JP2001101207A - Document summarizing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To exactly provide information desired by user and to provide the summary of the entire document. SOLUTION: Information under consideration is extracted based on a predetermined rule from a document inputted by a document information input part 1 and stored in a sentence information storage part 5 by a noticing information deciding means 2. When a character string in reference relation with the information under consideration exists, its reference information is stored in the sentence information storage part 5 by a reference information deciding means 6. When a character string having relation with the information under consideration exists, its related information is stored in the sentence information storage par 5 by a relation deciding means 10. A summary sentence of the inputted document is generated based on the information under consideration the reference information and the related information stored in the sentence information storage part 5 by a summary sentence generating means 11.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document summarizing apparatus for summarizing and displaying a document, and more particularly to a document summarizing apparatus for extracting and displaying contextual information.

[0002]

2. Description of the Related Art There is a document summarizing apparatus for extracting an abstract from given original text data based on a predetermined rule. Conventionally, such a document summarizing apparatus has been disclosed in, for example, JP-A-8-255172.

According to the technique disclosed in this document, a plurality of types of sentences for identifying the contents of a sentence such as opinions and proposals are set from an original text database storing original text data constituting a document. The document retrieval system includes an original sentence processing unit that creates excerpt sentence data in sentence units classified into these sentence types and stores the data as an excerpt sentence database. Here, the original sentence processing unit includes a type determination unit for extracting the excerpt sentence data corresponding to the specified sentence type, and a shaping unit for shaping the excerpt sentence data into a predetermined format such as, for example, a conjunction removed.

[0004] Its purpose is (1) to reduce the load required for a user's document search work by extracting and displaying only desired excerpts and information from the entire text of a searched document. ) Excerpt sentences and information can be extracted and displayed from the same document based on various viewpoints and criteria to realize search processing corresponding to various search purposes. (3) Full text of searched documents It is an object of the present invention to realize a search process capable of adjusting the amount of the excerpt sentences and information to be displayed in accordance with the priority order when only the excerpt sentences and information are extracted and displayed based on a predetermined criterion.

[0005]

In the case of extracting information of interest from a document and summarizing the information, it is necessary to grasp the relevance of each extracted information. In this case, in addition to extracting attention information for individual sentences, it is also important to grasp the anaphoric relation and reference relation of each attention information in the entire document as a range.

However, in the above-described conventional method, since the scope at the time of extracting information is realized by assuming processing in units of sentences, the relevance of contextual information over the entire document is reduced. There was a problem that it could not be grasped.

[0007] For example, in the above-described conventional technique, the user selects, for example, a type representing a subject of action such as "who" from a type of 5W1H excerpt information prepared in advance using a type determination dictionary. Then, for example, "Prime Minister" from the original text,
The names of persons who are the main actors of some operations, such as "the former Prime Minister" and "the Secretary of the Science and Technology Agency" are extracted and displayed on the excerpt information screen. Similarly, if the user selects a type that represents a place such as "where", the same sentence represents a place such as "Prime Minister's office" or "Domestic nuclear power plant". The words are extracted and displayed on the excerpt information screen.

However, in this method, it is possible to grasp the relationship between “Prime Minister” and “Prime Minister's residence” from a certain subject and the type of “where” in the sentence 1 with respect to a certain document. In the case where there is a “parliament building” as a “where” type, the range of “the prime minister” and “the parliament building” exceeds the recognition unit “sentence”.
It was not possible to associate the "Prime Minister" with the Parliament Building in sentence 2.

[0009]

The present invention employs the following structure to solve the above-mentioned problems. <Structure 1> Attention information determining means for extracting attention information from a character string in an input document based on a predetermined rule. If there is a character string in reference to the attention information in the input document, this is Reference information determining means for assigning reference information to the input document, and relevance determining means for assigning, to the input document, relevant information indicating a character string that is relevant to the attention information in the input document, if there is a character string in the input document. And a summary sentence generating means for generating a summary sentence of the input document based on the attention information, the reference information, and the related information.

<Structure 2> In the document summarizing apparatus described in Structure 1, when an arbitrary instruction is received from the user, the instruction of the input document is reflected by reflecting the instruction of the user based on the attention information, the reference information, and the related information. A document summarizing apparatus comprising: a customizing unit that generates an abstract.

<Structure 3> In the document summarizing apparatus described in structure 1 or 2, it is determined whether or not tag information is added to the input document. If tag information is added, the type and attribute of the tag information are determined. Tag information extracting means for extracting a tag, a summary sentence generating means for generating a summary of the input document based on the type and attribute of the tag information extracted by the tag information extracting means, and the attention information; A document summarizing means for generating a summary of the input document by reflecting a user's instruction based on the type and attribute of the attention information, the reference information, the related information, and the tag information. apparatus.

