JP2001325273A - Important sentence extraction device, method therefor and storage medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To appropriately specify an important sentence corresponding to the kind of the structure of a document. SOLUTION: In a document structure information importance table 3, the 'conditions' of the presence/absence of a document structure expression and the number of the paragraphs in the document and 'important parts' in the document are made to correspond to each other for registration. The document structure expression is an expression for clearly indicating the document structure and is a subtitle or the title of a chapter or a passage, for instance. A document analysis part 2 analyzes an input document, examines to which 'condition' of the document structure information importance table 3 it is pertinent and specifies a part pertinent to the 'important part' corresponding to the pertinent 'condition' as the important part of the input document.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an important sentence extracting apparatus and method capable of specifying an important part of a target document, a method thereof, and a storage medium.

[0002]

2. Description of the Related Art In recent years, with the spread of the Internet and the like, it has become necessary to handle a large amount of document data. Therefore, techniques for extracting important points in a document have been studied. Such techniques include those that specify important points in a document based on the frequency and expression of words, and those that determine important points in a document based on the structure of the document. In the latter case, in particular, since the structure differs depending on the document, the criteria for determining whether the document is important depends on the type of document to be processed.

On the other hand, JP-A-7-13967 discloses that
An “abstract sentence creation device” is disclosed, which has a scenario item for each field of a document, has a predicate table for each scenario part, and extracts important sentences of the document. In this method, important sentences corresponding to the field of the target document are extracted, and those sentences are formed to form an abstract sentence.

[0004]

However, the above-mentioned Japanese Patent Application Laid-Open No.
The technique disclosed in JP-A-13967 has the following disadvantages.

First, there is a problem that an important sentence cannot be obtained if the expression does not match the predicate expression.

Second, since attention is paid only to the predicate expression and no other linguistic information is used, an important sentence may not be extracted properly.

Third, there is a problem that the amount of important sentences to be extracted cannot be changed.

An object of the present invention is to appropriately specify an important sentence according to the type of the structure of a document.

Another object of the present invention is to appropriately specify an important sentence according to a document type.

Another object of the present invention is to identify, without dropping, a potentially important part of a document.

Another object of the present invention is to make it possible to specify an important part in a document in multiple stages according to the degree of importance.

Another object of the present invention is to specify an important part in a document so that the document can be read more efficiently.

Another object of the present invention is to specify a desired portion when additionally specifying an important portion or increasing the number of specified portions.

Another object of the present invention is to enable only a portion having higher importance to be specified in a document.

[0015]

According to a first aspect of the present invention, there is provided a storage device for storing a table in which a type of a document structure is described in association with the importance of a part in a document having the structure. An important sentence extracting apparatus comprising: means; and an important part specifying means for specifying an important part in a target document by referring to the table.

Therefore, an important sentence can be appropriately extracted according to the type of the structure of the document.

According to a second aspect of the present invention, in the important sentence extracting apparatus according to the first aspect, the table expresses the type of the document structure by using the presence or absence of an expression that specifies the document structure.

Therefore, it is possible to specify the type of the structure of the document based on the presence or absence of the expression that clearly indicates the document structure, and to appropriately extract important sentences.

According to a third aspect of the present invention, in the important sentence extracting apparatus according to the first or second aspect, the table includes:
The type of the document structure is expressed using the length of the document.

Therefore, the type of the structure of the document can be specified by the length of the document, and the important sentence can be appropriately extracted.

According to a fourth aspect of the present invention, there is provided a storage means for storing a table in which a type of a document is described in association with the importance of a part in the document of the type, and the table is referred to. Important section specifying means for specifying important points in a target document.

Therefore, important sentences can be appropriately extracted according to the document type.

According to a fifth aspect of the present invention, in the important sentence extracting apparatus according to the fourth aspect, the table describes a requirement of a non-important portion in the document, and the important portion specifying means includes: Referring to the table, a portion other than the portion corresponding to the requirement of the non-important portion in the target document is specified as the important portion.

Therefore, by defining unimportant parts in the document in the table, it is possible to extract other parts without leaking, and it is possible to extract parts that may be important without dropping them. it can.

According to a sixth aspect of the present invention, in the important sentence extracting apparatus according to the fourth aspect, the table describes requirements of important places in the document, and a magnitude of importance of each requirement. The important part specifying means refers to the table and attaches the magnitude of the importance to specify the important part in a target document.

Therefore, an important part in a document can be specified in multiple stages according to the degree of importance.

According to a seventh aspect of the present invention, in the important sentence extracting apparatus according to the fourth aspect, the table describes requirements of important places in the document and associates the requirements with each other. The important point specifying means specifies the important point by referring to the table and performing association between the important points based on the association in a target document.

Therefore, in addition to the part specified as an important part in the target document, a part related thereto can be specified as necessary, and the document can be read efficiently.

According to an eighth aspect of the present invention, in the important sentence extracting apparatus according to the seventh aspect, the table describes a relation name of the association, and the important location specifying means performs the association. When specifying the important part, the relation name is also associated.

Therefore, when an important part is additionally specified or the number of specified parts is increased, a desired part can be specified by using the relation name as a clue.

According to a ninth aspect of the present invention, in the important sentence extracting apparatus according to the sixth aspect, the table describes a condition relating to a description of a location where the requirement is satisfied in association with the requirement. The important location specifying means refers to the table and specifies a location having the requirement with the relatively high importance in the target document as the important location, and further determines the importance level from the identified important location. Is searched for a portion that satisfies the condition that is relatively large, and a portion that satisfies the requirement where the importance is relatively small in the target document and where the matching portion appears is also specified as the important portion.

Therefore, by analyzing the expression of a portion having a relatively high importance in the document and determining the importance of the other parts based on the analysis, it is possible to specify only the higher importance portion in the document. it can.

The invention according to claim 10 refers to a table stored in a predetermined storage device and describing the type of document structure and the importance of a part in a document having the structure in association with each other. The important sentence extraction method includes an important part specifying step of specifying an important part in a target document.

Therefore, an important sentence can be appropriately extracted according to the type of the structure of the document.

According to an eleventh aspect of the present invention, in the important sentence extracting method according to the tenth aspect, the table expresses the type of the document structure by using the presence or absence of an expression that specifies the document structure.

Therefore, it is possible to specify the type of the structure of the document based on the presence or absence of the expression that explicitly specifies the document structure, and to appropriately extract important sentences.

According to a twelfth aspect of the present invention, in the important sentence extracting method according to the tenth or eleventh aspect, the table expresses the type of the document structure using a document length.

Therefore, it is possible to specify the type of the structure of the document based on the length of the document, and to appropriately extract important sentences.

The invention according to claim 13 refers to a table stored in a predetermined storage device and describing the type of a document in association with the importance of a part in the document of the type. The important sentence extraction method includes an important part specifying step of specifying an important part in a target document.

Therefore, important sentences can be appropriately extracted according to the document type.

According to a fourteenth aspect of the present invention, in the important sentence extracting method according to the thirteenth aspect, the table describes a requirement of a non-important part in the document, and the important part specifying step includes the step of: Referring to the table, a portion other than the portion corresponding to the requirement of the non-important portion in the target document is specified as the important portion.

Therefore, by defining insignificant portions in the document in the table, it is possible to extract other portions without leaking, and to extract portions that may be important without dropping them. it can.

According to a fifteenth aspect of the present invention, in the important sentence extracting method according to the thirteenth aspect, the table describes requirements of important parts in the document, and the magnitude of importance of each requirement is described. Is described, the important location specifying step,
The important portion is specified in the target document by referring to the table and assigning the magnitude of the importance.

Therefore, an important part in a document can be specified in multiple stages according to the degree of importance.

According to a sixteenth aspect of the present invention, in the important sentence extracting method according to the thirteenth aspect, the table describes requirements of important places in the document and associates the requirements with each other. In the important part specifying step, the important part is specified by referring to the table and performing association between the important parts based on the association in a target document.

Therefore, in addition to the part specified as an important part in the target document, a related part can be specified as necessary, and the document can be read efficiently.

According to a seventeenth aspect of the present invention, in the important sentence extracting method according to the sixteenth aspect, the table describes a relation name of the association, and the important part specifying step performs the association. When specifying the important part, the relation name is also associated.

Therefore, when an important part is additionally specified or the number of specified parts is increased, a desired part can be specified by using the relation name as a clue.

According to an eighteenth aspect of the present invention, in the important sentence extracting method according to the fifteenth aspect, the table describes a condition relating to a description of a location having the requirement in association with the requirement. The important location specifying step refers to the table, specifies a location having the requirement where the importance is relatively large in the target document as the important location, and further determines the importance level from the identified important location. Is searched for a portion that satisfies the condition that is relatively large, and a portion that satisfies the requirement where the importance is relatively small in the target document and where the matching portion appears is also specified as the important portion.

Therefore, by analyzing the expression of a portion having relatively high importance in the document and judging the importance of the other portions, it is possible to specify only the portion having higher importance in the document. it can.

The invention according to claim 19 refers to a table stored in a predetermined storage device and describing the type of document structure and the importance of a part in a document having the structure in association with each other. And a computer-readable storage medium storing a program for causing a computer to execute an important portion specifying step of specifying an important portion in a target document.

Therefore, an important sentence can be appropriately extracted according to the type of the structure of the document.

According to a twentieth aspect of the present invention, in the storage medium according to the nineteenth aspect, a table that expresses the type of the document structure by using the presence or absence of an expression that explicitly indicates a document structure is used as the table. .

Therefore, the type of the document structure can be specified based on the presence or absence of the expression that clearly indicates the document structure, and the important sentence can be appropriately extracted.

According to a twenty-first aspect of the present invention, in the important sentence extracting method according to the nineteenth or twentieth aspect, a table expressing a type of the document structure using a document length is used as the table. .

