires that heuristics be prepared in advance. Therefore, it is not suitable for summarizing a document in which new information appears. Also, heuristics that are valid for one document type are not necessarily valid for another document type. For example, position information is effective for newspaper articles and the like. In academic papers, heuristics using clue words are effective.

[0007]

However, various types of documents are mixed on the Internet.
It is necessary to determine the document type automatically. However, at present, there is a problem that there is no effective technique for determining the document type. In addition, since there are many documents in which titles and headings do not exist and in which there are few clues and phrases, there is a problem that heuristics often do not work effectively.

The present invention has been made in view of the above points, and stores a document summarizing method and apparatus and a document summarizing program which can summarize a document in which new information appears without preparing heuristics or the like in advance. It is an object of the present invention to provide a storage medium that has been used.

[0009]

FIG. 1 is a diagram for explaining the principle of the present invention. According to a first aspect of the present invention, in a document summarizing method for extracting and characterizing information that characterizes a target document read from an input storage unit that stores document data, the target document to be summarized is converted into an entire document set that can be processed. Is compared with the standard document set, and the characteristic information characterizing the target document is calculated from the target document (step 1). Based on the characteristic information, the part that characterizes the target document is extracted from the target document (step 2). ), And outputs the extracted portion as a summary characterizing the target document (step 3).

FIG. 2 is a diagram showing the principle of the present invention. The present invention (claim 2) is a document summarizing apparatus that extracts information characterizing a target document read from an input storage unit 20 storing document data and summarizes the document, and the target document to be summarized can be processed. A feature information calculating unit 40 for comparing the standard document set which is the entire document set and calculating feature information for characterizing the target document from the target document; and, based on the feature information, a part which most characterizes the target document from the target document. It has a characteristic part extracting means 50 to be extracted, and a summary creating means 60 for outputting the part extracted by the characteristic part extracting means 50 as a summary characterizing the target document.

According to the present invention (claim 3), a standard document set updating means for acquiring a standard document set from an input storage means, and each document in the standard document set given to the standard document set updating means are analyzed. Standard document set analysis means for calculating words constituting a document and the appearance frequency of the word; standard document set analysis result storage means for storing words in the standard document set and the appearance frequency of the word in association with each other; A target document inputting means for receiving a target document from the means, a target document analyzing means for analyzing the target document and calculating words constituting the target document and an appearance frequency of the word, a word in the target document and an appearance of each word A target document analysis result storage unit that stores frequencies in association with each other, and characteristic information in the target document is stored in a standard document analysis result stored in the standard document set analysis result storage unit and a target document analysis result storage unit. A target document feature calculation unit that calculates the target document using the analysis result of the target document, a target document feature storage unit that stores feature information of the target document calculated by the target document feature calculation unit, and a feature information storage unit that stores the feature information. Feature expression extraction means for extracting a feature expression from a target document by using feature information of the target document, feature expression storage means for storing a feature expression extracted from the target document by the feature expression extraction means, and feature expression storage Means for outputting a feature expression stored in the means to the transfer medium.

According to the present invention (claim 4), the characteristic information calculating means includes means for using the chi-square test for calculating the characteristic information. According to the present invention (claim 5), in the target document feature calculation means, a word, a character, or a character string having a certain length is used as a calculation unit of the feature information score obtained by referring to the standard document set analysis result and the target document analysis result. Used.

According to a sixth aspect of the present invention, there is provided a storage medium storing a document summarizing program for extracting and characterizing information characterizing a target document read from an input storage unit storing document data, the program comprising: Compare the document with a standard document set, which is the entire set of documents that can be processed,
A feature information calculation process of calculating feature information characterizing the target document from the target document, a feature portion extraction process of extracting the most characteristic portion of the target document from the target document based on the feature information, and a feature portion extraction process. A summary creation process for outputting the extracted portion as a summary characterizing the target document.