<Structure 4> In the document summarizing apparatus described in Structure 3, it is determined whether or not there is image information to which tag information is added in the input document, and if there is image information, the image information is extracted. The extracting means generates a summary of the input document based on the type and attribute of the tag information extracted by the tag information extracting means and the attention information, and, when the image information is extracted by the image information extracting means, A summary sentence generating means for adding an image, and, when receiving an arbitrary instruction from the user, reflecting the instruction of the user based on attention information, reference information, related information, and type and attribute of tag information, and summarizing the input document. And a customizing means for adding an image when the image information is extracted by the image information extracting means.

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention will be described below in detail with reference to specific examples.

Embodiment 1 <Configuration> FIG. 1 is a configuration diagram showing Embodiment 1 of a document summarizing apparatus according to the present invention. The apparatus shown in the figure is composed of a microcomputer or the like, and includes a document information input unit 1, attention information determination means 2, a character string pattern database (character string pattern DB) 3,
Syntax pattern database (syntax pattern DB) 4, sentence information storage 5, reference information determining means 6, syntax analysis rule 7, word dictionary 8, attribute information determining means 9, relevance determining means 10, summary sentence generating means 11, summary An information output unit 12 is provided.

The document information input unit 1 is an input means for a user to input natural language information, such as a keyboard, a mouse, an OCR (optical character reading device), and voice input. The attention information determining means 2 has a function of extracting attention information from a character string in the document information input from the document information input unit 1 based on a predetermined rule. That is, the attention information determining means 2 includes a character string related to information desired by the user and a character string pattern database 3 in the input document.
Determines if there is a string that matches the string inside,
If there is, it has a function of marking a corresponding part as attention information.

The character string pattern database 3 includes, for example, affix patterns such as "san" of "san" and "ministry" of "Ministry of Finance", and "prime" of "prime OO".
It is a database that stores the meanings and attributes of each having a pattern of functional words such as “bank” of “Bank of Japan”.

In the syntactic pattern database 4, for example, “Fukui” of “I talked to Fukui” and “Fukui” of “Arrive at Fukui” are both character strings of “Fukui” in appearance. This is a database storing a specific word pattern for distinguishing between "person" and "place".

The sentence information storage unit 5 includes the attention information determination unit 2
This is a storage section for temporarily storing the character string of the attention information portion determined in step (1) and accompanying information associated therewith, position information in the document occupied by the attention information, and the like.

The reference information determining means 6 has a function of recognizing attention information indicating the same entity from attention information or a character string in the input information and marking the information. That is, it has a function of extracting a character string having a reference relationship with the attention information in the input document. The reference information determining means 6 uses a characteristic syntax pattern stored in the syntax pattern database 4 to determine a reference relationship, calls a word dictionary 8 and a syntax analysis rule 7 to generate a syntax, It is configured to analyze the structure of the style.

The parsing rule 7 is, for example, a level rule for grasping words, phrases, clauses, and the like in the parsing of the machine translator. The word dictionary 8 is a dictionary that stores general words and phrases necessary for the analysis processing of the syntax analysis rules 7.

The attribute information determining means 9 has a function of recognizing and marking or giving an attribute relating to the information of interest in the input information. That is, the attribute information determining means 9 has a function of determining which is correct when the attribute is ambiguous, such as a case where a plurality of attributes are considered for the attention information.

The relevancy determining means 10 is a functional unit for recognizing the relevancy between pieces of attention information in the input information and marking the information.

The summary sentence generating means 11 combines the information determined by the attention information determining means 2 to the relevance determining means 10 and marked and the information stored in the text information storage section 5 as a summary. This is the creation unit to generate. That is, the summary sentence generation unit 11 has a function of creating a summary based on the attention information, the reference information, and the related information given to the input document stored in the text information storage unit 5.

The summary information output unit 12 is composed of a CRT, a printer, a network, or the like, and is created in consideration of attention information extracted from natural language information input by a series of means and its attributes, relationships, and reference relationships. It has a function of outputting the summarized information.

The above-mentioned attention information deciding means 2, reference information deciding means 6, attribute information deciding means 9, and relevancy deciding means 1 are described above.
0, the summary sentence generating means 11 is composed of software corresponding to the function of each means and hardware such as a processor and a memory for executing the software. Further, the character string pattern database 3, the syntax pattern database 4, the text information storage unit 5, the syntax analysis rule 7, and the word dictionary 8 are realized on a storage device such as a magnetic disk device or a semiconductor memory.