Therefore, the type of the structure of the document can be specified by the length of the document, and the important sentence can be appropriately extracted.

The invention according to claim 22 refers to a table stored in a predetermined storage device and describing the type of a document in association with the presence or absence of importance of a part in the document of the type. The computer-readable storage medium stores a program for causing a computer to execute an important portion specifying step of specifying an important portion in a target document.

Therefore, important sentences can be appropriately extracted according to the document type.

According to a twenty-third aspect of the present invention, in the storage medium according to the twenty-second aspect, a table that describes requirements of an insignificant point in a document is used as the table, and the important point specifying step includes: With reference to the table, a portion other than the portion corresponding to the requirement of the non-important portion in the target document is specified as the important portion.

Therefore, by defining unimportant parts in the document in the table, it is possible to extract other parts without leaking, and it is possible to extract parts which may be important without dropping them. it can.

According to a twenty-fourth aspect of the present invention, in the storage medium according to the twenty-second aspect, the table describes requirements of important locations in a document and describes the magnitude of importance of each requirement. In the important portion specifying step, the important portion is specified in the target document by attaching the magnitude of the importance with reference to the table.

Therefore, important parts in a document can be specified in multiple stages according to the degree of importance.

According to a twenty-fifth aspect of the present invention, in the storage medium according to the twenty-second aspect, the table describes a requirement of an important part in a document and associates the requirement with the requirement. In the important part specifying step, the important parts are specified by referring to the table and performing association between the important parts based on the association in a target document.

Therefore, in addition to the part specified as an important part in the target document, a part related thereto can be specified as necessary, and the document can be read efficiently.

According to a twenty-sixth aspect of the present invention, in the storage medium according to the twenty-fifth aspect, the table describing the relation name of the association is used as the table. When performing the identification of the important part, the relation name is also associated.

Therefore, when an important part is additionally specified or the number of specified parts is increased, a desired part can be specified using the relation name as a clue.

According to a twenty-seventh aspect of the present invention, in the storage medium according to the twenty-fourth aspect, the table, which corresponds to the requirement and describes a condition relating to a description of a location having the requirement, is used as the table. The important point specifying step includes: referring to the table, specifying a part having the requirement having the relatively high importance in the target document as the important part, and determining the important part from the specified important part. Look for a part that meets the above condition with a relatively large degree,
In the target document, a part where the matching part appears in the document having the requirement with the relatively low importance is also specified as the important part.

Therefore, by analyzing the expression of a portion having relatively high importance in the document and judging the importance of the other portions, it is possible to specify only the portion having higher importance in the document. it can.

[0069]

DESCRIPTION OF THE PREFERRED EMBODIMENTS [First Embodiment of the Invention] One embodiment of the present invention will be described.

FIG. 1 is a functional block diagram of an important sentence extracting apparatus according to an embodiment of the present invention. As shown in FIG. 1, the important sentence extracting apparatus 1 includes a document analyzing unit 2, a document structure information importance table 3, and an extracting unit 4.

The document analysis unit 2 realizes an important place specifying means, and performs necessary analysis on an input document based on rules registered in the document structure information importance table 3. That is, words and parts of speech related to the structure and importance of the document are recognized by morphological analysis or character string pattern.

In the document structure information importance table 3, rules describing the structure of the document and its importance are registered. That is, it describes the conditions for the document structure,
In the case of a document that satisfies the condition, it is a table that describes an important part of which part is determined to be important. In this example, an important part is specified based on whether or not the target document has a clear expression representing the document structure.

The extraction unit 4 extracts important sentences from the input document based on the conditions registered in the document structure information importance table 3.

Next, a specific configuration of the important sentence extracting apparatus 1 that realizes the above functions will be described. FIG.
FIG. 2 is a block diagram showing an electrical connection of the important sentence extraction device 1. As shown in FIG. 2, the important sentence extracting apparatus 1 performs various operations to centrally control the entire important sentence extracting apparatus 1, and a ROM 12 storing a BIOS and the like.
And a RAM 13 serving as a work area of the CPU 11 are connected by a bus line 14. In addition, bus line 14
Has a magnetic disk device 15 such as a hard disk,
A communication control device 16 that connects the important sentence extraction device 1 to a network such as the Internet, a storage medium reading device 18 that reads a storage medium 17, an input device 19 such as a keyboard and a mouse, and an output device 20 such as an LCD include: They are connected via an interface (not shown).

As the storage medium 17, various types of media such as an optical disk such as a DVD and a CD, a magneto-optical disk, and a floppy (registered trademark) disk can be used. The storage medium 17 stores a program for realizing the functions of the document analysis unit 2 and the extraction unit 4 and fixed data as the contents of the document structure information importance table 3. The storage medium reading device 18 specifically uses an optical disk drive device, a magneto-optical disk drive device, a floppy disk drive device, or the like corresponding to the method of the storage medium 17, and stores the program stored in the storage medium 17. The fixed data is read by the CPU 11 via the storage medium reading device 18 and installed in the magnetic disk device 15, so that the important sentence extracting device 1 realizes functions such as the document analyzing unit 2 and the extracting unit 4. It becomes possible to do this. That is, the magnetic disk device 15 and the RAM 13 serve as storage means and a storage device for storing the document structure information importance table 3 which is a table.

Instead of installing the program and fixed data from the storage medium 17 to the magnetic disk device 15, the program and the fixed data are downloaded from a network such as the Internet via the communication control device 16 and installed on the magnetic disk device 15. It may be. In this case, a storage device that stores the program and the fixed data in the server device that has transmitted the program and the fixed data is also a storage medium of the present invention.

The program installed in the magnetic disk device 15 may be an application program that runs on a predetermined OS. In addition, this program may be a program that substitutes an OS or another application program for executing some of the functions performed by the document analysis unit 2 and the extraction unit 4 described below.

Next, a specific process performed by the important sentence extracting apparatus 1 will be described. First, an example of an input document to be input to the important sentence extraction device 1 will be described. The following example is a newspaper article (Mainichi Shimbun (1995)), and the title is "Northern Japan, stormy due to low pressure-transportation, disturbed".

<Example of input document> ----------------------------------------- -"The effects of the developed low pressure caused severe snowstorms and strong winds on the four days of Hokkaido and the Tohoku region due to the effects of the developed low pressure. As a result, traffic such as closing the highway and stopping the operation of the Tohoku Shinkansen was suspended. Institutional disruptions affected the U-turn rush to the Tokyo metropolitan area at the end of the New Year holidays, according to the Japan Meteorological Agency, where strong cold air below minus 36 degrees flowed over the northern part of the Japanese archipelago. A maximum wind speed of 15-200 meters on land and 20-25 meters on capes and sea was recorded.This strong wind was accompanied by snow and a snowstorm.The bad weather is likely to continue until the morning of the 5th. A strong wind of more than 30 meters was observed between Fukushima and Shiroishi Zao (Shiraishi City, Miyagi Prefecture) on the same day at 5:50 pm on the same day. Thirty-seven trains on this line were delayed up to two hours and six minutes, and some conventional lines in the Tohoku region were closed. According to the Japan Road Traffic Information Center, as of the morning of March 5, the Aomori Interchange and Sendai Izumi Interchange were closed off in Tohoku Expressway and intermittently in Fukushima Prefecture. The road is temporary,
The road was closed. ------------------------------------------

When there is such an input document, first, the document analysis unit 2 performs necessary analysis with reference to the document structure information importance table 3 shown in FIG.

More specifically, the document structure information importance table 3 has registered therein a condition as to whether or not there is a document structure expression. The document structure expression is an expression that specifies the document structure, and corresponds to, for example, a subheading or a chapter or section heading.

The subheadings are as follows:-There is indentation, and a record whose noun is at the end is a subheading. -A record with a symbol at the beginning is a subheading. And so on. Here, whether or not a word is a noun may be determined by performing morphological analysis on a record with indentation. The paragraph of the document can be determined by whether or not there is indentation.

FIG. 4 is a flowchart showing the processing performed by the document analysis unit 2. This process implements an important location specifying step. As shown in FIG. 4, first, the document analysis unit 2 determines whether there is an unprocessed sentence or record in the input document (step S1). If there is an unprocessed sentence or record (step S1) Y), one sentence or one record is specified from the head of the unprocessed sentence or record (step S2), and it is determined whether or not the one sentence or one record has a document structure expression (step S3). The result of the determination is recorded in a predetermined area of the RAM 13 (step S
4) Return to step S1. After making such a determination for all sentences and records (N in step S1),
From the result of the presence / absence of the document structure expression recorded in the predetermined area of the RAM 13 as described above, the presence / absence of the document structure expression is determined for the entire input document, and the number of paragraphs of the entire input document is determined (step). S5). Then, referring to the document structure information importance table 3, an important part corresponding to the condition that matches the determination result is specified in the input document (step S6).

In the input document of the above example, it is understood that there is no document structure expression such as a subheading and the number of paragraphs is four. Then, by referring to the document structure information importance table 3 and examining important points that match the condition obtained in the previous analysis result, under the condition of “no document structure expression” and “5 or less paragraphs”, Since "extract the first paragraph" is described as an important part, only the first paragraph is specified as an important document from the input document in this example.

The extraction section 4 extracts important points in the input document specified in step S 6 and outputs the extracted important points to the output device 20.

In the input document example described above, important portions in the input document extracted in this manner are as follows.

<Example of extracted document> ----------------------------------------- -"The effects of the developed low pressure caused severe snowstorms and strong winds on the four days of Hokkaido and the Tohoku region due to the effects of the developed low pressure. As a result, traffic such as closing the highway and stopping the operation of the Tohoku Shinkansen was suspended. The institution was disrupted, and the U-turn rush to the Tokyo metropolitan area after the New Year holidays was affected. "--------------------------- ---------------

According to the important sentence extracting apparatus 1 described above,
Depending on the type of the structure of the document, an important sentence can be appropriately extracted by using factors (number of paragraphs, etc.) characterizing the structure.

