JPH10340271A - Document abstract preparation device, and storage medium where document abstract generation program is recorded - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a document abstract preparation device for preparing an abstract for judging the necessity of a document obtained as the result of document retrieval with a simple operation in accordance with retrieval intention. SOLUTION: For preparing the abstract of a document in the document abstract preparation device, a document division means 1 divides document data into document elements. A retrieval intent holding means 2 holds inputted retrieval intent. An adaptability calculation means 3 calculates adaptability on the retrieval intention of the divided document element. A document element extraction means 4 extracts an adapted document element from the document element based on calculated adaptability. An abstract generation means 5 generates the abstract of the document by the adapted document element.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document abstract creation apparatus and a storage medium storing a document abstract creation program.
In particular, the present invention relates to a document abstract creation apparatus for creating an abstract of a document, and a storage medium storing a document abstract creation program for creating an abstract of a document.

[0002]

2. Description of the Related Art Conventionally, various methods have been considered in order to retrieve a document from a large amount of document data as needed. As the amount of document data to be searched increases, faster and more appropriate search processing is required, and document search using a computer or the like is also used today.

[0003] The ratio of a required document to the result of a document search is called a relevance ratio. At present, various techniques for increasing the relevance rate have been developed, but ultimately, the user will need to read the documents obtained as a result of the search one by one to determine whether or not the documents are necessary.

However, if the number of obtained documents is large,
The time and effort required to make this determination can be enormous. Therefore, there is a demand for presenting the document of the search result in the form of an abstract and supporting this determination.

[0005] As a method of preparing an abstract of a search result document, first, an important word (keyword) is registered in advance for each field of the document, and a method of extracting a sentence containing the keyword is disclosed in Japanese Patent Laid-Open Publication No. HEI 2 (1998) -197686. No. 93866.

However, according to this method, only an abstract of a document belonging to a limited field in which a keyword is registered in advance can be created. In addition, the user must sequentially register additional keywords, and the effort is not negligible.

[0007] Also, the document "Fast Generation of Abstrac"
ts from General Domain Text Corpora by Extracting
Relevant Sentences "(Zechner, COLONG'96) creates a weighted table of keywords. When creating an abstract, the sum of the weights of the keywords contained in each sentence constituting the document is calculated,
Score for each sentence. Then, a certain number of sentences having a high score are extracted and arranged in the order of appearance to create an abstract of the document.

However, in this method, a score is calculated for each sentence mechanically regardless of the length of the sentence. Therefore, the score of the sentence calculated from the weight of each keyword appropriately represents the importance of the sentence. Not necessarily.

In any of the above-described methods, an abstract is created irrespective of the user's intention or purpose of document search. Therefore, an unextracted portion may include information necessary for the user. This is insufficient as support for determining whether or not the document obtained as a result of the document search matches the user's intention.

[0010] Therefore, it has been considered to use keywords, search formulas, and the like input for document search when creating an abstract. Japanese Patent Laid-Open Publication No. Hei 6-309368 discloses a method of extracting a sentence including a character string input for a document search in the process of searching for a structured document and listing the extracted sentences to create an abstract. Is shown in

In addition, the context structure of the document obtained as a result of the search is analyzed in detail, and the results are used to present backgrounds and examples related to topics specified by the user, and to create an abstract from the viewpoint desired by the user. The technology is disclosed in JP-A-7-182373.

[0012]

However, in the method of extracting and enumerating sentences including a character string input for document search, if the character string frequently appears in a document,
There was a problem that the size of the abstract became huge, not much different from the original document.

[0013] Further, in the technology of creating an abstract from a viewpoint desired by a user by analyzing a context structure, a detailed description is given of how a topic designated by the user is positioned in an original document. There is a problem in that if a large number of documents are retrieved as a result of analysis, complicated operations are required only to create abstracts.

SUMMARY OF THE INVENTION The present invention has been made in view of the above points, and provides a document abstract in which an abstract for judging the necessity of a document obtained as a result of a document search is created by a simple operation according to a user's intention. It is an object to provide a creation device.