According to the present invention (claim 7), a standard document set updating process for acquiring a standard document set from an input storage means;
Each document in the standard document set given to the standard document set update process is analyzed, the words constituting the document and the frequency of occurrence of the words are calculated, and the standard document set analysis result storage means stores A standard document set analysis process for storing the words in association with the frequency of occurrence of the word, a target document input process for receiving the target document from the input storage means, and analyzing the target document to determine the words constituting the target document and the A target document analysis process of calculating the frequency of occurrence of a word and storing the word in the target document and the frequency of occurrence of each word in the target document analysis result storage means in association with each other;
The target document is calculated using the standard document analysis result stored in the standard document set analysis result storage unit and the target document analysis result stored in the target document analysis result storage unit, and the calculated target is stored in the target document feature storage unit. A feature expression is extracted from the target document by using the target document feature calculation process for storing the feature information of the document and the feature information of the target document stored in the feature information storage unit, and the feature expression is stored in the feature expression storage unit. It has a feature expression extraction process for storing, and a feature expression output process for providing the feature expression stored in the feature expression storage means to the transfer medium.

As described above, according to the present invention, a target document is compared with a standard document set, information characterizing the target document is calculated from the target document, and a part that characterizes the target document is determined based on the characteristic information. By extracting from a document and outputting the extracted portion as a summary that characterizes the document, it is possible to provide a large amount of documents with a timely and appropriate amount of summary characterizing each document. Further, it is not necessary to prepare knowledge to be used for document summarization in advance, and the present invention can be applied to summarization of various documents.

[0016]

FIG. 3 shows the configuration of a document summarizing apparatus according to the present invention. The document summarizing apparatus shown in FIG.
0, the input storage device 20, the transfer medium 30, the standard document set update unit 101, the standard document set analysis unit 102, the target document input unit 103, the target document analysis unit 104, the target document feature calculation unit 105, and the feature expression extraction unit 106 , Feature expression output unit 10
7, a standard document set analysis result storage unit 201, a target document analysis result storage unit 202, a target document feature storage unit 203, and a feature expression storage unit 204 are connected.

Here, each of the processing units 101 to 107 is composed of, for example, a digital computer,
ROM for storing an operation program and data for executing the operation program, and R used as a working memory
AM. Note that all the processing units may be constituted by one digital computer. In addition, each of the storage units 201 and 2
04 is stored in a memory such as a hard disk memory, for example.

The input storage unit 20 stores standard document sets and target documents provided to the apparatus in a certain order. The input storage unit 20 includes a semiconductor memory device or
It can be realized by a hard disk or a floppy disk. The transfer medium 30 is a communication channel or a recording medium to which the processing result of the present apparatus is given. In the following description, the “standard document set” refers to an entire document set that can be processed by the present apparatus or a standard document set having the document set as a parent set. The “target document” indicates a document to be summarized. The “feature expression” is an expression extracted from the “target document”, and the “target document”
Is an expression that characterizes

The standard document set updating unit 101 receives a standard document set from the input storage device 20. The standard document set analyzing unit 102 analyzes each document in the standard document set given to the standard document set updating unit 101, and calculates the words constituting the document and the appearance frequency of the words.

The standard document set analysis result storage unit 201 stores the words in the standard document obtained by the standard document set analysis unit 102 in association with the appearance frequencies of the words. The target document input unit 103 receives the target document to be summarized from the input storage device 20 and passes it to the target document analysis unit 104. The target document analysis unit 104 analyzes the target document passed from the target document input unit 103, and calculates words constituting the document and the appearance frequency of the words.

The target document analysis result storage unit 202 stores the frequency of occurrence of a word in the target document obtained by the target document analysis unit 104 in association with the document. The target document feature calculation unit 105 compares the feature information in the target document with the standard document analysis result stored in the standard document set analysis result storage unit 201 and the target document analysis result storage unit 202
Is calculated using the target document analysis result stored in the.

The target document feature storage unit 203 records the feature information of the target document calculated by the target document feature calculation unit 105. The characteristic expression extraction unit 106 extracts a characteristic expression from the target document using the characteristic information of the target document stored in the target document characteristic storage unit 203. The characteristic expression storage unit 204 stores the characteristic expressions extracted from the target document by the characteristic expression extraction unit 106.

The feature expression output unit 107 gives the feature expression stored in the feature expression storage unit 204 to the transfer medium 30.

[0024]

Embodiments of the present invention will be described below with reference to the drawings. First, the standard document set analysis result storage unit 201, target document analysis result storage unit 202, target document feature storage unit 203, and feature expression storage unit 204 connected to the monitoring control unit 10 shown in FIG. 3 will be described.