In the above example, it is described in the document structure information importance table 3 that the document does not have a document structure expression and has a paragraph number of 5 or less and is relatively short. In many cases, the document is a document that seeks immediacy such as a flash report, and extracts an important sentence based on the idea that the beginning of the document is often suitable for almost the entire summary. In addition, when there is a subheading of a paragraph (with a document structure expression), the document is often longer than the above, and extraction can be performed in accordance with the document.

[Second Embodiment of the Invention] Another embodiment of the invention will be described as a second embodiment of the invention.

The important sentence extracting device 1 according to the second embodiment of the present invention is different from the first embodiment in that the document structure information importance table 3 shown in FIG. 3 is replaced with the document structure shown in FIG. Using the information importance table 3, FIG.
6 in that the processing shown in FIG. 6 is performed instead of the processing shown in FIG. The other contents are the same as those of the first embodiment of the present invention.
Detailed description is omitted.

Document structure information importance table 3 shown in FIG.
Is a description that associates a document type with a component (a group with the document structure expression at the top) that is an important point in the document type so that important points can be specified based on the type of the input document. It is.

Next, a specific process performed by the important sentence extracting apparatus 1 will be described. First, an example of an input document to be input to the important sentence extraction device 1 will be described. The following document is an example of a published patent gazette. (In this specification, when a patent gazette is indicated, for convenience, parentheses in the sentence are replaced with "<
>").

<Example of input document> ----------------------------------------- --- `` <Title of Invention> Text Summarization Support Device <Claims><Claim1> A display unit that displays digitized text, and characters and phrases in the text displayed on the display unit Contains the input part for the user to input the position specification to specify the character and the phrase specified by the position specification input from the user
A sentence cutout unit that cuts out one sentence, a storage unit, an analysis dictionary, and analyzes the structure of the cutout sentence using the analysis dictionary, including a specified character or phrase, and focusing on one predicate. When a unit sentence detecting unit that takes out a unit sentence that is a unitary unit and stores the unit sentence in the storage unit, and the processing by the sentence cutout unit and the unit sentence detecting unit ends for all position designations input by a user, A sentence summarization support device comprising: an output unit that outputs all unit sentences stored in the storage unit as summary sentence creation data that is a source of the sentence summary. BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a sentence summarization support device that assists a user when summarizing an input sentence. <0002><Priorart> When a human summarizes a sentence, he takes the steps of marking important points while reading the sentence, extracting the marked part, and assembling a summary sentence based on it. It is common. <Problems to be Solved by the Invention> In the conventional concept of automatic summarization, the longer a sentence is, the higher the probability of including a keyword is, and the ratio of extracting a sentence containing many modifiers is relatively high. Another problem is that sentences containing only one keyword are difficult to extract. For this reason, automatic summarization is actually only a device for extracting basic data of summaries, and human
It is indispensable to confirm the extracted sentences, relieve omission of extraction, and process the extracted sentences. [0006] The present invention has been made in view of such circumstances, and an object of the present invention is to provide a text summarization support apparatus that reduces the burden on a user who summarizes text. <Means for Solving the Problem> In order to achieve the above object, a sentence summarization support device of the present invention includes a display unit for displaying digitized sentences, and a sentence displayed on the display unit. An input unit for the user to input a position specification that specifies a character or phrase in the inside, a sentence cutout unit that cuts out one sentence including the character or phrase specified by the position specification input from the user,
A storage unit, an analysis dictionary, and the structure of the cut-out sentence are analyzed using the analysis dictionary, and a specified sentence, including a specified character or phrase, is extracted as a unit sentence that is a united unit centered on one predicate. When the unit sentence detection unit stored in the storage unit and the processing by the sentence cutout unit and the unit sentence detection unit for all position designations input by the user are completed, all of the units stored in the storage unit And an output unit for outputting the unit sentence of the sentence as summary sentence creation data serving as a basis for summarizing the sentence. <Operation> According to the present invention, a user specifies (omitted) in a sentence in order to create an abstract sentence. <Example> Next, an example of the present invention will be described with reference to the drawings. (Omitted) <Effect of the Invention> As described above, the present invention displays a digitized text on a display or the like, and includes a character or a phrase specified in the text when the position is specified by the user. A sentence is cut out, the structure of the sentence is analyzed by natural language analysis, a unit sentence including a specified character or phrase is taken out and stored, and is stored when a series of position designations by the user are completed. By outputting the unit sentence as data for creating a summary sentence, there is an effect that the user can easily create a summary sentence from the output unit sentence. <Brief Description of the Drawings><FIG.1> FIG. 1 is a block diagram illustrating a configuration of a text summarization support device according to an embodiment of the present invention. <FIG. 2> A flowchart showing the processing of the sentence segmentation unit 3. <FIG. 3> A flowchart showing the processing of the unit sentence detection unit 5. FIG. 4 is a diagram illustrating an example of an analysis result of natural language analysis performed by the unit sentence detection unit 5. FIG. 5 is a diagram showing an example of a range of a unit sentence obtained by natural language analysis by the unit sentence detection unit 5. --------------------------------------------

When the above document is input, the type of the document is specified. The specification may be performed by the user directly inputting the type of the document.
May analyze the document structure expression in the input document and automatically specify the document structure expression based on the analysis result. In the example of the input document described above, the document type in the document structure information importance table 3 shown in FIG. 5 corresponds to “patent”. However, the user may directly input the document type “patent”. Further, the document analysis unit 2 extracts a document structure expression such as “claims” or “conventional technology” defined corresponding to the document type “patent” in the document structure information importance table 3. Thus, it may be determined that the document type is “patent”.

The specific processing of the document analysis unit 2 when the document analysis unit 2 automatically specifies important points in the input document is as follows.
This is as shown in the flowchart of FIG. This process implements an important location specifying step. That is, first,
The document analysis unit 2 determines whether there is an unprocessed statement or record in the input document (step S11). If there is an unprocessed statement or record (Y in step S11),
One sentence or one record is specified from the beginning of the unprocessed sentence or record (step S12), and the one sentence or one
Document structure expressions unique to patent specifications such as "Title of Invention" and "Prior Art" listed as important places corresponding to the document type "Patent" in the document structure information importance table 3 in the record It is determined whether or not there is (Step S13), the determination result is recorded in a predetermined area of the RAM 13 (Step S14), and the process returns to Step S11. After making such a determination for all sentences and records (N in step S11), the type of the input document is specified from the result of the presence / absence of the document structure expression recorded in the predetermined area of the RAM 13 as described above. (Step S15) Referring to the document structure information importance table 3, an important part in the input document corresponding to the document type is specified (Step S16).

That is, in the example of the document structure information importance table 3 shown in FIG. 5, a document structure expression unique to a patent specification such as "name of invention" or "conventional technology" is included in the input document. Sometimes we identify the type of input document as "patent"
When such a document structure expression is not included, the type of the input document is specified as, for example, “column”. Then, important points registered corresponding to each document type are specified in the input document.

[0098] Based on such a result of the document analysis unit 2, the extraction unit 4 extracts an important part in the specified input document from the input document, and outputs the result to the output device 20. In the example of FIG. 5, if the document type is "patent",
The texts in the columns of “Title of Invention”, “Claims”, “Prior Art”, and “Problem to be Solved by the Invention” are extracted.

The following is an extraction example of the example of the input document.

<Example of extracted document> ----------------------------------------- ------ `` <Title of Invention> Text Summarization Support Device <Claims><Claim1> A display unit that displays digitized text, and a text in the text displayed on the display unit Includes an input part for the user to input the position specification to specify characters and phrases, and the characters and phrases specified by the position specification entered by the user
A sentence cutout unit that cuts out one sentence, a storage unit, an analysis dictionary, and analyzes the structure of the cutout sentence using the analysis dictionary, including a specified character or phrase, and focusing on one predicate. When a unit sentence detecting unit that takes out a unit sentence that is a unitary unit and stores the unit sentence in the storage unit, and the processing by the sentence cutout unit and the unit sentence detecting unit ends for all position designations input by a user, A sentence summarization support device comprising: an output unit that outputs all unit sentences stored in the storage unit as summary sentence creation data that is a source of the sentence summary. <Conventional technology> When a human summarizes a sentence, it is common practice to mark important points while reading the sentence, then extract the marked part, and assemble a summary based on that. It is a target. <Problems to be Solved by the Invention> In the conventional concept of automatic summarization, the longer a sentence is, the higher the probability of including a keyword is, and the ratio of extracting a sentence containing many modifiers is relatively high. Another problem is that sentences containing only one keyword are difficult to extract. For this reason, automatic summarization is actually only a device for extracting the basic data of summarization, and humans need to confirm extracted sentences, rescue omissions, and extract Sentence processing is essential.
[0006] The present invention has been made in view of such circumstances, and an object of the present invention is to provide a text summarization support apparatus that reduces the burden on a user who summarizes text. -------------------------------------------

If the target document is a column, as in the example described in the document structure information importance table 3 in FIG. And only the subheadings are extracted as important parts.

The following is an example of an input document as a column, which is an article of the Mainichi Shimbun (1995). The title is "Sugar is" sweet "and the reality is not sweet?