Another object of the present invention is to provide a document abstract creation program for causing a computer to create an abstract for judging the necessity of a document obtained as a result of a document search by a simple operation according to a user's intention. It is to provide a recorded storage medium.

[0016]

According to the present invention, in order to solve the above-mentioned problems, in a document abstract creation apparatus for creating an abstract of a document, a document dividing means for dividing the document into document elements, Means for holding user intention, means for calculating the degree of suitability of the divided document elements with respect to the user intention, and extracting a suitable document element from the document element based on the degree of suitability A document abstract creation device is provided, comprising: a document element extraction unit; and an abstract creation unit that creates an abstract of the document using the compatible document element.

When an abstract of a document is created by such a document abstract creating apparatus, the document dividing means divides the document into document elements. The user intention holding means holds the input user intention. The relevance calculating means calculates the relevance of the divided document elements to the user's intention. The document element extracting means extracts a compatible document element from the document element based on the degree of relevance. Then, the abstract creating means creates an abstract of the document using the relevant document element.

As described above, in the first document abstract creation apparatus of the present invention, a document is divided into document elements, and a document element having a high degree of compatibility with the user's intention is extracted to create an abstract of the document. An abstract for judging the necessity of the user can be created by a simple operation according to the user's intention.

According to another aspect of the present invention, there is provided a document abstract creation apparatus for creating an abstract of a document.
User intention holding means for holding input user intention, important keyword extracting means for extracting important keywords from the user intention, and important sentence extracting means for extracting important sentences containing the important keywords from the document And an abstract creating means for creating an abstract of the document based on the important sentence.

When an abstract of a document is created by such a document abstract creation device, the user intention holding means holds the input user intention. The important keyword holding means extracts important keywords from the user's intention. The important sentence extracting means extracts an important sentence including an important keyword from the document. Then, the abstract creating means creates an abstract of the document with the important document.

As described above, in the second document abstract creation apparatus of the present invention, an important sentence including an important keyword is extracted from a document to create an abstract of the document. It can be created by a simple operation according to the user's intention.

Further, according to the present invention, in order to solve the above-mentioned problems, in a storage medium storing a document abstract creation program for creating an abstract of a document, a computer is provided with a document dividing means for dividing the document into document elements. Means for holding user intention, means for calculating the degree of suitability of the divided document elements with respect to the user intention, and extracting a suitable document element from the document element based on the degree of suitability A storage medium storing a document abstract creation program characterized by functioning as a document element extraction unit and an abstract creation unit for creating an abstract of the document using the relevant document element is provided.

When a computer prepares an abstract of a document using such a program for preparing a document abstract, the computer is provided as a document dividing means for dividing the document into document elements and a user intention holding means for holding the input user intention. Function. Further, the computer also functions as a degree-of-fit calculating means for calculating the degree of suitability of the divided document element to the user's intention, and as a document element extracting means for extracting a relevant document element from the document element based on the degree of matching. Further, the computer functions as an abstract creating unit that creates an abstract of the document using the compatible document element.

As described above, in the document abstract creation program of the present invention, a function of extracting an important sentence including an important keyword from a document and creating an abstract of the document is realized, so that the necessity of the document is determined. Abstract for
It can be created by a simple operation according to the user's intention.

[0025]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the principle configuration of the first document abstract creation device of the present invention. The first document abstract creation device according to the present invention comprises a document division unit 1, a user intention holding unit 2, a matching degree calculation unit 3, a document element extraction unit 4, and an abstract creation unit 5. An abstract is created from the data according to the user's intention.

Here, the document dividing means 1 divides the document data into document elements A, B, C. The user intention holding means 2 holds the input user intention. The relevance calculating means 3 calculates the divided document elements A, B, C,.
Calculate the degree of conformity to the user's intention. Document elements A, C
If the relevance calculated for is high, the document element extraction unit 4 extracts the document elements A and C as the relevant document elements based on the relevance. The abstract creating means 5 creates an abstract of the document based on the extracted conforming document elements A and C.