The standard document set analysis result storage unit 201 stores and holds a standard document set analysis result, which is a processing result of the standard document set analysis unit 102. Here, the “standard document set” is the entire document set to be searched by the information search device to which the present apparatus is connected, or the standard set of this document set. The “analysis result of the standard document set” is obtained by morphologically analyzing the sentences described in all the documents in the standard document set, and correlating the expression of each word with the frequency of appearance of each word. The “analysis result” is expressed, stored, and held as a table including two columns of a word expression and an appearance frequency.
In this table, each row represents the correspondence between each word expression and the appearance frequency of that word. This table has a structure in which a corresponding row can be searched using each word expression as a key.

The target document analysis result storage unit 202 stores and holds the analysis result of the target document obtained as the processing result of the target document analysis unit 104. Here, "target document"
This is the document to be summarized. The "analysis result of the target document" is obtained by morphologically analyzing a sentence in the target document and associating each word expression with each word appearance frequency in the target document. The “analysis result” has the same format as the analysis result of the standard document set, and is expressed, stored, and held as a table including two columns of a word expression and an appearance frequency. In this table, each row represents the correspondence between each word expression and the appearance frequency of that word. This table has a structure in which a corresponding row can be searched using each word expression as a key.

The target document feature storage unit 204 stores and retains information obtained as a result of processing by the target document feature calculation unit 105 in which the features of the target document with respect to the standard document set are scored. Here, the "characteristics of the target document" means that the frequency distribution of words in the standard document set and the frequency distribution of words in the target document are compared, and the magnitude of the difference in the distribution is quantified for each word. The larger the difference between the appearance frequency distribution in the standard document set and the appearance frequency distribution in the target document, the larger the numerical value. A characteristic word in the target document, that is, a word having a larger difference between the appearance frequency distribution in the standard document set and the appearance distribution in the target document has a larger numerical value. The “feature of the target document” is expressed as a table including two columns of each word expression and a feature score obtained by quantifying the feature of the appearance distribution of the word. Each line represents the correspondence between the expression of each word and the feature score. One table corresponds to one target document.

The feature expression storage unit 204 stores and holds the feature expressions in the target document obtained as a result of processing by the feature expression extraction unit 106 in association with the document from which the information is extracted. The characteristic expression in the target document is a word string that is included in the target document and has a predetermined number of words, or
A predetermined number of sentences, or a word string constituting a predetermined part, and the average of the feature scores of all the words constituting the continuous word string, the sentence, and the word string constituting the part is the largest. Part.

Next, each processing unit will be described. The monitoring control unit 10 is a module that controls the processing units 101 to 107 and controls a data flow. FIG. 4 is a flowchart of the monitoring control process executed by the monitoring control unit according to one embodiment of the present invention.

The processing will be described below with reference to FIG. Step 101) It is determined whether or not the standard document set has been updated. If updated, step 102
Then, if it is not updated, the process of step 101 is repeated. Step 102) The updated standard document set is transferred from the input storage device 20 to the standard document set update unit 101. At this point, the standard document set update unit 101 executes a standard document set update process on the transferred standard document set,
The monitoring control unit 10 receives the standard document set update result as the processing result.
Output to

Step 103) Standard document set updating unit 1
The standard document set analysis unit 102 transfers all the standard document set update results output from the standard document set 01. At this time, the standard document set analysis unit 102 executes the standard document set analysis processing, and outputs the processing result to the monitoring control unit 10. Step 104) Transfer all the standard document set analysis results output from the standard document set analysis unit 102 to the standard document set analysis result storage unit 201, update the contents, and store and hold the newly transferred values. .

Step 105) It is determined whether or not the target document has been input. If input, step 10
The process proceeds to step 6 and if not entered, step 10
Step 5 is repeated. Step 106) The input target document is transferred from the input storage device 20 to the target document input unit 103. The target document input unit 103 executes a target document input process on the input target document, and outputs a processing result to the monitoring control unit 10.

Step 107) Target document input unit 103
Is transferred to the target document analysis unit 104. At this time, the target document analysis unit 104
Executes the target document analysis process and outputs the analysis result to the monitoring control unit 10. Step 108) The target document analysis result output from the target document analysis unit 104 is transferred to the target document analysis result storage unit 202, the content of the storage unit is updated, and the newly transferred value is stored and held.