<Example of input document> ----------------------------------------- -"Sugar consumption is declining every year. Drinks and sweets are all about" low-sweetness, "but the industry is working hard to expand its consumption by promoting the utility of sugar.消費 Consumption decreased at the peak of FY73--Industry and PR sprung up According to the Ministry of Agriculture and Fisheries, per capita consumption of sugar was the highest in the 197kg sugar year (October-September of the following year). Since then, it has been declining, falling below 25 km in FY1990 and below 20 km since FY193. In addition, the technology for mass production of isomerized sugars using starch as a raw material has been established, and has been widely used for soft drinks, cans, confections, breads and the like for more than ten years. Isomerized sugar is not included in sugar in the classification, so it is in the form of consumption of sugar.However, consumption of isomerized sugar also peaked at 28 kg in FY78, Is 25 km, poor. According to the Ministry of Sugars, "Sucrose and sweetness based on incorrect health knowledge that sugar causes obesity, tooth decay and diabetes have become noticeable in the Heisei era. And confectionery manufacturers say, "If you reduce sweetness, it will sell well," and switching production has a major effect. "(Eiji Shinta, Assistant Manager). In response to this downward trend, eight sugar groups, including the Sugar Refining Industry Association and the Japan Sugar Import and Export Council, formed the Sugar “True” Age Council at the end of 1990. Continue campaigning to get more sugar from consumers. Last spring, she produced 30,000 copies of the book "Sweet Story", which introduces delicious Japanese and Western confectionery and artisan skills, based on programs aired on TV two years ago, but says that there is almost no remainder. The second volume will be published this spring. In addition, from February 19 to 25, an art object using sugar will be exhibited at Sony Square in Ginza, Tokyo, and publicized to shoppers.使用 Successful usage and effects << Health >> Sugar industry asks questions about health related consumers, such as "If you eat sugar, you get fat,""Do you not lose tooth decay,""Do you lose vitamin B1 and calcium?" The meeting is (1)
Sugar is an energy source that quickly recovers from fatigue. There is no problem if you eat well-balanced and move your body well. After that, if you brush your teeth, you can prevent tooth decay. (3) B1 and calcium will not be deprived, but it is reasonable to take milk, fruits, vegetables, etc. in a well-balanced manner. "Nine Wonders" Are there two more uses of sugar than "Seven Wonders"? (1) Hydrophilicity: When immersed in meat, it binds protein and moisture and softens meat. (2) Foam stability = whisking in addition to egg white provides a firm foam meringue. (3) Osmoticness = added when making fruit wine to bring out the taste and aroma of fruit by osmotic pressure. (4) Jellification = In addition to making jam, pectin contained in fruits is made into jelly. (5) Prevention of starch aging = Since it has the function of keeping starch soft, it prevents it from becoming hard when added to sushi rice. (6) Promotion of fermentation = Activate yeast fermentation to make plump bread with good baking color. (7) Temperature change = about 150 ° C to syrup, 115 ° C to fondant, 1650 ° C to 165 ° C, caramel at 195 ° C or more. (8) Antioxidation of fat = Preserve the buttery flavor of cakes and cookies. (9) Antiseptic = absorbs water required for the growth of mold and bacteria. (From “Sweet Story”) 《Preservation》 According to the Tokyo Consumer Center, inquiries about sugar and complaints from April 1990 to November last year were relatively low, totaling 28 cases. . Most of the content is about preservation, such as "Old sugar came out, is it OK?" Sugar is easy to absorb odor and moisture, so put it in a can or plastic container that can be sealed. When the smell is transferred, spread it in a well-ventilated place and expose it to the air. If it hardens, it can be spread and sprayed or crushed. << Plate >> The custom of distributing sugar in the shape of Thailand for the celebration has been common since ancient times, but in the Mitsukoshi Ginza store in Tokyo, the name of the forest of Kaminomi = Moto Akasaka, Minato-ku Released a sugar plate that can be inserted. The plate is about the size of the palm of an adult. You can put words like "Congratulations, chan." The price is 3,000 yen with sugar baby shoes. There are several types of seasonal products, such as a heart-shaped accessory case for Valentine's Day (4,000 yen). We also produce original products according to orders. Kind Features / Usage ◆ Kami-shirasu So-called white sugar. Moist flavor.
All-purpose seasoning, confectionery, drinks, etc. ◆ Three-colored sugar. When used for boiled foods and tsukudani, sweetness and richness can be obtained. ◆ Granulated sugar High purity and smooth. Widely used for drinks, confectionery, and cooking with a light sweetness. It has a flavor,
Used for cooking and noodles. Kaku Sugar Granulated sugar is hardened into a square. Convenient for making drinks, cooking and confectionery ◆ Crying sugar Ideal for fruit wine. When you boil Inari sushi, it gives a beautiful luster. Powdered sugar Sprinkled on fruits, cakes and cookies,
Used for making Western confectionery ◆ Granular sugar Porous, granule-like high-purity sugar. Put it in a cold drink or put it on fruit. Crystals are very small and are prized for Japanese sweets ◆ Brown sugar The sugarcane juice was boiled down. Strong flavor and sweetness = from the data of the Sugar Refining Industry Association "----------------------------------- --------

The following is an example in which the document type of the above example of the input document is recognized as “column”, and the important sentence is extracted by referring to the document structure information importance table 3.

<Extraction Example> ------------------------------------------- "Sugar consumption is decreasing year by year. Although drinks and sweets are all about" low-sweetness, "the industry is working hard to expand its consumption by promoting the utility of sugar.消費 Decrease in consumption at the peak of FY73--Industry and PR booming ◇ Good usage and effects 《Health》 《Nine wonder》 《Preservation》 《Plate》 --------------- ----------------------------

According to the important sentence extracting apparatus 1 described above,
Since important points in the document are defined for each document type, an appropriate sentence can be extracted according to the document type.

[Third Embodiment of the Invention] Another embodiment of the invention will be described as a third embodiment of the invention.

The important sentence extracting device 1 according to the third embodiment of the present invention is different from the second embodiment in that the document structure information importance table 3 shown in FIG. 5 is replaced with the document structure shown in FIG. Using the information importance table 3, FIG.
8 in that the processing shown in FIG. 8 is performed instead of the processing shown in FIG. The other contents are the same as those of the second embodiment of the present invention, so that the same reference numerals as those of the second embodiment of the present invention are used.
Detailed description is omitted.

The important sentence extracting apparatus 1 uses the document structure information importance table 3 shown in FIG. 7 such that a document type and its document can be identified based on the type of the input document. It is described in association with constituent elements that are not important parts that need not be extracted from the type (groups starting with the document structure expression).

FIG. 8 is a flowchart illustrating a specific process performed by the document analysis unit 2. This process implements an important location specifying step. As shown in FIG.
First, in this example, the user directly inputs the type of the input document. When this input is made (Y in step S21), the document analysis unit 2 determines whether there is an unprocessed sentence or record in the input document (step S22), and there is an unprocessed sentence or record. At this time (Y in step S22), one sentence or one record is specified from the head of the unprocessed sentence or record (step S23), and the sentence or record is added to the sentence or record in step S21 of the document structure information importance table 3 in step S21. An analysis necessary to determine the presence or absence of contents listed as non-important parts corresponding to the specified document type is performed (step S24), and the determination result is recorded in a predetermined area of the RAM 13 (step S24). S25),
It returns to step S22. After performing such an analysis for all sentences and records (N in step S22),
Based on the analysis result recorded in the predetermined area of the RAM 13 as described above, a non-important portion in the input document is specified (step S26).

More specifically, when the document type is a column, in the document structure information importance table 3 shown in FIG. 7, the description of the column, ie, "itemization and enumeration of the hierarchy below the subheading" is not shown. It is an important place. Therefore, the document analysis unit 2 analyzes and determines “subheadline”, “item list”, and “enumeration expression”.

The itemized list and the enumerated expression include, for example, a part where the beginning is a symbolic expression or a numerical expression. A part where the beginning is a symbol and the end of the record is a series of consecutive noun records.

In the analysis for each sentence (step S24),
The expression of the above condition is detected and stored in the RAM 13,
After analyzing all sentences (N in step S22), a continuous part is determined as an item, and if not, it is determined as a subheading (step S26).

Then, the extracting unit 4 extracts only non-important locations from the description of the non-important locations stored in the RAM 13 as important locations.

In the example of the input document shown in the second embodiment of the present invention, in the document structure information importance table 3 shown in FIG.
As a result of performing the above-described processing by selecting the column as the document type, the following documents are extracted.

<Example of extracted document> ----------------------------------------- --- "Sugar consumption is decreasing year by year. Drinks and sweets are all about" low-sweetness, "but the industry is working hard to expand its consumption by promoting the utility of sugar.消費 Consumption decreased at the peak of FY73--Industry and PR sprung up According to the Ministry of Agriculture and Fisheries, per capita consumption of sugar was the highest in the 197kg sugar year (October-September of the following year). Since then, it has been declining, falling below 25 km in FY1990 and below 20 km since FY193. In addition, the technology for mass production of isomerized sugars using starch as a raw material has been established, and has been widely used for soft drinks, cans, confections, breads and the like for more than ten years. Isomerized sugar is not included in sugar in the classification, so it is in the form of consumption of sugar.However, consumption of isomerized sugar also peaked at 28 kg in FY78, Is 25 km, poor. According to the Ministry of Sugars, "Sucrose and sweetness based on incorrect health knowledge that sugar causes obesity, tooth decay and diabetes have become noticeable in the Heisei era. And confectionery manufacturers say, "If you reduce sweetness, it will sell well," and switching production has a major effect. "(Eiji Shinta, Assistant Manager). In response to this downward trend, eight sugar groups, including the Sugar Refining Industry Association and the Japan Sugar Import and Export Council, formed the Sugar “True” Age Council at the end of 1990. Continue campaigning to get more sugar from consumers. Last spring, she produced 30,000 copies of the book "Sweet Story", which introduces delicious Japanese and Western confectionery and artisan skills, based on programs aired on TV two years ago, but says that there is almost no remainder. The second volume will be published this spring. In addition, from February 19 to 25, an art object using sugar will be exhibited at Sony Square in Ginza, Tokyo, and publicized to shoppers. ◇ Good use and effects 《Health》 “If you eat sugar, you will get fat”
Consumers concerned with health, such as "Can vitamin B1 and calcium be deprived?" The Sugar Refining Manufacturers Association (1) sugar is an energy source that quickly recovers from fatigue. There is no problem if you eat well-balanced and move your body well. After that, if you brush your teeth, you can prevent tooth decay. (3) B1 and calcium will not be deprived, but it is reasonable to take milk, fruits, vegetables, etc. in a well-balanced manner. "Nine Wonders" Are there two more uses of sugar than "Seven Wonders"? << Conservation >> According to the Tokyo Consumer Center, there have been relatively few inquiries and complaints about sugar since April 1990, up to November last year, a total of 28 cases. Most of the content is about preservation, such as "Old sugar came out, is it OK?" Sugar is easy to absorb odor and moisture, so put it in a can or plastic container that can be sealed. When the smell is transferred, spread it in a well-ventilated place and expose it to the air. If it hardens, it can be spread and sprayed or crushed. << Plate >> The custom of distributing sugar in the shape of Thailand for the celebration has been common since ancient times, but in the Mitsukoshi Ginza store in Tokyo, the name of the forest of Kaminomi = Moto Akasaka, Minato-ku Released a sugar plate that can be inserted. The plate is about the size of the palm of an adult. You can put words like "Congratulations, chan." The price is 3,000 yen with sugar baby shoes. There are several types of seasonal products, such as a heart-shaped accessory case for Valentine's Day (4,000 yen). We also produce original products according to orders. ------------------------------------------