As described above, in the first document abstract creation apparatus of the present invention, a document is divided into document elements, and a document element having high relevance to the user's intention is extracted to create an abstract of the document. An abstract for judging the necessity of the user can be created by a simple operation according to the user's intention.

FIG. 2 is a diagram showing an embodiment of a document search device to which the first document abstract creation device of the present invention is applied.
The illustrated document search device 10 is connected to an input device 20 that receives an input from a user and an output device 30 that outputs an abstract created for the user. This document search device 10 may be realized by a computer or the like. Note that the output device 30 may be configured with only a display device such as a monitor.

Here, the document retrieval device 10 includes a document data storage unit 11, a document designation unit 12, a document division unit 13, a user intention storage unit 14, a fitness calculation unit 15, a fitness storage unit 16,
It is composed of an abstract creation unit 17.

The document data storage section 11 stores the input document data, and when the designation of the document data is input, outputs the corresponding document data. Here, the document data storage unit 1
The document data stored in 1 may be a group of documents obtained as a result of the document search. The document specification unit 12 receives specification of document data for which an abstract is to be created among the stored document data.

The document dividing section 13 divides the designated and output document data into document elements. Note that the document division unit 1
In No. 3, a plurality of division units for dividing the document are defined. If the abstract creation processing does not proceed smoothly, the division unit is changed and the document data is divided again.

The user intention storage unit 14 receives an input of a user intention, which is the purpose of preparing an abstract of a document, and stores it. The fitness calculating unit 15 calculates
The degree of conformity with the user intention is calculated for each of the divided document elements. The relevance storage unit 16 stores the relationship between the calculated relevance and the document element.

Then, the abstract creating unit 17 extracts a matching document element from the document elements based on the matching degree stored in the matching degree storage unit 16 and creates an abstract of the document. Next, a procedure for creating an abstract of a document in the document search device 10 will be described.

FIG. 3 is a flowchart for explaining the procedure for creating a document abstract in the document search apparatus 10 shown in FIG. It is assumed that document data is stored in the document data storage unit 11 in advance. Hereinafter, the description will be given along the step numbers in the figure. [S1] The user intention storage unit 14 receives, via the input device 20, an input of a search expression indicating the user intention, which is the purpose of creating an abstract of a document, and stores the input. [S2] The document specification unit 12 receives the specification of the target document for preparing the abstract via the input device 20, and inputs this to the document data storage unit 11. [S3] The document division unit 13 supplied with the specified document from the document data storage unit 11 divides this document into document elements in units of sentences. [S4] The relevance calculation unit 15 acquires the user intention from the user intention storage unit 14 and calculates the relevance for all the document elements input from the document division unit 13. The calculated relevance is stored in the relevance storage unit 16 together with the correspondence with the document element. [S5] The abstract creation unit 17 determines whether or not there is a document element with a suitability of 1.0 based on the suitability of each document element stored in the suitability storage unit 16. If there is a document element with a fitness of 1.0, the process proceeds to step S6. Fitness 1.0
If no document element exists, the process proceeds to step S7. [S6] Since the document elements divided from the document dividing section 13 are input to the abstract creating section 17, the document elements having the relevance of 1.0 are extracted as the conforming document elements. [S7] The abstract creating unit 17 calculates the
It is determined whether the division unit can be increased. If the division unit can be increased, the process proceeds to step S8. If the division unit is already the maximum and cannot be increased any further, the process proceeds to step S9. [S8] The abstract creation unit 17 requests the document division unit 13 to change the division unit. Upon receiving the division unit change request, the document classification unit 13 redivides the document supplied from the document data storage unit 11 into large division units, and repeats step S4.
Proceed to. At this time, if the division unit is “sentence” at this time, the division unit is “paragraph”, and if the division unit is “paragraph”, it is “paragraph”. If the division unit is “section”, the section is “section”. If the division unit is “section”, the section is “chapter”. [S9] Since the document elements divided from the document dividing section 13 are input to the abstract creating section 17, the document elements having the maximum relevance among them are extracted as the conforming document elements. [S10] The abstract creation unit 17 creates an abstract of the document from the extracted relevant document information. The created abstract is output to the output device 30. [S11] It is determined whether to create an abstract of the next document using the same search formula. When preparing an abstract, the process proceeds to step S2. If the abstract is not created, the process of this flowchart ends.