Step 109) The standard document set analysis result stored in the standard document set analysis result storage unit 201 is transferred to the target document feature calculation unit 105, and the target document stored in the target document analysis result storage unit 202 is stored. The analysis result is transferred to the target document feature calculation unit 105. The target document feature calculation unit 105 executes a target document feature calculation process based on the transferred standard document set analysis result and the target document analysis result. The processing result is output to the monitoring control unit 10.

Step 110) Target document feature calculation unit 1
The target document feature output from step 05 is transferred to the target document feature storage unit 203, the content is updated, and the newly transferred value is stored and held. Step 111) Transfer the target document features stored in the target document feature storage unit 203 to the feature expression extraction unit 106. At this time, the feature expression extraction unit 106 performs a feature expression extraction process based on the target document features. The feature expression as the processing result is output to the monitoring control unit 10.

Step 112) Feature expression extraction unit 106
Is transferred to the feature expression storage unit 204, its contents are updated, and the newly transferred values are stored and held. Step 113) The feature expression stored in the feature expression storage unit 204 is transferred to the feature expression output unit 107. At this time, the feature expression output unit 107 performs a feature expression output process on the transferred feature expression. The processing result is
It is output to the monitoring control unit 10.

Step 114) Special information output unit 107
Output to the transfer medium 30. Step 115) It is determined whether or not all the processes have been completed. If all the processes have been completed, the monitoring control process is completed. If the processing has not been completed, the process proceeds to step 101, and the above processing is repeated.

The standard document set updating unit 101 executes a standard document set updating process on the standard document set transferred from the monitoring control unit 10. This processing is a pre-processing of the subsequent processing, and removes a part unnecessary for processing by the present apparatus from each document in the input standard document set. Also, it is converted into a corresponding character code in the subsequent processing. The processing result is output to the monitoring control unit 10.

The standard document set analysis unit 102 executes a standard document set analysis process on the standard document set update processing result transferred from the monitoring control unit 10. In this process, for each document, a sentence described in the document is morphologically analyzed, and the expression of each word and the frequency of occurrence of each word in the transferred standard document set are recorded in association with each other. is there. The analysis result is output to the monitoring control unit 10 as a table including two columns of a word expression and an appearance frequency.
In this table, each row describes each word expression and the appearance frequency of the word. This table also
The structure is such that all the corresponding lines can be searched using each word expression as a key.

In the target document input unit 103, the monitoring control unit 1
The target document input process is executed for the target document transferred from 0. This processing is a pre-processing of the subsequent processing, and removes a part of the transferred target document that is not necessary for the processing by the present apparatus. In addition, it is converted into a corresponding character code in the subsequent processing. The processing result is output to the monitoring control unit 10.

The target document analysis unit 104 includes the monitoring control unit 10
The target document analysis processing is executed on the target document input processing result transferred from the server, and the target document analysis result as the processing result is output to the monitoring control unit 10. In the target document analysis process, first, a sentence described in a document is morphologically analyzed, the expression of each word is associated with the frequency of occurrence of each word in the document, and recorded in a table for each document. One table is created for one target document. This table is output to the monitoring control unit 10 as a target document analysis result. This table is composed of two columns, a column for recording a word expression and a column for recording the frequency of occurrence of the word. Further, this table has a structure in which all corresponding rows can be searched using each word expression as a key.

The target document feature calculation unit 105 calculates a target document feature based on the standard document set analysis result and the target document analysis result transferred from the monitoring control unit 10.
The “target document feature” is obtained by comparing the frequency distribution of words in the standard document set with the frequency distribution of words in the target document, and quantifying the difference in the distribution for each word. A word having a larger difference between the appearance frequency distribution in the standard document set and the appearance frequency distribution in the target document has a larger numerical value.
That is, words that are characteristic of the target document have larger numerical values.

In this embodiment, the target document feature is calculated for each word by using the chi-square test for the appearance frequency distribution of the word. The chi-square test can test “whether there is a difference in the distribution of a certain variable in some groups”. In the present invention, this variable is a word in the document. Expectation of the appearance frequency of each word in the target document calculated from the total number of all words in the target document and the standard document set, the total number of all words in the target document, and the total number of all words in the standard document set The χ 2 value is calculated from the distribution of the frequency of occurrence of each word whose value distribution is actually observed in the target document. The larger this value is, the more the distribution is different, and the more such a word is, the more unevenly appearing. In the present invention, the target document feature of each word is calculated using this value.