According to the important sentence extracting apparatus 1 described above,
By defining unimportant parts in the document for each document type in the document structure information importance table 3, it is possible to accurately extract important sentences according to the type of the document.

Further, by defining unimportant portions in the document in the document structure information importance table 3, it is possible to extract other portions without leaking, and to remove portions that may be important. It can be extracted without.

[Fourth Embodiment of the Invention] Another embodiment of the invention will be described as a fourth embodiment of the invention.

The important sentence extracting apparatus 1 according to the fourth embodiment of the present invention is different from the third embodiment in that the document structure information importance table 3 shown in FIG. 7 is replaced with the document structure shown in FIG. Using the information importance table 3, FIG.
Is that the processing shown in FIG. 10 is performed instead of the processing shown in FIG.
Other details are the same as those of the third embodiment of the present invention, and therefore, the same reference numerals as those of the third embodiment of the present invention are used, and the detailed description is omitted.

In the important sentence extracting apparatus 1, in the document structure information importance table 3 shown in FIG. 9, based on the type of the input document, the components of the document (the unit having the document structure expression at the head) and Describes the degree and order of importance of the location.

The specific processing of the document analysis unit 2 when the document analysis unit 2 automatically specifies important points in the input document is as follows.
This is as shown in the flowchart of FIG. This process implements an important location specifying step. That is, first, the document analysis unit 2 determines whether there is an unprocessed sentence or record in the input document (step 31). If there is an unprocessed sentence or record (Y in step S31), the process proceeds to step S31. One sentence or one record is specified from the beginning of the processing sentence or record (step S32), and the condition corresponding to the document type “patent” in the document structure information importance table 3 for the sentence or record is specified. It is determined whether or not there is a specific document structure expression in the patent specification such as “name of invention” or “prior art” (step S33), and the determination result is recorded in a predetermined area of the RAM 13. Then (step S34), the process returns to step S31. After making such a determination for all sentences and records (N in step S31), the type of the input document is specified from the result of the presence / absence of the document structure expression recorded in the predetermined area of the RAM 13 as described above. (Step S35), referring to the document structure information importance table 3, specify an important part from conditions corresponding to the document type (Step S35).
36).

Then, the extracting unit 4 extracts the important part and outputs the extracted important part to the output device 20. In the document structure information importance table 3 shown in FIG. The order of the degrees is determined. In this example, the one having the importance level of 1 is set to the maximum importance level. Since the order of importance is determined in this way, when an important part in the input document is specified (step S36), only the output of the output device 20 having the importance of 1 is output first. When the importance 2 is output only when instructed by the user, or when the importance 1 and 2 are displayed by distinguishing font sizes and colors, or when the document size to be output is specified, Various outputs in accordance with the magnitude of importance are possible, such as outputting importance 2 only when there is room in the size of the importance 1.

An example in which the input document example shown in the second embodiment of the present invention is processed by the important sentence extracting apparatus 1 and only the importance 1 is output is shown below.

<Example of extracted document> ----------------------------------------- "Sugar consumption is decreasing year by year. Although drinks and sweets are all about" low-sweetness, "the industry is working hard to expand its consumption by promoting the utility of sugar.消費 Decrease in consumption at the peak of FY73--Industry and PR booming ◇ Good usage and effects 《Health》 《Nine wonder》 《Preservation》 《Plate》 --------------- ----------------------------

According to the important sentence extracting apparatus 1 described above,
Importance is assigned to each component of the document based on the type of the input document, so important sentences can be extracted according to the document type, and extracted in multiple stages according to the user's wishes and output restrictions. It is possible to output a document.

[Fifth Embodiment] Another embodiment of the present invention will be described as a fifth embodiment of the present invention.

The important sentence extracting device 1 according to the fifth embodiment of the present invention is different from the fourth embodiment in that the document structure information importance table 3 shown in FIG. 9 is replaced with the document structure shown in FIG. The point is that the same processing as in the second embodiment of the invention shown in FIG. 6 is performed instead of the processing shown in FIG. 10 using the information importance table 3. The other contents are the same as those of the fourth embodiment of the invention, so that the same reference numerals as those of the fourth embodiment of the invention are used and the detailed description is omitted.

In the document structure information importance table 3 shown in FIG. 11, based on the type of the input document, the document structure expression in the document and the parts related to the constituent elements (the unit having the document structure expression at the head) are related. Is described. In the example of FIG. 11, “Claims” are related in content to “Examples”. In particular, in this example, the component of the document structure expression written on the leftmost side and the component of the document structure expression described on the right side are related.

Such a document structure information importance table 3
, The document analysis unit 2 performs the same processing as in the second embodiment of the invention shown in FIG. In this case, step S13
Determines whether there is a document structure expression listed in the set of document structure expressions corresponding to the document type "patent" in the document structure information importance table 3, and inputs such a document structure expression in step S15. Document type "patent"
And specify. Then, in step S16, a component corresponding to the document structure expression listed in the set of document structure expressions corresponding to the document type "patent" is specified as an important part of the input document. When performing this identification, identify whether it is the component of the document structure expression written at the leftmost in the column of the document structure expression set, or the component of the document structure expression described to the right of it. The components of the same set of document structure expressions are specified in association with each other.

Then, the extracting unit 4 sets the document type "patent"
Of the set of document structure expressions corresponding to (1), the components of the leftmost document structure expression (in the example of FIG. 5, “claims” and “problem to be solved by the invention”) are extracted. Then, the data is output to the output device 20. At this time, the user is instructed about the components related to each component (in the example of FIG. 5, “effect”, “example”, and “action”), or the component of the document structure expression described at the leftmost. If only the amount of extraction is small, the relevant part is output.
For example, when the user designates the “problem to be solved by the invention” part, the related “action” and “effect” parts are output. Note that the associated portion may be an identification display or the like as described in Embodiment 3 of the present invention.

In the example of the input document of the patent publication shown in the second embodiment of the invention, an example in which important portions are extracted by the important sentence extracting device 1 is as follows.

<Example of extracted document> ----------------------------------------- -"<Problems to be Solved by the Invention> In the conventional concept of automatic summarization, the longer a sentence is, the higher the probability of including a keyword is, and the ratio of extracting a sentence containing many modifiers is relatively high. There is also a problem that sentences containing only one keyword are difficult to extract, so even if an automatic summarization is used, it is actually just a device for extracting basic data for summarization, and a final summary sentence is created. It is indispensable for a human to check the extracted text, remedy the omission of extraction, and process the extracted text. <0006> The present invention has been made in view of such circumstances, and its purpose is to summarize the text. It is an object of the present invention to provide a sentence summarization support device that can reduce the burden on a user who performs summarization. , A unit sentence containing a character or phrase specified by the user (a unit of one predicate) is extracted, so that the user can create a summary sentence from the extracted unit sentence more easily than in the past. <Effects of the Invention> As described above, the present invention displays digitized text on a display or the like, and includes a character or phrase specified in the text when the position is specified by the user. A sentence is cut out, the structure of the sentence is analyzed by natural language analysis, a unit sentence including a specified character or phrase is taken out and stored, and is stored when a series of position designations by the user are completed. By outputting the unit sentence as data for creating a summary sentence, there is an effect that a user can easily create a summary sentence from the output unit sentence. "------------ ------------------------------

According to the important sentence extracting apparatus 1 described above,
By associating each component with each document type, it is possible to output related component parts as necessary, and to present important parts and related parts, and to read documents efficiently. You can proceed.

[Sixth Embodiment of the Invention] Another embodiment of the invention will be described as a sixth embodiment of the invention.

The important sentence extracting device 1 according to the sixth embodiment of the present invention is different from the fifth embodiment in that the document structure information importance table 3 shown in FIG. 11 is replaced with the document structure shown in FIG. The point is that the information importance table 3 is used. Other details are the same as those of the fifth embodiment of the present invention, so that the same reference numerals as those of the fifth embodiment of the present invention are used.
Detailed description is omitted.

In the document structure information importance table 3 shown in FIG. 12, a set of document structure expressions is described in the same manner as in the fifth embodiment of the invention shown in FIG. In the set of document structure expressions, a “relation name” indicating the type of the relationship is described in association with the relationship. That is, in the example of FIG. 12, "claims" is related to "embodiment" in that information is "refined", and "claims" is referred to as "operation" and "simplification". Related in.