The document retrieval apparatus 10 of the present invention creates an abstract of a document according to the above procedure. Here, how the document retrieval apparatus 10 actually creates an abstract of a document will be described step by step.

FIG. 4 shows the document data storage unit 1 shown in FIG.
FIG. 2 is a diagram illustrating an example of two documents stored in a first document; When the document data 40 and 50 are stored in the document data storage unit 11 and the user wants to create an abstract of the document relating to the joint development and the joint research of “A Co., Ltd.”, the search formula is “A and Kyoko and ( Development or research) "is input. This search expression indicates that the keyword "Co. A" and the keyword "joint" and the keyword "development" are included at the same time, or that the keyword "stock A" and the keyword "joint" and the keyword "research" are included simultaneously I have.

This search expression is received by the user intention storage unit 14 via the input device 20, and the document data 40, 50
Is specified, the document creating apparatus 10 first divides the document data 40, 50 into document elements in units of "sentence".

FIG. 5 is a diagram showing a state where the document data 40 shown in FIG. 4 is divided into document elements in units of [sentence]. When the document is divided, each document element has a NO. Is appended. Here, when the relevance of each document element with the search formula is calculated, the document element NO. 3 includes the keyword “A”, the keyword “joint”, and the keyword “development” at the same time, and the matching degree is 1.0. The document element No. 6 includes only the keyword “stock company A”, the matching degree is 0.33. Note that the other document elements do not include any keywords, so the relevance is 0.

In this case, since there is a document element having a relevance of 1.0, there is no need to further divide the document. FIG. 6 is a diagram showing an abstract created according to the calculation result of the degree of suitability of the document element shown in FIG.

In the present invention, the abstract creating section 17 can create three types of abstracts. The abstract 41 is created by extracting only the document elements having the relevance of 1.0. Since the purpose of the user to create the abstract is to search for documents related to the joint development and joint research of “A Co., Ltd.”, when the abstract 41 is displayed, the user can determine whether the document data 40 is necessary.

The abstract 42 is created by extracting and adding the title of the original document to a document element having a relevance of 1.0. By clarifying the subject of the created abstract included in the document, it is possible to assist the user's judgment.

Further, the abstract 43 is prepared so as to extract a document element having a relevance of 1.0 and highlight the keywords contained in the document element. Here, the presence of the keyword is emphasized by reverse display, but the number of points and the font may be changed or underlined. When the abstract to be created is long, the keyword can be emphasized in this way to support the user's judgment.

FIG. 7 is a diagram showing a state in which the document data 50 shown in FIG. 4 is divided into document elements in units of “sentence”. When the relevance of each document element with the search formula is calculated, the document element NO. 1 includes the keyword “joint” and the keyword “development” at the same time, and the relevance is 0.67.
Becomes The document element No. 3. The keyword "joint"
And the keyword "development" are included at the same time, and the degree of matching is 0.67. Further, the document element No. 5 and N
O. 7 includes the keyword “A Co., Ltd.”, the matching degree is 0.33. Since the other document elements do not include any keywords, the relevance is 0.

In this case, since there is no document element having a relevance of 1.0, the unit of document division is increased. FIG.
FIG. 5 is a diagram showing a state in which the document data 50 shown in FIG.

When the relevance of each document element with the retrieval formula is calculated, the document element NO. 1 includes the keyword “joint” and the keyword “development” at the same time, and the degree of matching is 0.67. Document element No. 3 is the keyword "A", the keyword "joint" and the keyword "development"
Are included at the same time, and the matching degree is 1.0. The document element No. 4 includes the keyword “A Co., Ltd.”, and the matching degree is 0.33.

In this case, since there is a document element having a relevance of 1.0, there is no need to further divide the document. FIG. 9 is a diagram showing an abstract created in accordance with the calculation result of the degree of suitability of the document element shown in FIG.