<Operation> FIG. 2 is a flowchart showing the operation of the document summarizing apparatus. The document information input from the document information input unit 1 is document information composed of a text character string typified by a newspaper article or an e-mail, and is input in a lump or in a unit (step S101). ).

FIG. 3 is an explanatory diagram showing an example of the input information. At this time, it is assumed that a sentence "Prime Minister O △ has announced ... on April 1" as shown in the input document information.

The input document information is referred to the character string pattern database 3 by the attention information determining means 2 (step S102), and a character string matching the pattern in the database or a character string related to the information desired by the user. It is checked whether there is a column (step S10).
3). At this time, the character string pattern database 3 contains
For example, assume that the following information is stored.

FIG. 4 is an explanatory diagram of a character string pattern stored in the character string pattern database 3. here,
[...] indicates that any character can appear any number of times. If the character is followed by a character string such as "President" or "Prime Minister", the type is person name and the attribute is political. It indicates that it is a house.

If the input information of FIG. 3 has a character string pattern as shown in FIG. 4, the pattern "Prime Minister" matches. Therefore, the character string “○ と い う” of “Prime Minister △” in the document information is determined as the attention information, and SG
ML (standard generalized markup language) and HT
ML (hypertext markup language), XML (extensi
This is marked by means such as tagging such as ble markup language (step S104) and stored in the text information storage unit 5 (step S105).

FIG. 5 is an explanatory diagram showing an example of document information in which attention information is marked. As shown in FIG. 5A, since the pattern of "Prime Minister" in the document information matches, this is determined as attention information. If it is determined that the information is attention information, it is tagged as shown in FIG. 5B, and a type and an attribute as shown in FIG. 4 are given.

If it does not match the character string pattern in the character string pattern database 3 in step S103, the process proceeds to step S106.

Next, the reference information determining means 6 checks the reference relation of the attention information. First, the syntactic analysis rule 7 is activated (step S106). The marked information stored in the text information storage unit 5 is subjected to the attention information and the noun phrase using the syntax pattern of the syntax pattern database 4 and the word dictionary 8. It is determined whether there is a reference destination or a reference source (step S107).

For example, in the case of a noun phrase such as the attention information “○ △”, “he”, “primary”, “the politician”
Such paraphrasing often appears. In this case, since these paraphrases all point to the same entity “△ Δ”, they can be said to be in a reference relationship with each other.

Therefore, the character string determined to be paraphrased as described above is marked in the document where the paraphrase is found (step S108) and stored in the text information storage unit 5 (step S109). On the other hand, if the reference information is not found in step S107, the process proceeds to step S110.

FIG. 6 is an example of document information in which reference information is marked. Note that the attention information (person name: PM, politician) indicates the type of “person name” and the attributes “Prime Minister” and “politician” as shown in FIG. In this case, since the attribute is an information set characterizing a person's name, a plurality of attributes may exist. Also, reference information (003, 005, 006, 007: he)
Means that "he" is a paraphrase, and 003 etc. indicates the position where the paraphrase appears. In this case, the appearance position indicates, for example, what number noun phrase appears in the input document or what number noun phrase in the input document appears in the input document. However, any position information may be used as long as its position can be specified in the target document.

The attribute information determining means 9 checks the attribute information of the information of interest. First, the syntax analysis rule 7 is activated, and the syntax pattern database 4 and the word dictionary 8 are referred to (step S110). The syntax analysis rule determines whether attribute information can be obtained with respect to the noted information of the marked document information stored in the text information storage unit 5 (step S111). If the attribute information is obtained, the marked document information is determined. And the information is stored in the text information storage unit 5 (steps S112 and S113).

FIG. 7 is an explanatory diagram showing an example of document information in which attribute information is marked. Here, the reason why the reference information in FIG. 6 is changed to the attention information is that, for example, when “congress” is the attention information, the reference destination of the “lower house assembly” is the attention information. This is because the reference information is regarded as the same as the attention information. However, in the illustrated example, the target information is changed to the attention information as described above. However, if it is possible to know which noun phrase the reference information refers to, it is not always necessary to change the reference information to the attention information.

If no attribute information has been obtained in step S111, or if attribute information has already been given, the flow advances to step S114. Here, the attribute information is obtained from the syntax pattern database 4 or the word dictionary 8,
It may be obtained from the information stored in the text information storage unit 5.

FIG. 8 is an explanatory diagram showing an example of a syntax pattern in the syntax pattern database 4. FIG. 9 is an explanatory diagram illustrating an example of other document information in which attribute information is marked.