The target document feature is represented as a table including two columns of an expression of each word and a feature score obtained by digitizing the feature of the word. Each line represents the correspondence between each word expression and the feature information score. The calculation result is transmitted to the monitoring control unit 10
Output to The characteristic expression extraction unit 106 extracts characteristic expressions (characteristic expressions) of the target document from the target document using the target document characteristic information transferred from the monitoring control unit 10 and outputs the extracted expressions to the monitoring control unit 10. .

The feature expression in the target document is a word string having a predetermined number of consecutive words included in the document, a predetermined number of sentences, or a predetermined partial structure in the document. For example, paragraphs)
The average of the feature information scores of the words included in the word strings constituting the continuous word strings, sentences, and partial structures is the largest part.

When a sentence is extracted from a document as a feature expression, the procedure for extracting the feature expression is as follows. First, a feature information score is assigned to all words in the document. Next, for each sentence in the document, the average of the feature information scores assigned to the words constituting the sentence is determined.

The sentence having the maximum feature information score is extracted as the feature expression of the document. Feature expression output unit 10
At 7, the characteristic expression for the target document transferred from the monitoring control unit 10 is transferred to the transfer medium 3 through the monitoring control unit 10.
Output to 0. FIG. 5 shows a standard document set according to an embodiment of the present invention, and FIG. 6 shows an example of a part of a target document according to an embodiment of the present invention.

Hereinafter, a standard document set partially shown in FIG.
The operation of this apparatus will be described with reference to the flowchart of FIG. 4 using an example of a process when it is assumed that the target document shown in FIG. 6 is provided. First, it is determined whether the standard document set has been updated (step 101). Here, it is assumed that it has been found that it has been updated. The standard document set partially shown in FIG. 5 is transferred to the standard document set update unit 101.

The standard document set is a document set, and it is desirable that the number of documents is sufficiently large. In the standard document set updating unit 101, portions unnecessary for the subsequent processing are removed from each document in the standard document set. For example, in the case of an HTML document, the HTML tag is removed. In the case of a word processing document, character decoration and the like are removed. Further, when the codes of the characters constituting the document are various, they are unified into one code. This result is output to the monitoring control unit 10 (Step 102).

The monitoring control unit 10 transfers this to the standard document set analysis unit 102. The standard document set analysis unit 102
Analyze the transferred standard document set. That is, for each document, the document described by the document is subjected to morphological analysis, and the word expression in the standard document set and the appearance frequency of the word are obtained (step 103). FIG. 7 shows an example of a standard document set analysis result according to an embodiment of the present invention. The result is output to the monitoring control unit 10.

The monitoring control unit 10 transfers the standard document set analysis result to the standard document set analysis result storage unit 201. The standard document set analysis result storage unit 201 stores and holds the transferred standard document set analysis result (step 10).
4). As described above, the process regarding the standard document set is completed. Next, it is determined whether or not the target document has been input (step 105). If it has been entered, the process is executed as follows. The target document shown in FIG. 6 is input to the target document input unit 103. The target document input unit 103
Unnecessary parts for subsequent processing are removed from the input document set. Also, it is converted into a character code corresponding to the subsequent processing. The processing result is output to the monitoring control unit 10 (Step 106).

The monitoring control unit 10 includes a target document input unit 103
Is transferred to the target document analysis unit 104. The target document analysis unit 104 analyzes the transferred target document. The target document analysis result is a table in which word expressions constituting the target document and the appearance frequency of each word are recorded. FIG. 8 shows a part of the result. This result is output to the monitoring controller 10 (step 107).

The monitoring control 10 transfers the target document analysis result to the target document analysis result storage unit 202. The target document analysis result storage unit 202 stores and holds the transferred target document analysis result (step 108). Next, target document features are calculated. First, the monitoring control unit 10 analyzes the standard document set analysis result stored and held in the standard document set analysis result storage unit 201 and the target document analysis result stored and held in the target document analysis result storage unit 202. The document is transferred to the document feature calculation unit 105.