The abstract 51 extracts only document elements having a relevance of 1.0. When the abstract 51 is displayed, the user can determine whether the document data 50 is necessary. In addition, the abstract 52 extracts and adds the title of the original document to a document element having a matching degree of 1.0. Further, the abstract 53 extracts a document element having a relevance of 1.0, and highlights and displays the keywords included in the document element.

As described above, in the embodiment of the present invention, a document is divided into document elements, and a document element having high relevance to the user's intention is extracted to create an abstract of the document.
Abstracts for determining the necessity of a document can be created by a simple operation according to the user's intention.

In the above description, the division units used for dividing the document by the document division unit 13 are defined as “sentence”, “paragraph”, “chapter”, and “section”. Not all need to be defined. That is, it is also possible to define only frequently used division units such as “sentence” and “paragraph”.

Further, when there is no document element having a relevance ratio of 1.0 in units of "sentence", the document unit may be changed to "consecutive n sentences". “N sentence” is extracted so as not to be restricted by the logical structure of the sentence. That is, if a document composed of the sentence “a”, the sentence “b”, the sentence “c”, the sentence “d”... Is divided into the document units “two sentences”, the document element becomes the sentence “ab” , Sentence "bc", sentence "cd", and so on.

Further, when the abstract to be created is too large, filtering may be performed so that the number of document elements constituting the abstract becomes an appropriate amount. For example, in a document element having a unity such as a paragraph, a gist is often described at the beginning or end of the sentence, and therefore, the document element at such a position is preferentially extracted. A document element in which the distance of the keyword group included in the search formula is the smallest may be extracted, or a document element including the keyword newly input by the user may be extracted. Further, it is also possible to extract only the first certain number of document elements of the created abstract.

By reducing the size of the abstract using these methods, the user can read the abstract in a shorter time and determine whether a document is necessary or not. In the first document retrieval apparatus 10 described above, among the newspaper articles for about one year, abstracts were obtained for 26 documents that were hit by the retrieval formula "Aand joint and (development or research)". An experiment was conducted to determine the necessity of the document.

As a result, in about 77% of abstracts, it was possible to make the same judgment as when reading and judging the original document. The 26 abstracts created by the document retrieval apparatus 10 were composed of about 27% of the number of characters compared to the original document.

In a conventional abstract creation apparatus that extracts the text of the first paragraph when creating an abstract of a newspaper article, the scale of the abstract compared to the original document is 34% on average. In addition, about 57% of the abstracts created in this way were judged to be unnecessary or unnecessary after reading the original document.

Next, a second document abstract creation apparatus of the present invention will be described. FIG. 10 is a block diagram showing the principle configuration of the second document abstract creation device of the present invention.

The second document abstract creation device of the present invention comprises a user intention holding means 61 and an important keyword extraction means 62
, An important sentence extracting unit 63 and an abstract creating unit 64, which create an abstract according to the input user's intention from the document data.

Here, the user intention holding means 61 holds the input user intention. Important keyword extraction means 62
Extracts an important keyword from the user intention held in the user intention holding means 61. Important sentence extraction means 63
Extracts important sentences A and C including important keywords from document data. The abstract creating means 64 creates an abstract of the document based on the extracted important sentences A and C.

As described above, in the second document abstract creation apparatus of the present invention, an important sentence including an important keyword is extracted from a document to create an abstract of the document. It can be created by a simple operation according to the user's intention.

FIG. 11 is a diagram showing an embodiment of a document search device to which the second document abstract creation device of the present invention is applied. The illustrated document search device 70 is connected to an input device 80 that receives an input from a user and an output device 90 that outputs an abstract created for the user. This document search device 70 may be realized by a computer or the like. In addition,
The output device 90 may be composed of only a display device such as a monitor.

Here, the document search device 70 comprises a document data storage unit 71, a user intention storage unit 72, a keyword extraction mechanism unit 73, a keyword holding mechanism unit 74, an important sentence selection mechanism unit 75, and an abstract creation unit 76. ing.