In FIG. 8, a pattern number indicates a pattern-specific number in the syntax pattern database 4, and a syntax pattern is an abstraction of a combination of syntaxes for obtaining necessary information in the attribute information determining means 9. It was done. For example, in pattern 22, a pattern of two consecutive nouns separated by a comma is shown, and the noun at the front indicates that the noun indicates "person" and the noun at the back indicates "unknown". .

For example, there is a series of nouns such as "section manager" and "patent Ichiro". Is determined to be a noun (phrase), but is not known in the word dictionary, so it is assumed to be an unknown word. At this time, “section manager (person), Ichiro Patent (unknown word)” matches the pattern number 22, and the “kind” of the pattern is determined as a person name, and the pattern is interpreted. “Interpretation of a pattern” indicates what relationship should be interpreted between the elements of the matched syntax pattern (two nouns in the pattern 22). In the above example, according to this interpretation, "section manager" is "patent Ichiro"
Can be interpreted as "personal name". Therefore, in the example of FIG. 9, the pattern is “Former Prime Minister, Taro Patent”. Note that the document information shown in FIG. 9 shows an example of attribute information based on a syntax pattern, and has no direct relationship with the document information shown in FIG.

The relevancy determining means 10 determines the relevancy between the pieces of attention information. First, the document information with a mark stored in the text information storage unit 5 is referred to, a syntax analysis rule 7 is activated for this document information, and the syntax pattern database 4 and the word dictionary 8 are referred to (step S11).
4).

In the syntax analysis rule 7, the text information storage 5
About the attention information of the document information with the mark stored in
It is determined whether there is a relationship with other attention information (step S115).

Here, the syntax analysis rule 7 refers to the syntax pattern database 4. FIG. 10 is an explanatory diagram illustrating an example of the syntax pattern. If a syntax pattern as shown in FIG. 10 is used, for example, in document information such as “... the prime minister participates on behalf of the Japanese government”, “the prime minister” is “the Japanese government”
And both are determined to be related. Furthermore, using the reference information, it is possible to ascertain the relevance of the Japanese government to “○ △”, which is the substance of the “Prime Minister”. FIG. 11 is an explanatory diagram of another document information in which attribute information indicating such an example is marked.

As described above, when there is relevance, information is added to the noticed information in the marked document information and stored in the text information storage unit 5 (steps S116 and S11).
7).

FIG. 12 is an explanatory diagram of document information in which attribute information is marked. This state shows the result of the recognition of the "relevance" between "Prime Minister △ @" and "the Japanese government". The reason for the relevance (003: government) (005: government) is as follows. .

First, as shown in FIG. 7, the document information (before the stage shown in FIG. 12) is “... <attention information (person name: PM,
Politician) | Reference information (003, 005, 006, 007: he) (00
4,008: Prime Minister) (008: The politician)>○> Prime Minister <attention information> ... ". In this case, FIG.
Reference information (004, 008: Prime Minister) of “Prime Minister” and “○
It is determined from the fact that the reference information (004, 008: prime minister) of “Prime Minister” matches (in short, has a reference relationship).

If the document information at the previous stage shown in FIG. 12 is “... <attention information (personal name: Prime Minister, politician) | reference information (003, 005, 006, 007: him) (004, 008) : Prime)
(008: The politician) | Relevance (003, 005: Government)> ○ △
Prime Minister <Attention information> ... "
The relevance is also determined from the fact that the reference information (003, 005: government) of “Japanese government” shown in FIG. 1 matches the related information (003, 005: government) of “Prime Minister △”.

If no relevance is found in step S115, the process proceeds directly to step S118.

Note that the steps S101 to S101
Each time a series of processing in step 117 proceeds, the attention information in the document and the information related thereto are updated as marked document information.
The text information is stored in the text information storage unit 5. The mark information held here can be referred to and used later or in a similar process performed in the same process. Further, in the character string pattern database 3 and the syntax pattern database 4, pattern information of words and phrases required when the above series of processing is performed and the combination thereof are stored in advance as databases, Referred to by

Finally, the summary sentence generating means 11 refers to the marked document information from the sentence information storage 5 (step S118). The summary sentence generating unit 11 creates a summary sentence by combining important or desired information of interest with reference to the marked information (step S).
119). Then, the summary information created by the summary sentence generation unit 11 is output from the summary information output unit 12 (step S1).
20).

FIG. 13 is an explanatory diagram showing an example of the summary information. The illustrated summary information shows details of the XX reform plan. As the output method, for example, as the default, “April 1,
We announced the reform plan. Is output, and the character string of the XX reform plan is displayed differently. And this ○○
○ A method of sequentially outputting the details of the reform proposal as needed can be considered. Such summary information can also be extracted from information such as attention information and reference information regardless of where it is located in the input document.