Such document structure information importance table 3
, The document analysis unit 2 performs the same processing as in the fifth embodiment of the invention shown in FIG. As described above, step S
In step S16, the component corresponding to the document structure expression listed in the set of the document structure expression corresponding to the document type "patent" is specified as an important part of the input document. Identify whether it is a component of the document structure expression written on the leftmost in the column of expression set or a component of the document structure expression described on the right of it, and configure the same set of document structure expressions The elements are specified in association with each other. When this identification is performed, the relation names listed in the document structure information importance table 3 are also associated with the components of the same set of document structure expressions.

Then, the extraction unit 4 regards the unit of the component “claims” described at the leftmost of the document structure information importance table 3 as an important part first, and
Output to 0. At this time, the document structure information importance table 3
It is also possible to output the component name related to the component of the document structure expression described at the leftmost and the related name indicating this relationship, when there is an instruction from the user, or when the amount of important parts extracted is small Can display its associated components.

For example, in the example of FIG. 12, when the user designates the "action" portion of the relation name "simplification" for "claims", the related "action" portion is output to the output device 20.
Output from Note that the associated portion may be an identification display or the like as described in Embodiment 3 of the present invention. This may be set in advance as to which relation name is better for additionally displaying information, depending on the use or the like.

In the example of the input document of the patent publication shown in the second embodiment of the invention, an example in which important portions are extracted by the important sentence extracting device 1 is as follows.

<Example of Extracted Document> ---------------------------------------- "<Claims><Claim1> A display for displaying a digitized sentence, and a position for designating a character or a phrase in the sentence displayed on the display for the user to input. Includes characters and phrases specified by the input part and the position specified by the user
A sentence cutout unit that cuts out one sentence, a storage unit, an analysis dictionary, and analyzes the structure of the cutout sentence using the analysis dictionary, including a specified character or phrase, and focusing on one predicate. When a unit sentence detecting unit that takes out a unit sentence that is a unitary unit and stores the unit sentence in the storage unit, and the processing by the sentence cutout unit and the unit sentence detecting unit ends for all position designations input by a user, A sentence summarization support device comprising: an output unit that outputs all unit sentences stored in the storage unit as summary sentence creation data that is a source of the sentence summary. -> (Simplification) Action <Action> The present invention extracts a unit sentence (a unit of one predicate) containing a character or phrase specified by the user from a sentence in an attempt to create an abstract sentence. Therefore, the user can easily create a summary sentence from the extracted unit sentence as compared with the related art. -----------------------------------------

According to the important sentence extracting apparatus 1 described above,
By retaining related components in the document and their relationships, it is possible to output information desired by the user when information is additionally displayed or the output amount is increased.

In the above example, the “claims”, which are long and complex as sentences, and the “actions” which are simplified from them.
By displaying together with the expression of the portion, more appropriate information presentation can be performed.

[Embodiment 7] Another embodiment of the present invention will be described as a seventh embodiment of the present invention.

The important sentence extracting apparatus 1 according to the seventh embodiment differs from the sixth embodiment in that the document structure information importance table 3 shown in FIG. 12 is replaced with the document structure shown in FIG. The point is that the information importance table 3 is used. Other details are the same as those in the sixth embodiment of the present invention, and thus the same reference numerals as those in the sixth embodiment of the present invention are used.
Detailed description is omitted.

In the document structure information importance table 3 shown in FIG. 13, according to the document type, the “document structure expression” at the head of a component (a group starting from the document structure expression) which is an important part in the document type ", A predetermined" condition "for a phrase included in the component, and" importance "of the component are described in association with each other.

Such a document structure information importance table 3
, The document analysis unit 2 performs the same processing as in the fourth embodiment of the invention shown in FIG. However, in step S36 of FIG. 10, the important part of the input document is specified as follows. That is, first, a component having “document structure expression” listed in the document structure information importance table 3 is specified in the input document. Then, among these constituent elements, syntax analysis, pattern matching, morphological analysis, and the like are performed on the component part having the highest “importance” (importance 1), and a word that matches the corresponding “condition”, that is, , The word in a long qualifier sentence, the last word in a long qualifier sentence,
Or, keep words that are frequently used. In the example of FIG. 13, the last word of the sentence having a long modifier is “condition”.
With respect to this component itself, the whole may be set to the importance level 1, or particularly, only the long sentence of the modifier may be set to the importance level.
Other components with low "importance" (importance 2)
With regard to, if the held phrase appears in the "condition" described in the document structure information importance table 3, the sentence of the component is extracted as an important part.

In the example of the document structure information importance table 3 shown in FIG. 13, for the “claims” having “importance” of 1, a long sentence of a qualifying phrase (how long does it mean? Is set in advance in the document structure information importance table 3), and its phrase is obtained. When the input document is an example of the patent publication disclosed in the second embodiment of the invention, "the position designation for designating a character or a phrase in a sentence displayed on the display unit for the user to input is designated. `` Input unit '', `` Sentence cutout unit that cuts out one sentence containing the character or phrase specified by the position specified by the user '', `` Analyze and specify the structure of the cutout sentence using the analysis dictionary, and specify Including united characters and phrases, a unit sentence detection unit that extracts unit sentences centered on one predicate and stores the unit sentence in the storage unit '', `` for all position designations input by the user, When the processing by the sentence segmentation unit and the unit sentence detection unit is completed, all of the unit sentences stored in the storage unit are output as summary sentence creation data that is the basis for summarizing the sentence. Are so long that the words The first to hold.

If there is a document structure expression “conventional technology” or “embodiment”, for the component corresponding to the “conventional technology” or “embodiment”, the preceding term is described in the document structure information importance table. It is determined whether or not the “condition” listed in 3 has appeared in the form of “proposal“ -ha ””. If there is a sentence that appeared in the form of “proposal“ -ha ””, that sentence Is specified and stored in the RAM 13.

Then, the extraction unit 4 extracts the sentence specified as described above and outputs it from the output device 20.

The following document is an example of the same published patent publication as that described in the second embodiment of the present invention.

<Example of input document> ----------------------------------------- `` <Title of invention> Text summarization support device <Claims><Claim1> Designating a display unit that displays digitized text, and characters and phrases in the text displayed on the display unit Includes an input part for the user to input the position specification, and the characters and phrases specified by the position specification input by the user
A sentence cutout unit that cuts out one sentence, a storage unit, an analysis dictionary, and analyzes the structure of the cutout sentence using the analysis dictionary, including a specified character or phrase, and focusing on one predicate. When a unit sentence detecting unit that takes out a unit sentence that is a unitary unit and stores the unit sentence in the storage unit, and the processing by the sentence cutout unit and the unit sentence detecting unit ends for all position designations input by a user, A sentence summarization support device comprising: an output unit that outputs all unit sentences stored in the storage unit as summary sentence creation data that is a source of the sentence summary. (Omitted) <0012> The storage unit 4 stores sentences to be summarized. The display unit 1 is, for example, a CRT display, and displays a sentence stored in the storage unit 4. The input unit 2 is a mouse, a touch key, etc., and the user looks at the text displayed on the display unit 1 and inputs characters or phrases that are considered important.
The sentence segmentation unit 3 that marks (positions specified) using
As shown in FIG. 2, a boundary between a sentence and a sentence is searched for from the mark toward the beginning of the sentence, the position is set to “S” (step 11), and then the sentence and the sentence are sent from the mark toward the end of the sentence. And find its position as “E” (step 1
2) Finally, cut out the range from S to E as one sentence.
(Step 13). As shown in FIG. 3, the unit sentence detection unit 5 performs a natural language analysis of the sentence cut out by the sentence extraction unit 3 using the analysis dictionary 6 (see Reference: Shin Nagao, "Japanese Information Processing" (electronic communication (Academic Society, 1985)) (Step 21), determine the range of unit sentence that is a unit around one predicate (Step 22), detect the unit sentence including the marked character or phrase, and store it in the storage unit 4. It is stored (step 23). Output unit 7
When the processing by the sentence extraction unit 3 and the unit sentence detection unit 5 is completed for all the marks in the printer, for example, the storage unit
All the unit sentences stored in 4 are output as summary sentence creation data which is the basis of the sentence summary. <0013> Next, the operation of the text summarization support apparatus of the present embodiment will be described.
A case where the following document is the subject of summarization will be described. <0014> “Toin, a leading manufacturer of product instruction manuals, has put into practical use a machine translation system that translates English manuals into Japanese.
Introduced in the creation process. Since the computer automatically handles the English-Japanese machine translation work, the delivery time is reduced by half and the translation level is also constant. A full-fledged business system is unusual, and the company has already begun operation of the system by focusing on technical manuals for foreign-affiliated companies' products for Japan. One year later, it will have a 10,000 page monthly system. <0015> The developed system is TDMS. It has started operation for English-language technical manuals, and has received orders from several US electronic component and communication equipment manufacturers. To promote this business,
We have recently set up a specialized section, the “Machine Translation Department”. [0016] In this system, an English-language manuscript is received on a magnetic medium such as a floppy disk, and a series of operations such as pre-editing for setting a format, translation, and production of a copy are processed by a computer. Most of the translation work by humans was eliminated, and the translators only worked on parts that could not be translated by the machine, and on the rewrite part to make natural Japanese. <0017> It is said that machine translation can be completed in four hours if one translator takes one month. It does not change the parts of the manual that do not need to be translated, such as numbers and units, to prevent data entry errors during Japanese translation. The original print style can be reproduced in Japanese as it is. (Hereinafter abbreviated; from the article of Nihon Keizai Shimbun on July 1, 1991) The user is the display unit 1
It is assumed that the portion indicated by the underline is marked by the input unit 2 while looking at the above-mentioned sentence displayed in. The sentence segmentation unit 3 starts segmentation of a sentence when the user inputs a mark. For example, for the “Machine Translation” mark on the first line, first look for the boundaries of the sentence according to the flowchart of FIG. 2 and reach the beginning of the sentence.
The head of the sentence, that is, “the instruction manual for the product ...” is the starting position S of the sentence. Next, a search is made for a boundary between sentences, and the punctuation mark "... introduced in the work process" in the middle of the second line becomes the boundary. One sentence containing the phrase thus marked is cut out in order. <0018> The unit sentence detection unit 5 extracts unit sentences in accordance with the flowchart of FIG. First, a natural language analysis of the extracted sentence is performed to determine the structure of the sentence. FIG. 4 shows an example of the structure of a sentence obtained by this analysis. The range of each unit sentence as shown in FIGS. 5 (1) and 5 (2) is determined with a unit sentence centered on one predicate obtained by this structure as a unit sentence, and the unit sentence includes a word marked by the user. Things are extracted in order. In this example, there is a mark in the "Machine Translation" part, so a united product centered on the predicate "Practical use""A major manufacturer of instruction manuals for products has implemented a machine translation system that translates English manuals into Japanese. Is extracted. In this way, the following summary sentence generation data is obtained. After these are stored in the storage unit 4,
Output by the output unit 7. <0019> Summarized sentence creation data Toin, a major manufacturer of product instruction manuals, has commercialized a machine translation system that converts English manuals into Japanese sentences. The delivery time is halved compared to humans, and the translation level is constant.
One year later, it will have a 10,000 page monthly system. The developed system is TDMS. Operation started for technical manuals in English. It has received orders from several US electronic component and communication equipment manufacturers. A translator can take a month to complete a machine translation job in four hours. <0020> Since these summary sentence creation data are sentences composed of the minimum necessary elements including characters and phrases marked by the user, by appropriately combining these, the following summary sentence can be obtained. It can be easily created by the user. In this example, focusing on the "technical manual" with the system name, the two sentences are combined and expressed, and the order of the other sentences is changed. <0021> Example of creating a summary sentence "Toin, a major manufacturer of product instruction manuals, has commercialized a machine translation system" TDMS "that converts English manuals into Japanese sentences. It has received orders from several US electronic component and communication equipment manufacturers. A translator can take a month to complete a machine translation job in four hours. Delivery time is halved compared to manual labor,
The translation level also becomes constant. One year later, it will have a 10,000 page monthly system. '''' -------------------------------------