The document data storage section 71 stores the document data of the document group obtained as a result of the document search, and outputs the corresponding document data when the user inputs the designation of the document data via the input device 80. .

The user intention storage unit 72 receives an input of the user intention, which is the purpose of preparing the abstract of the document, and stores it. The user intention input here may be a search expression composed of words and logical operators or a natural language.

The keyword extraction mechanism 73 extracts important keywords from the input user intention. When the user's intention is input by a search formula, words included in the search formula are extracted as important keywords. Further, when the user intention is input in a natural language, the user intention is subjected to morphological analysis, and an independent word is extracted from the included words to be an important keyword.

If there is an instruction, the keyword extraction mechanism 73 extracts a document keyword from the input document data. Here, tf * ID is used to extract the document keyword.
A method using the F product is adopted. This is because tf (term frequencnc) is used for all words that appear in the document data.
y, the frequency of occurrence of the word in the document) and DF (Doc
In this method, the inverse IDF (Inverted DF) of the ument Frequency, the number of documents in the document set where the word appears, is calculated, and the tf * IDF product shown in the following equation (1) is calculated.

[0065]

Tf * IDF = tf * log (N / DF) (1) where N indicates the total number of documents in the document set. The keyword extraction mechanism 73 sets a certain number of words whose calculated tf * IDF products are equal to or greater than a certain value or words whose calculated tf * IDF products are large as a document keyword. At this time, t
Even if the value of the f * IDF product is large, a word that does not need to be a document keyword is stored as a stop word list, and a word selected as a document keyword can be made appropriate.

The keyword holding mechanism 74 holds the extracted important keywords and document keywords. The important sentence selection mechanism 75, based on the input keyword,
An important sentence is selected from the supplied document data. The selection of the important sentence will be described later with reference to a flowchart.

The abstract creation section 76 arranges the selected important sentences in the order in which they appear, and creates an abstract of the document. here,
The procedure for selecting an important sentence in the important sentence selection mechanism 75 will be described. There are two methods for selecting important sentences.

FIG. 12 is a flowchart showing the procedure of the first important sentence selection method in the important sentence selection mechanism 75 shown in FIG. Hereinafter, description will be given along the step numbers in the figure. [S21] The tf * IDF product of each word in the supplied document is calculated. [S22] The tf * IDF products of all important keywords are set to values higher than any tf * IDF products calculated in S21. [S23] A word list L of important keywords is created. [S24] For each sentence in the supplied document, the score is calculated by summing the tf * IDF products of the words included in the sentence. [S25] A fixed number (N) of sentences are selected in descending order of the score, and these are set as important sentences. [S26] It is determined whether an important keyword is included in the important sentence selected in step S25. If an important keyword is included, the process proceeds to step S27. If no important keyword is included, the process proceeds to step S28. [S27] The important keyword determined to be included in the important sentence in step S26 is deleted from the word list L. [S28] It is determined whether the word list L is empty. If the word list L is empty, the processing of this flowchart ends. If the word list L is not empty, the process proceeds to step S29. [S29] One is added to N to obtain a new N. [S30] It is determined whether an important keyword is included in the Nth sentence S of the score. If an important keyword is included, the process proceeds to step S31. If no important keyword is included, the process proceeds to step S29 again. [S31] The important keywords determined to be included in the sentence S in step S30 are deleted from the word list L. [S32] The sentence S is additionally selected as an important sentence, and the process proceeds to step S28 again.

In the second document retrieval apparatus 70 of the present invention, the important sentences selected in this way are arranged in the order of appearance, and an abstract of the document is created, thereby making it possible to create a document abstract according to the user's intention. it can.

Next, a second method for selecting an important sentence will be described. In the second method, when a document keyword for document data is weighted beforehand, an important sentence is selected using the document keyword and the weight.