<Effects> According to the specific example 1 as described above,
Since the summary sentence in the input document is generated based on the attention information, the reference information, and the related information in the input document, the following effects can be obtained.

(1) Automatically determines the notable part of the text information or the information to be noticed with respect to the information desired by the user, and creates summary information by grasping the types and attributes of the noticed information and their relevance to each other. Therefore, the information desired by the user can be provided more accurately and simply as compared with the related art.

(2) Since the process is carried out while holding the document information in which each information is marked, it is possible to grasp the relevance and reference relationship of the attention information for the entire document.

<< Specific Example 2 >> In the specific example 2, in addition to the configuration of the specific example 1, a configuration for customizing and outputting summary information is added.

<Structure> FIG. 14 is a diagram showing the structure of a document summarizing apparatus according to the second embodiment. The apparatus shown in the figure includes a document information input unit 1, attention information determining means 2, a character string pattern database (character string pattern DB) 3, a syntax pattern database (syntax pattern DB) 4, a text information storage unit 5, and a reference information determining means 6. , A syntax analysis rule 7, a word dictionary 8, attribute information determining means 9, relevance determining means 10, summary sentence generating means 11, summary information output section 12, and customizing means 13. Here, the configuration other than the customizing unit 13 is the same as that of the first embodiment, and the description is omitted here.

The customizing means 13 combines the marked information determined by the attention information deciding means 2 to each of the relevance deciding means 10 and the information stored in the text information storage section 5 according to the user's instruction. For example,
It has a function to customize the summary information by performing a list, an index, highlighting a point of interest, and the like.

<Operation> FIG. 15 is a flowchart showing the operation of the second embodiment. In the specific example 2, the input from the document information input unit 1 is document information constituted by a text character string typified by a newspaper article or an e-mail, and is input collectively or in a unit (step). S201).

In the specific example 2, the operation from step S201 to step S217 is the same as the operation from step S101 to step S117 in the specific example 1, and the description is omitted here.

When the marked document information is stored in the sentence information storage unit 5 in step S217, or when there is no relevant attention information in step S215, next, the summary sentence generation unit 11 and the customization unit 13 Marked document information is referenced from the text information storage unit 5 (step S218). Here, it is determined whether customization of the summary information is specified by the user (step S219). If it is specified, the process proceeds to step S220. If not, the process proceeds to step S221.

In step S220, the customizing means 13 customizes the summary information in a form suited to the user's purpose by combining important or desired user's attention information while referring to the marked information.

FIG. 16 is an explanatory diagram of summary information customized as a personal name list. FIG. 17 is an explanatory diagram of summary information customized as a product catalog.

In the example of FIG. 16, summary information is generated for items such as occupation for a person's name, affiliation organization, and location of the organization. In the example of FIG. 17, summary information is generated for each product in the form of the type of the product, the source of the product, the source of the sale, the location, and the situation.

On the other hand, in step S221, the summary sentence generating means 11 creates a summary sentence by combining important or desired information of interest by the user while referring to the marked information.

Then, the summary information customized by the summary sentence generating means 11 or the customizing means 13 is output from the summary information output section 12 (step S222).

<Effects> As described above, according to the second embodiment, a means for customizing the summary information is provided in addition to the configuration of the first embodiment. The effect is obtained.

(1) A noteworthy portion of text information is automatically determined, and the association of the noticeable information can be summarized according to the purpose of the user and output as a list.

(2) Even if the document information cannot be displayed all at once on the display device (summary information output unit 12) of the computer, it is necessary to display the necessary information at once without dropping necessary information. Becomes possible.

<< Third Specific Example >> In a third specific example, tag information is extracted from document information including tag information.

<Structure> FIG. 18 is a diagram showing the structure of a document summarizing apparatus according to the third embodiment. The apparatus shown in the figure includes a document information input unit 1, attention information determining means 2, a character string pattern database (character string pattern DB) 3, a syntax pattern database (syntax pattern DB) 4, a text information storage unit 5, and a reference information determining means 6. , Syntax analysis rules 7, word dictionary 8, attribute information determining means 9, relevance determining means 10, summary sentence generating means 11, summary information output section 12, customizing means 13a, tag pattern database (tag pattern DB) 14, tag information It consists of extraction means 15. Here, the tag pattern database 14, the tag information extracting means 15, and the customizing means 13
Except for a, the configuration is the same as that of the specific example 2, and the description is omitted here.

The tag pattern database 14 is a database for storing basic tag set formats and patterns. The tag information extraction unit 15 determines whether tag information such as SGML, HTML, or XML has been added to the input information.
The tag information is accessed and determined, and when tag information is added, the type and position information are extracted and stored in the text information storage unit 5.