In such an input document example, the following parts are extracted for “Example”.

<Example of Extracted Document> ------------------------------------- <Example><0012> The display unit 1 is, for example, a CRT display, and displays a text stored in the storage unit 4. The input unit 2 is a mouse, a touch key, or the like, and the user uses the input unit 2 to mark (position designation) a character or a phrase that is considered important while viewing the text displayed on the display unit 1. Sentence segmentation part 3
As shown in FIG. 2, a search is made for a sentence-to-sentence boundary from the mark toward the beginning of the sentence, the position is set to "S" (step 11), and the sentence is then sent from the mark toward the end of the sentence. Then, the boundary between the sentence and the sentence is searched, the position is set to “E” (step 12), and finally the range from S to E is cut out as one sentence. (Step 13). As shown in FIG. 3, the unit sentence detection unit 5 performs a natural language analysis of the sentence cut out by the sentence extraction unit 3 using the analysis dictionary 6 (see Reference: Shin Nagao, "Japanese Information Processing" (electronic communication Society, 1985)) (Step 21), 1
The range of unit sentence, which is a group around one predicate, is determined (step 22), unit sentences including marked characters and phrases are detected, and stored in the storage unit 4 (step 23). The output unit 7 is, for example, a printer, when the processing by the sentence extraction unit 3 and the unit sentence detection unit 5 is completed for all the marks,
All the unit sentences stored in the storage unit 4 are output as summary sentence creation data which is a source of the sentence summary. -------------------------------------

According to the important sentence extracting apparatus 1 described above,
Analyzes the expression of the most important part in the input document, retains the important words, and determines the importance of other components based on whether or not the important words are included in a specific form. Therefore, it is possible to specify only a portion having higher importance.

[0157]

According to the first aspect of the present invention, an important sentence can be appropriately extracted according to the type of the structure of the document.

According to a second aspect of the present invention, in the important sentence extracting apparatus according to the first aspect, the type of a document structure is specified based on the presence or absence of an expression that specifies the document structure, and an important sentence is appropriately extracted. Can be.

According to a third aspect of the present invention, in the important sentence extracting apparatus according to the first or second aspect, the type of the structure of the document is specified by the length of the document, and the important sentence can be appropriately extracted. .

According to the present invention, an important sentence can be appropriately extracted according to a document type.

According to a fifth aspect of the present invention, in the important sentence extracting apparatus according to the fourth aspect, by defining unimportant portions in a document in a table, other portions can be extracted without leaking. Possible and potentially important parts can be extracted without dropping.

According to a sixth aspect of the present invention, in the important sentence extracting apparatus according to the fourth aspect, an important part in a document can be specified in multiple stages according to the importance.

According to a seventh aspect of the present invention, in the important sentence extracting apparatus according to the fourth aspect, in addition to a part specified as an important part in the target document, a part related thereto can be specified as necessary. This makes it possible to read the document efficiently.

According to an eighth aspect of the present invention, in the important sentence extracting apparatus according to the seventh aspect, an important part is additionally specified,
Alternatively, when the number of locations to be specified is increased, a desired location can be specified using the relation name as a clue.

According to a ninth aspect of the present invention, in the important sentence extracting apparatus according to the sixth aspect, the expression of a portion having a relatively high importance in the document is analyzed, and thereby the importance of the other part is determined. By doing so, it is possible to specify only a portion having higher importance in the document.

According to the tenth aspect, an important sentence can be appropriately extracted according to the type of the structure of the document.

According to an eleventh aspect of the present invention, in the important sentence extracting method according to the tenth aspect, the type of the structure of the document is specified based on the presence or absence of an expression specifying the document structure, and the important sentence is appropriately extracted. Can be.

According to a twelfth aspect of the present invention, in the important sentence extracting method according to the tenth or eleventh aspect, the table expresses the type of the document structure using a document length.

Therefore, the type of the structure of the document can be specified based on the length of the document, and the important sentence can be appropriately extracted.

The invention according to claim 13 refers to a table stored in a predetermined storage device and describing a document type in association with the presence or absence of importance of a part in the document of the type. The important sentence extraction method includes an important part specifying step of specifying an important part in a target document.

Therefore, important sentences can be appropriately extracted according to the document type.

According to a fourteenth aspect of the present invention, in the important sentence extracting method according to the thirteenth aspect, the table describes requirements of a non-important portion in the document, and the important portion specifying step includes the step of: Referring to the table, a portion other than the portion corresponding to the requirement of the non-important portion in the target document is specified as the important portion.

Therefore, by defining unimportant parts in the document in the table, it is possible to extract other parts without leaking, and it is possible to extract parts that may be important without dropping them. it can.

According to a fifteenth aspect of the present invention, in the important sentence extracting method according to the thirteenth aspect, the table describes requirements at important locations in the document, and the magnitude of the importance of each requirement. Is described, the important location specifying step,
The important portion is specified in the target document by referring to the table and assigning the magnitude of the importance.

Therefore, an important part in a document can be specified in multiple stages according to the degree of importance.

According to a sixteenth aspect of the present invention, in the important sentence extracting method according to the thirteenth aspect, the table describes requirements of important places in the document and associates the requirements with each other. In the important part specifying step, the important part is specified by referring to the table and performing association between the important parts based on the association in a target document.

Therefore, in addition to the part specified as an important part in the target document, a related part can be specified as necessary, and the document can be read efficiently.

According to a seventeenth aspect of the present invention, in the important sentence extracting method according to the sixteenth aspect, the table describes a relation name of the association, and the important part specifying step performs the association. When specifying the important part, the relation name is also associated.

Therefore, when an important part is additionally specified or the number of specified parts is increased, a desired part can be specified by using the relation name as a clue.

According to an eighteenth aspect of the present invention, in the important sentence extracting method according to the fifteenth aspect, the table describes conditions relating to a description of a location having the requirement in association with the requirement. The important location specifying step refers to the table, specifies a location having the requirement where the importance is relatively large in the target document as the important location, and further determines the importance level from the identified important location. Is searched for a portion that satisfies the condition that is relatively large, and a portion that satisfies the requirement where the importance is relatively small in the target document and where the matching portion appears is also specified as the important portion.

Therefore, by analyzing the expression of a portion having a relatively high importance in the document and determining the importance of the other parts based on the analysis, it is possible to specify only the higher importance portion in the document. it can.

The invention according to claim 19 refers to a table which is stored in a predetermined storage device and describes the type of document structure and the importance of a part in a document having the structure in association with each other. And a computer-readable storage medium storing a program for causing a computer to execute an important portion specifying step of specifying an important portion in a target document.

Therefore, important sentences can be appropriately extracted according to the type of the structure of the document.

According to a twentieth aspect of the present invention, in the storage medium according to the nineteenth aspect, a table expressing the type of the document structure by using the presence or absence of an expression for explicitly indicating a document structure is used as the table. .

Therefore, the type of the structure of the document can be specified based on the presence / absence of the expression that clearly indicates the document structure, and the important sentence can be appropriately extracted.

According to a twenty-first aspect of the present invention, in the important sentence extracting method according to the nineteenth or twentieth aspect, a table expressing the type of the document structure using a document length is used as the table. .

Therefore, the type of the structure of the document can be specified by the length of the document, and the important sentence can be appropriately extracted.

The invention according to claim 22 refers to a table stored in a predetermined storage device and describing, in association with each other, the type of document and the importance of a part in the document of the type. The computer-readable storage medium stores a program for causing a computer to execute an important portion specifying step of specifying an important portion in a target document.