FIG. 13 is a flowchart showing the procedure of the second important sentence selection method in the important sentence selection mechanism 75 shown in FIG. Hereinafter, description will be given along the step numbers in the figure. [S41] A keyword table is created from document keywords and important keywords. The weight of the important keyword is set to a value larger than the maximum value of the weight of the document keyword. [S42] A score for each sentence in the document is calculated, and a sentence table is created. The score is calculated by summing the weights of the keywords included in the text. [S43] The sentence with the highest score is regarded as an important sentence, and the sentence is moved from the sentence table to the important sentence list. [S44] The keyword table is updated. That is,
In step S43, the weight added to the keyword included in the sentence that is regarded as an important sentence is reduced, and a check that the sentence is used is added. [S45] It is determined whether or not there is a keyword in the keyword table to which a weight greater than the lower weight limit has been added.
If there is a keyword to which a weight greater than the lower weight limit is added, the process proceeds to step S42. If there is no such keyword, the process of this flowchart ends.

Here, an example in which an important sentence is selected by the second selection method and an abstract of a document is created will be described.
FIG. 14 is a diagram illustrating an example of document data stored in the document data storage unit 71 illustrated in FIG.

The document data 100 is composed of a title and nine paragraphs. FIG. 15 shows a weighted document keyword added to the document data shown in FIG.

A table 101 of weighted document keywords is prepared in advance. Here, when the search expression “(China | Chinese) * cuisine” (“|” indicates a logical sum and “*” indicates a logical product) is input, “China” or “Chinese” and “ “Cooking” becomes an important keyword, and a keyword table is created.

FIG. 16 is a diagram showing a keyword table created first from the document data shown in FIG. The keyword table 102 describes a document keyword and an important keyword, and the important keyword is weighted with a value (here, 5) larger than the maximum value (here, 4) of the weight of the document keyword. In addition, a used column for checking whether or not each keyword has been used is provided.

FIG. 17 is a diagram showing a sentence table created based on the keyword table 102 shown in FIG. The sentence table 103 describes the appearance order and score of each sentence.

An abstract of a document is created from the document data 100, the keyword table 102, and the sentence table 103 as described above. FIG. 18 is a diagram showing an abstract of the document data shown in FIG. 14, wherein (A) is an abstract created by the second document abstracting apparatus of the present invention, and (B) is created from only document keywords. It is a figure which shows an abstract.

Here, the weight of the keyword is reduced to 1/1 by updating the keyword table in step S44.
The abstract prepared when the weight lower limit value is set to 0 and the weight lower limit value is set to 2 is described as an example.

In the figure, the abstract 104 includes sentences 104 including “China”, “China”, and “cooking” included in the retrieval formula.
It can be seen that a is included. As described above, according to the second document search device of the present invention, it is determined whether or not the document 100 which does not originally have the search formula “(China | Chinese) * cooking” as a subject matches the search purpose. Abstracts that can be judged quickly can be created.

In the above description, the sum of the weights of the keywords included in the text is calculated as the score of the text. However, the product of the frequency and the weight of the keyword included in each text may be summed to obtain the score of the text. .

Also, by changing the weight lower limit value,
It is possible to control the amount of abstract documents.

[0082]

As described above, in the first document abstract creation apparatus of the present invention, a document is divided into document elements, and a document element having high relevance to the user's intention is extracted to create an abstract of the document. With this configuration, an abstract for determining whether or not a document is necessary can be created by a simple operation according to the user's intention.

Further, as described above, the second document abstract creation apparatus of the present invention is configured to extract an important sentence including an important keyword from a document to create an abstract of the document. An abstract for judging can be created by a simple operation according to the user's intention.

Further, the document abstract creation program of the present invention has a function of extracting an important sentence including an important keyword from a document and realizing a function of creating an abstract of the document. An abstract for determining unnecessaryness can be created by a simple operation according to the user's intention.

[Brief description of the drawings]

FIG. 1 is a block diagram showing the principle configuration of a first document abstract creation device of the present invention.

FIG. 2 is a diagram showing an embodiment of a document search device to which the first document abstract creation device of the present invention is applied.

FIG. 3 is a flowchart illustrating a procedure for creating a document abstract in the document search device shown in FIG. 2;

FIG. 4 is a diagram illustrating 2 stored in a document data storage unit shown in FIG.
FIG. 4 is a diagram showing an example of one document.