Further, the customizing means 13a determines the marked information determined by each means of the relevance determining means 10 from the tag information extracting means 15 and the information stored in the text information storage section 5 according to the user's instruction. And has a function of customizing the summary information by performing, for example, a list, an index, highlighting of a point of interest, and the like.

<Operation> FIGS. 19 and 20 are flowcharts (parts 1 and 2) showing the operation of the third embodiment.

In the specific example 3, the document input from the document information input unit 1 is SGML stored in the CD-ROM.
H as represented by tagged newspaper articles and WWW information
Tagged document information including tag information such as TML or SHTML and text character string information, and is input collectively or in a unit (step S301).

FIG. 21 is an explanatory diagram showing an example of an input document. For the tagged document information as shown in the figure, the tag pattern database 14 is referred to by the tag information extracting means 15 (step S302), and whether there is tag information matching the tag pattern in the database or tag information designated by the user is present. Is checked (step S30).
3).

FIG. 22 is an explanatory diagram showing an example of a tag pattern. In step S303, if tag pattern information as shown in FIG. 22, for example, is stored in the tag pattern database 14, tag information in a document that matches this is determined to be a tag, and its type and attribute are grasped.

For example, “<DATE>” and “</ DATE>” of “<DATE> April 1 </ DATE>” are SGML or H
It is determined that the tag information is TML tag information, and the character string between them is grasped as date information. Also, “<TITLE>
… </ TITLE> ”for“ <TITLE> ”and </ TITLE
> ”Is determined to be SGML tag information, and the character string between them is grasped as a title. Further, since the type of the tag is uniquely determined, the tag-attached document is represented by S
It is recognized as GML.

The grasped tag information is marked by means such as tagging so that it can be recognized in the summarization apparatus and is easy to process (step S304).
(Step S305). FIG. 23 is an example of tagged document information in which tag information is marked.

Returning to FIG. 19, if there is no character string that matches the tag pattern in step S303, the flow advances to step S306. Hereinafter, since operations from step S306 to step S321 are the same as steps S202 to S217 in the specific example 2,
The description here is omitted.

In step S322, the summary sentence generating means 1
1 and the customizing unit 13 refer to the marked document information from the text information storage unit 5. Next, it is determined whether customization of the summary information is specified by the user (step S323). If it is specified, the process proceeds to step S324. If not, the process proceeds to step S325.

The summary sentence generating means 11 creates a summary by combining important or desired user's attention information while referring to the marked information. The tag information is also input at a necessary portion when the tag information is input. (Step S325).

The customizing means 13 combines important or desired information of interest by referring to the marked information, for example, a personal name list as shown in FIG. 16 or a product catalog as shown in FIG. In step S324, the summary information is customized according to the purpose of the user, and tag information is added to necessary portions according to the type of tag at the time of input.

Finally, the summary information generating section 11 and the customizing section 13 convert the summary information in consideration of the attention information desired by the user into a form desired by the user (normal text document or tagged document) in the summary information output section. 12 (step S326).

<Effects> As described above, according to the specific example 3, in addition to the configuration of the specific example 2, a means for extracting the tag information from the document information including the tag information is provided. In addition to the above, the following effects can be further obtained.

(1) Even for a tagged document such as SGML, HTML, or XML, attention information can be grasped and summary information can be provided in the same manner as ordinary document information.
(2) Since the tag pattern information is held, it is possible to add a tag to the summary information of normal document information and to output the summary information, or to provide the summary information of a tagged document as heterogeneous tagged information.

<< Specific Example 4 >> In the specific example 4, when image information is included in a document, the image information is also extracted.

<Structure> FIG. 24 is a diagram showing the structure of a document summarizing apparatus according to the fourth embodiment. The apparatus shown in the figure includes a document information input unit 1, attention information determining means 2, a character string pattern database (character string pattern DB) 3, a syntax pattern database (syntax pattern DB) 4, a text information storage unit 5, and a reference information determining means 6. , Syntax analysis rules 7, word dictionary 8, attribute information determination means 9, relevance determination means 10, summary sentence generation means 11a, summary information output unit 12, customization means 13b, tag pattern database (tag pattern DB) 14, tag information It comprises an extracting unit 15, an image information extracting unit 16, and an image information storage unit 17. Here, except for the summary sentence generation unit 11a, the customization unit 13b, the image information extraction unit 16, and the image information storage unit 17, the description is omitted because it is the same as that of the specific example 3.

The image information extracting means 16 has a function of referring to the marked document information stored in the text information storage section 5 and checking the image information in the document. The tag pattern database 14 is a database that stores basic tag set formats and patterns. The tag information extracting means 15 adds, for example, SGML or HTM to the input information.
Accessing the tag pattern database 14 to determine whether tag information such as L, XML, etc. has been added,
When tag information has been given, it has a function of extracting those types and position information and storing them in the text information storage unit 5.