Therefore, important sentences can be appropriately extracted according to the document type.

According to a twenty-third aspect of the present invention, in the storage medium according to the twenty-second aspect, a table describing requirements of an insignificant part in a document is used as the table. With reference to the table, a portion other than the portion corresponding to the requirement of the non-important portion in the target document is specified as the important portion.

Therefore, by defining unimportant parts in the document in the table, it is possible to extract other parts without leaking, and it is possible to extract parts which may be important without dropping them. it can.

According to a twenty-fourth aspect of the present invention, in the storage medium according to the twenty-second aspect, the table describes requirements of important places in a document and describes a degree of importance of each requirement. In the important portion specifying step, the important portion is specified in the target document by attaching the magnitude of the importance with reference to the table.

Therefore, an important part in a document can be specified in multiple stages according to the degree of importance.

According to a twenty-fifth aspect of the present invention, in the storage medium according to the twenty-second aspect, a table that describes requirements of important places in a document and associates the requirements with each other is used as the table. In the important part specifying step, the important parts are specified by referring to the table and performing association between the important parts based on the association in a target document.

Therefore, in addition to the part specified as an important part in the target document, a part related thereto can be specified as necessary, and the document can be read efficiently.

According to a twenty-sixth aspect of the present invention, in the storage medium according to the twenty-fifth aspect, the table describing the relation name of the association is used as the table. When performing the identification of the important part, the relation name is also associated.

Therefore, when an important part is additionally specified or the number of specified parts is increased, a desired part can be specified using the relation name as a clue.

According to a twenty-seventh aspect of the present invention, in the storage medium according to the twenty-fourth aspect, the table, which corresponds to the requirement and describes a condition relating to description of a location having the requirement, is used as the table. The important point specifying step includes: referring to the table, specifying a part having the requirement having the relatively high importance in the target document as the important part, and determining the important part from the specified important part. Look for a part that meets the above condition with a relatively large degree,
In the target document, a part where the matching part appears in the document having the requirement with the relatively low importance is also specified as the important part.

Therefore, by analyzing the expression of a portion having relatively high importance in the document and determining the importance of the other portions by using the expression, it is possible to specify only the portion having higher importance in the document. it can.

[Brief description of the drawings]

FIG. 1 is a functional block diagram of an important sentence extracting apparatus according to an embodiment of the present invention.

FIG. 2 is a block diagram showing an electrical connection of the important sentence extracting device.

FIG. 3 is an explanatory diagram of a document structure information importance level table of the important sentence extraction device according to the first embodiment of the present invention;

FIG. 4 is a flowchart illustrating processing of a document analysis unit of the important sentence extraction device.

FIG. 5 is an explanatory diagram of a document structure information importance level table of the important sentence extraction device according to the second embodiment of the present invention;

FIG. 6 is a flowchart illustrating a process of a document analysis unit of the important sentence extraction device.

FIG. 7 is an explanatory diagram of a document structure information importance table of an important sentence extracting apparatus according to Embodiment 3 of the present invention;

FIG. 8 is a flowchart illustrating processing of a document analysis unit of the important sentence extraction device.

FIG. 9 is an explanatory diagram of a document structure information importance table of an important sentence extracting apparatus according to Embodiment 4 of the present invention;

FIG. 10 is a flowchart illustrating processing of a document analysis unit of the important sentence extraction device.

FIG. 11 is an explanatory diagram of a document structure information importance table of an important sentence extracting apparatus according to Embodiment 5 of the present invention;

FIG. 12 is an explanatory diagram of a document structure information importance table of an important sentence extracting apparatus according to Embodiment 6 of the present invention;

FIG. 13 is an explanatory diagram of a document structure information importance table of an important sentence extracting apparatus according to Embodiment 7 of the present invention;

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 important sentence extraction device 2 important location specifying means 3 table 13 storage means, storage device 15 storage means, storage device

Claims (27)

[Claims] 1. A storage means for storing a table in which a type of a document structure is associated with the presence or absence of importance of a part in a document having the structure, and a target document with reference to the table. An important part specifying means for specifying an important part in the important sentence extracting apparatus. 2. The important sentence extracting apparatus according to claim 1, wherein the table expresses the type of the document structure by using the presence or absence of an expression that clearly indicates the document structure. 3. The important sentence extracting apparatus according to claim 1, wherein the table expresses the type of the document structure using a document length. 4. A storage unit for storing a table in which a document type is associated with the presence or absence of importance of a part in a document of the type, and a table in the target document with reference to the table. Important part specifying means for specifying an important part of the sentence. 5. The table describes requirements of an insignificant place in the document, and the important location specifying means corresponds to the requirement of the insignificant place in a target document by referring to the table. The important sentence extracting apparatus according to claim 4, wherein a part other than the part to be executed is specified as the important part. 6. The table describes requirements of important points in a document, and describes the magnitude of importance of each requirement. The important point specifying means refers to the table and refers to the table. The important sentence extracting apparatus according to claim 4, wherein the important part is specified in a target document by assigning a magnitude of importance. 7. The table describes requirements of important points in a document and associates the requirements with each other, and the important point specifying means refers to the table and performs the association in a target document. The important sentence extracting apparatus according to claim 4, wherein the important points are specified by associating the important points based on the information. 8. The method according to claim 7, wherein the table describes a relation name of the association, and the important location specifying unit also associates the relation name when the association is performed to specify the important location. Significant sentence extraction device. 9. The table describes conditions relating to the description of a location having the requirement in association with the requirement, and the important location specifying means refers to the table and writes the condition in a target document. A location having the requirement having a relatively high importance is specified as the important location, and a portion matching the condition with the relatively high importance is searched for from the specified important location, and in the target document, 7. The important sentence extracting apparatus according to claim 6, wherein a part having the requirement having the relatively small importance and having the matching part appearing is also specified as the important part. 10. Referring to a table stored in a predetermined storage device and describing the type of document structure and the importance of a portion in the document having the structure in association with each other, the target document An important sentence extraction method including an important place specifying step of specifying important places. 11. The important sentence extraction method according to claim 10, wherein the table expresses the type of the document structure by using the presence or absence of an expression that clearly indicates the document structure. 12. The table according to claim 10, wherein the table represents the type of the document structure using a document length.
2. The important sentence extraction method according to 1. 13. A target document is referred to by referring to a table stored in a predetermined storage device and describing the type of the document and the importance of a part in the document of the type in association with each other. An important sentence extraction method including an important point specifying step of specifying important points in a document. 14. The table describes requirements of a non-important location in a document, and the important location specifying step corresponds to the requirement of the non-important location in a target document by referring to the table. 14. The important sentence extracting method according to claim 13, wherein a part other than the part to be executed is specified as the important part. 15. The table describes requirements of important locations in a document, and describes the magnitude of importance of each requirement. The important location specifying step refers to the table and refers to the table. 14. The important sentence extracting method according to claim 13, wherein the important part is specified in the target document by assigning a magnitude of importance. 16. The table describes requirements of important locations in a document and associates the requirements with each other. The important location specifying step refers to the table and performs the association in a target document. 14. The important sentence extraction method according to claim 13, wherein the important points are specified by associating the important points with each other. 17. The method according to claim 16, wherein the table describes the relation name of the association, and the important location specifying step also associates the relation name when the association is performed to specify the important location. Extraction method of important sentence of description. 18. The table describes conditions relating to the description of a location having the requirement in association with the requirement, and the important location specifying step refers to the table and refers to the condition in a target document. A location having the requirement having a relatively high importance is specified as the important location, and a portion matching the condition with the relatively high importance is searched for from the specified important location, and in the target document, 16. The important sentence extraction method according to claim 15, wherein a part having the requirement with the relatively small importance and in which the matching part appears is also specified as the important part. 19. Referring to a table stored in a predetermined storage device and describing the type of the document structure and the importance of a part in the document having the structure in association with each other, A computer-readable storage medium storing a program for causing a computer to execute an important part specifying step of specifying an important part. 20. The storage medium according to claim 19, wherein a table expressing a type of the document structure by using presence / absence of an expression for explicitly indicating a document structure is used as the table. 21. The important sentence extracting method according to claim 19, wherein a table expressing the type of the document structure using a document length is used as the table. 22. A target document is referred to by referring to a table stored in a predetermined storage device and describing the type of the document in association with the importance of a portion in the document of the type. A computer-readable storage medium storing a program for causing a computer to execute an important portion specifying step of specifying an important portion in the computer. 23. A table in which requirements of an insignificant point in a document are described as the table, and the step of identifying an insignificant point includes the step of identifying the insignificant point in the target document by referring to the table. 23. The storage medium according to claim 22, wherein a part other than the part corresponding to the requirement is specified as the important part. 24. As the table, a table describing requirements of important points in a document and describing the magnitude of importance of each requirement is used. 23. The storage medium according to claim 22, wherein the important portion is specified in the target document by referring to and assigning the magnitude of the importance. 25. A table which describes requirements of important parts in a document and associates the requirements with each other as the table, and in the important part specifying step, the table refers to the table to be used as a target. 23. The storage medium according to claim 22, wherein the important part is specified by performing the association between the important parts based on the association. 26. A table in which a relation name of the association is described as the table, wherein the important part specifying step associates the relation name when the association is performed to specify the important part. Item 29. The storage medium according to Item 25. 27. A table in which conditions relating to the description of a location having the requirement are described in correspondence with the requirement, and the important location specifying step refers to the table and refers to a target document. Identify a location having the requirement in which the importance is relatively large as the important location, and search for a portion that satisfies the condition in which the importance is relatively large from the identified important location, and 25. The storage medium according to claim 24, wherein a part of the document having the requirement with the relatively low importance and in which the matching part appears appears as the important part.