FIG. 5 is a diagram showing a state in which the document data shown in FIG. 4 is divided into document elements in units of [sentence].

FIG. 6 is a diagram showing an abstract created in accordance with a calculation result of the degree of suitability of the document element shown in FIG. 5;

FIG. 7 is a diagram showing a state where the document data shown in FIG. 4 is divided into document elements in units of “sentence”.

FIG. 8 is a diagram showing a state where the document data shown in FIG. 7 is divided into document elements in units of “paragraphs”.

9 is a diagram showing an abstract created according to the calculation result of the degree of matching of the document element shown in FIG. 8;

FIG. 10 is a block diagram showing the principle configuration of the second document abstract creation device of the present invention.

FIG. 11 is a diagram showing an embodiment of a document search device to which the second document abstract creation device of the present invention is applied.

FIG. 12 is a flowchart showing a procedure of a first important sentence selection method in the important sentence selection mechanism shown in FIG. 11;

13 is a flowchart showing a procedure of a second important sentence selection method in the important sentence selection mechanism shown in FIG. 11;

FIG. 14 is a diagram illustrating an example of document data stored in the document data storage unit illustrated in FIG. 11;

FIG. 15 illustrates a weighted document keyword added to the document data illustrated in FIG. 14;

FIG. 16 is a diagram showing a keyword table created first from the document data shown in FIG. 14;

FIG. 17 is a diagram showing a sentence table created based on the keyword table shown in FIG. 16;

18A and 18B are diagrams showing an abstract of the document data shown in FIG. 14, wherein FIG. 18A is an abstract created by the second document abstract device of the present invention, and FIG. It is a figure which shows an abstract.

[Explanation of symbols]

 1 Document dividing means 2 User intention holding means 3 Relevance calculating means 4 Document element extracting means 5 Abstract making means

Claims (8)

[Claims] 1. A document abstract creating apparatus for creating an abstract of a document, comprising: a document dividing unit that divides the document into document elements; a user intention holding unit that holds an input user intention; Relevance calculating means for calculating relevance of the element to the user's intention; document element extracting means for extracting a relevant document element from the document element based on the relevance; An abstract creation means for creating a document abstract, and a document abstract creation apparatus comprising: 2. The document division unit defines a plurality of division units for dividing the document, and changes the division unit when a suitable document element is not extracted by the document element extraction unit. The document abstract creation apparatus according to claim 1, wherein 3. An important document which extracts an important document element from the relevant document element based on the stored important document element condition, and inputs the important document element to the abstract creation means instead of the relevant document element. 2. The document abstract creation device according to claim 1, further comprising an element extraction unit. 4. A document abstract creation apparatus for creating an abstract of a document, a user intention holding means for holding an input user intention, an important keyword extraction means for extracting an important keyword from the user intention, A document abstract creating apparatus, comprising: an important sentence extracting unit that extracts an important sentence including the important keyword from a document; and an abstract creating unit that creates an abstract of the document based on the important sentence. 5. The document abstract creation apparatus according to claim 4, wherein said important keyword extracting means sets a document keyword extracted from said document as said important keyword. 6. The document abstract creation apparatus according to claim 4, further comprising display means for highlighting the important keyword when displaying the abstract. 7. A storage medium storing a document abstract creation program for creating an abstract of a document, a computer comprising: a document dividing means for dividing the document into document elements; and a user intention holding for input user intention. Means for calculating a degree of suitability of the divided document elements with respect to the user intention; a document element extracting means for extracting a suitable document element from the document element based on the degree of suitability; A storage medium storing a document abstract creation program, which functions as an abstract creation means for creating an abstract of the document. 8. A storage medium storing a document abstract creation program for creating an abstract of a document, a computer, a user intention holding means for holding the input user intention, and extracting an important keyword from the user intention. An important keyword extracting means for extracting an important sentence including the important keyword from the document; an abstract creating means for creating an abstract of the document based on the important sentence; A storage medium that stores an abstract creation program.