The target document feature calculation unit 105 calculates a feature information score by referring to the transferred standard document set analysis result and the target document analysis result. The calculation result is output to the monitoring control unit 10 (Step 109). Some of the results are shown in FIG. The monitoring control unit 10 transfers the target document feature to the target document feature storage unit 203. Target document feature storage unit 20
3 stores and holds the transferred target document features (step 110).

Next, a characteristic expression is extracted from each document in the target document. First, the monitoring control unit 10 transfers the feature information stored and held in the target document feature storage unit 203 to the feature expression extraction unit 107. The feature expression extraction unit 107 extracts a feature expression from the target document by using the feature information transferred from the monitoring control unit 10, and compares the result with the monitoring control unit 10.
(Step 111). FIG. 10 shows the extraction result. In the example shown in the figure, “2”, “war”,..., And “report” have values of 100 or more, and a sentence including these feature expressions is “on April 1... ] Is extracted.

The monitoring control unit 10 includes a feature expression extraction unit 107
Is transferred to the feature expression storage unit 204. The feature expression storage unit 204 stores and holds the transferred feature expression (step 112). The feature expression output unit 107
A feature expression output process is performed in association with the feature expression stored in the feature expression storage unit 204 and the document from which the feature expression has been extracted (step 113).

The monitoring control unit 10 includes a feature expression storage unit 204
Is associated with the document from which it is extracted and output to the transfer medium 30 (step 1).
14). Thus, what is output to the transfer medium 30 is
The contents shown in FIG. 10 from the target document shown in FIG. 6 (“April 1...” On the second to fourth lines from the top in FIG. 6)
Becomes

In the above embodiment, various defined values are used, but these values are design values, and may be changed as necessary as described below. -Although the concept of the chi-square test is used for calculating the characteristic information, it may be calculated by another method. The word is used as the unit for calculating the feature score, but this unit may be a character or a character string of a fixed length.

Although the above embodiment has been described based on the configuration of FIG. 3, the present invention is not limited to this example.
The monitoring control unit 10 and the standard document set updating unit 10 shown in FIG.
1. Standard document set analysis unit 102, target document input unit 10
3. Target document analysis unit 104, target document feature calculation unit 10
5. The feature expression extraction unit 106 and the feature expression output unit 107 are constructed as programs and stored in a disk device connected to a computer used as a document summarization device, or a portable storage medium such as a floppy disk or a CD-ROM. In addition, the present invention can be easily realized by installing the present invention when implementing the present invention.

The present invention is not limited to the above embodiments, but can be variously modified and applied within the scope of the claims.

[0061]

As described above, according to the present invention, it is possible to provide a document summarizing apparatus capable of extracting feature information and adding it to a document as a summary without preparing feature information extraction knowledge in advance. It is possible. As a result, it is possible to appropriately summarize even when there is a difference between the content assumed at the start of use and the content of the target document or when new information is included.

Further, since the document information characterizing each document can be extracted from each document in the document set as a summary, the document information can be efficiently selected from the document set of the search result by applying to the output editing device of the document search system.・ Can be browsed.

[Brief description of the drawings]

FIG. 1 is a diagram for explaining the principle of the present invention.

FIG. 2 is a principle configuration diagram of the present invention.

FIG. 3 is a configuration diagram of a document summarizing apparatus of the present invention.

FIG. 4 is a flowchart of a monitoring control process executed by the monitoring control unit according to one embodiment of the present invention.

FIG. 5 is an example of a standard document set according to an embodiment of the present invention.

FIG. 6 is an example of a part of a target document according to an embodiment of the present invention.

FIG. 7 is an example of a standard document set analysis result according to an embodiment of the present invention.

FIG. 8 is an example of a target document analysis result according to an embodiment of the present invention.

FIG. 9 is an example of a part of target document feature information according to an embodiment of the present invention.

FIG. 10 is an example of a feature expression according to an embodiment of the present invention.