The summary sentence generating means 11a stores the information determined and marked by each means from the tag information extracting means 15 to the relevancy determining means 10 and the sentence information storage 5 and the image information storage 17. It has a function of combining the information with the information, and embedding the image position information as necessary, and generating an image-attached summary.

Further, the customizing means 13b stores the information determined and marked by the respective means of the relevance determining means 10 from the tag information extracting means 15 and the text information storage section 5 and the image information storage section 17 according to the user's instruction. Combine with the stored information, and, if necessary, in the form of embedding image position information and attaching the image, for example, perform a list, index, highlight of the point of interest,
It has a function to customize summary information.

<Operation> FIGS. 25 and 26 are flowcharts (Nos. 1 and 2) showing the operation of the fourth embodiment.

In the specific example 4, the document input from the document information input unit 1 is SGML stored in the CD-ROM.
H as represented by tagged newspaper articles and WWW information
Tagged document information including tag information such as TML or SHTML and text character string information, and is input collectively or in a unit (step S401).

The input tag-added document information is referred to the tag pattern database 14 by the tag information extracting means 15 (step S402), and tag information matching the tag pattern in the database or tag information designated by the user is stored. It is checked whether there is (Step S4)
03).

The operations in steps S403 to S405 are the same as the operations in steps S303 to S305 in the third embodiment.

Next, the image information extracting means 16 refers to the marked document information stored in the text information storage section 5 (step S406), and checks the image information in the document (step S407). As a result, when an image or a part for designating image display is found in the mark document, the image information extracting unit 16 stores the image display part and the corresponding character string part in the image information storage unit 17 as a set. (Steps S408 and S409).

FIG. 27 is an explanatory diagram showing an example of tagged document information. FIG. 28 is an explanatory diagram illustrating an example of the image information stored in the image information storage unit 17.

For example, in a sentence as shown in FIG.
Character string part "OKTAC45000" with tag
And link information indicating the position of image information “<img name =“… ok
tac45000.gif ">" is set as a set and stored in the image information storage unit 17 as shown in FIG. On the other hand, if the image information portion is not found, the process proceeds to step S410.
Proceed to.

Hereinafter, steps S410 to S4
The operations up to 25 are the same as the operations from step S306 to step S321 in the specific example 3.

After the operation from step S401 to step S425 is completed, the summary sentence generating means 11 and the customizing means 13b refer to the marked document information from the text information storage section 5 (step S427). Here, if customization of the summary information is specified by the user, the process proceeds to step S428, and if not, the process proceeds to step S429.

In step S428, the summary sentence generating means 1
1 combines the important or desired information of interest with reference to the marked information, creates a summary with image information as needed, and, where necessary, adds the tag information to the tag at the time of input. It is given according to the type of.

On the other hand, in step S429, while referring to the information marked by the customizing means 13b,
Combining important or user attention information, for example,
Customize the summary information with the image information according to the purpose of the user, such as a person name list or a product catalog, and also add the tag information where necessary according to the type of tag at the time of input. Granted.

FIG. 29 is an explanatory diagram of summary information customized as a personal name list. FIG. 30 is an explanatory diagram of summary information customized as a product catalog.

As described above, the summary sentence generation unit 11 and the customization unit 13b obtain the summary information in consideration of the attention information desired by the user in a form desired by the user (normal text document or tagged document). (Step S430). Image information is added to the obtained summary information, and in the case of a document with a tag, the summary information output unit 12 provides the sample with a reduced image or the like attached thereto.

The user can refer to the document information holding the image information as shown in FIGS. 29 and 30 by designating the reduced image or the reference portion as necessary.

<Effects> As described above, according to the fourth embodiment,
Since the image information extracting means for extracting the image information included in the document is provided in addition to the configuration of the specific example 3, the following effect can be obtained in addition to the effect of the specific example 3.

(1) A noteworthy portion of text information is automatically determined, and further, related image information can be displayed in the same manner when displaying a summary. (2) When summarizing and displaying search results, by providing image information as an illustration, it is possible to intuitively convey information (color, shape, etc.) that is difficult to transmit only by text.

In the above specific example 4, an image is added to the summary information. However, when audio information is included in the input document, the audio information is added to the abstract in the same manner as the image information. You may make it.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing a specific example 1 of a document summarizing apparatus of the present invention.

FIG. 2 is a flowchart illustrating an operation of the document summarizing apparatus according to the first embodiment.

FIG. 3 is an explanatory diagram illustrating an example of input information.

FIG. 4 is an explanatory diagram of a character string pattern stored in a character string pattern database.

FIG. 5 is an explanatory diagram illustrating an example of document information in which attention information is marked.

FIG. 6 is an explanatory diagram showing an example of document information in which reference information is marked.

FIG. 7 is an explanatory diagram illustrating an example of document information in which attribute information is marked.

FIG. 8 is an explanatory diagram showing an example of a syntax pattern in a syntax pattern database 4.

FIG. 9 is an explanatory diagram showing an example of other document information marked with attribute information.

FIG. 10 is an explanatory diagram showing an example of a syntax pattern.

FIG. 11 is an explanatory diagram of other document information marked with attribute information.

FIG. 12 is an explanatory diagram of document information in which attribute information is marked.

FIG. 13 is an explanatory diagram illustrating an example of summary information.

FIG. 14 is a configuration diagram of a document summarization device of a specific example 2.

FIG. 15 is a flowchart illustrating an operation of a specific example 2;

FIG. 16 is an explanatory diagram of summary information customized as a personal name list.

FIG. 17 is an explanatory diagram of summary information customized as a product catalog.

FIG. 18 is a configuration diagram of a document summarizing apparatus according to a third embodiment.

FIG. 19 is a flowchart (part 1) illustrating the operation of the third embodiment.

FIG. 20 is a flowchart (part 2) illustrating the operation of the third embodiment;

FIG. 21 is an explanatory diagram illustrating an example of an input document.

FIG. 22 is an explanatory diagram illustrating an example of a tag pattern.

FIG. 23 is an explanatory diagram showing an example of tagged document information in which tag information is marked.

FIG. 24 is a configuration diagram of a document summarizing apparatus according to a fourth embodiment.

FIG. 25 is a flowchart (part 1) illustrating the operation of the specific example 4;

FIG. 26 is a flowchart (part 2) illustrating the operation of the specific example 4;

FIG. 27 is an explanatory diagram showing an example of tagged document information.

FIG. 28 is an explanatory diagram illustrating an example of image information stored in the image information storage unit 17;

FIG. 29 is an explanatory diagram of summary information customized as a personal name list.

FIG. 30 is an explanatory diagram of summary information customized as a product catalog.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Document information input part 2 Attention information determination means 5 Text information storage part 6 Reference information determination means 9 Attribute information determination means 10 Relevance determination means 11 Abstract sentence generation means 13, 13a, 13b Customization means 15 Tag information extraction means 16 Image information Extraction means 17 Image information storage

 ────────────────────────────────────────────────── ─── Continuing on the front page (72) Inventor Mitsuo Shimohata 1-7-12 Toranomon, Minato-ku, Tokyo Oki Electric Industry Co., Ltd. F-term (reference) 5B075 ND03 NK02 NS01

Claims (4)

[Claims] 1. An attention information determining unit for extracting attention information from a character string in an input document based on a predetermined rule, and a case where there is a character string in a reference relationship with the attention information in the input document. Reference information determining means for providing reference information indicating this to the input document; and when there is a character string related to the attention information in the input document, providing relevant information indicating this to the input document. A document summarizing apparatus comprising: a relevance determining unit; and a summary sentence generating unit that generates a summary sentence of the input document based on the attention information, the reference information, and the related information. 2. The document summarizing apparatus according to claim 1, wherein when an arbitrary instruction is received from a user, the input document is reflected by reflecting the instruction of the user based on attention information, reference information, and related information. A document summarizing apparatus comprising: a customizing unit that generates an abstract. 3. The document summarizing apparatus according to claim 1, wherein it is determined whether tag information is added to the input document. If the tag information is added, the type and attribute of the tag information are determined. Tag information extraction means for extracting the input document; a summary sentence generation means for generating a summary of the input document based on the type and attribute of the tag information extracted by the tag information extraction means; and attention information; Is received, based on attention information, reference information, related information, and the type and attribute of the tag information,
A document summarizing device for generating a summary of the input document by reflecting the instruction of the user. 4. The document summarizing apparatus according to claim 3, wherein it is determined whether or not there is image information to which tag information is added in the input document, and if there is such image information, the image information to be extracted is extracted. Extracting means for generating a summary of the input document based on the type and attribute of the tag information extracted by the tag information extracting means and the attention information; and when the image information is extracted by the image information extracting means, A summary sentence generating means for adding the image, and when an arbitrary instruction from the user is received, based on attention information, reference information, related information, and the type and attribute of the tag information,
A document summarizing means for generating an abstract of the input document by reflecting the instruction of the user, and customizing means for adding the image when the image information is extracted by the image information extracting means. apparatus.