[Explanation of symbols]

 REFERENCE SIGNS LIST 10 monitoring control unit 20 input storage device, input storage means 30 transfer medium 40 feature information calculation means 50 feature part extraction means 60 summary creation means 101 standard document set update section 102 standard document set analysis section 103 target document input section 104 target document analysis Unit 105 target document feature calculation unit 106 feature expression extraction unit 107 feature expression output unit 201 standard document set analysis result storage unit 202 target document analysis result storage unit 203 target document feature storage unit 204 feature expression storage unit

Claims (7)

[Claims] 1. A document summarizing method for extracting and characterizing information characterizing a target document read from an input storage unit storing document data, wherein the target document to be summarized is a standard document which is an entire set of documents that can be processed. Comparing the document with a set of documents, calculating characteristic information characterizing the target document from the target document, extracting, from the target document, a portion that most characterizes the target document based on the characteristic information; A document summarization method characterized by outputting as a summary characterizing a document. 2. A document summarization apparatus for extracting information characterizing a target document read from an input storage unit storing document data and summarizing the information, wherein the summarization target document is processed by a whole document set that can be processed. A feature information calculating means for comparing feature information that characterizes the target document from the target document by comparing with a standard document set; and a feature for extracting, from the target document, a part that most characterizes the target document based on the feature information. A document summarizing apparatus, comprising: a part extracting unit; and a summary creating unit that outputs a part extracted by the characteristic part extracting unit as a summary characterizing the target document. 3. A standard document set updating means for acquiring a standard document set from the input storage means; and analyzing each document in the standard document set given to the standard document set updating means to constitute the document. A standard document set analysis unit for calculating a word and an appearance frequency of the word; a standard document set analysis result storage unit for storing a word in the standard document set and an appearance frequency of the word in association with each other; A target document input unit that receives the target document; a target document analysis unit that analyzes the target document and calculates words constituting the target document and an appearance frequency of the word; and a word in the target document and each word. A target document analysis result storage unit that stores appearance frequencies in association with each other, and the feature information in the target document, the standard document analysis result stored in the standard document set analysis result storage unit,
A target document feature calculation unit that calculates using the target document analysis result stored in the target document analysis result storage unit; and a target document that stores feature information of the target document calculated by the target document feature calculation unit. A feature storage unit; a feature expression extraction unit that extracts a feature expression from the target document by using feature information of the target document stored in the feature information storage unit; 3. The document summarizing apparatus according to claim 2, further comprising: a feature expression storage unit configured to store the extracted feature expression; and a feature expression output unit configured to provide the feature expression stored in the feature expression storage unit to a transfer medium. 4. The document summarizing apparatus according to claim 3, wherein the feature information calculating unit includes a unit using a chi-square test for calculating the feature information. 5. The target document feature calculation means, as a unit for calculating a feature information score obtained by referring to the standard document set analysis result and the target document analysis result,
4. The document summarizing apparatus according to claim 3, wherein a character or a character string having a fixed length is used. 6. A storage medium storing a document summarization program for extracting and characterizing information characterizing a target document read from an input storage means storing document data, wherein the target document to be summarized is processed. A feature information calculating process of comparing the obtained document set with a standard document set and calculating feature information characterizing the target document from the target document; and, based on the feature information, a part that most characterizes the target document. A storage medium storing a document summarization program, comprising: a feature portion extraction process for extracting from a document; and a summary creation process for outputting a portion extracted in the feature portion extraction process as a summary characterizing the target document. 7. A standard document set updating process for acquiring a standard document set from the input storage means, and analyzing each document in the standard document set given to the standard document set updating process to constitute the document. A standard document set analysis process for calculating a word and an appearance frequency of the word, and storing the word in the standard document set and the appearance frequency of the word in the standard document set analysis result storage means in association with each other; A target document input process for receiving the target document from the target document, analyzing the target document, calculating the words constituting the target document and the appearance frequency of the word, and storing the words in the target document in the target document analysis result storage means. A target document analysis process for storing the frequency of occurrence of each of the words in association with each other; and the standard information stored in the standard document set analysis result storage means. And written analysis result,
A target document feature calculation process that calculates using the target document analysis result stored in the target document analysis result storage unit and causes the target document feature storage unit to store the calculated feature information of the target document; A feature expression extraction process of extracting a feature expression from the target document using feature information of the target document stored in the feature information storage unit, and storing the feature expression in the feature expression storage unit; 7. A storage medium storing the document summarizing program according to claim 6, further comprising: a characteristic expression output process of providing the characteristic expression stored in the means to a transfer medium.