JPH10214268A - Method and device for retrieving document - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce the required capacity of storage device by preparing the index of registered document from a summarized sentence automatically generated from this document and retrieving the document corresponding to a desired retrieval request based on this index. SOLUTION: A document retrieval system is composed of an input/output part 10, automatic document summarizing part 20, indexing part 30, document compressing/extending part 40, document retrieval part 50, document storage part 60, index storage part 70, document managing file storage part 80 and control part 90. Then, the summarized sentence of registered document is automatically generated from this document, the index of document is prepared from this automatically generated summarized sentence and based on this index, the document corresponding to the desired retrieval request is retrieved. Besides, when the number of times of access to the registered document in the past exceeds a specified value within a fixed period, the index of document is prepared from the source sentence of this registered document and based on this index, the document corresponding to the desired retrieval request is retrieved.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document retrieval method and apparatus for retrieving a document using a document summary, and more particularly, to reducing the required capacity of a storage device for storing registered documents and the like while suppressing the influence on the document retrieval accuracy. The present invention relates to a document search method and apparatus capable of greatly reducing the number of search information and providing more search information to a user.

[0002]

2. Description of the Related Art Heretofore, there have been known document search methods and apparatuses employing various methods as a document search method and apparatus for searching a desired document from a plurality of registered documents in response to a desired search request.

For example, one for setting an index corresponding to each document and searching for a desired document based on the index, another for creating a summary sentence corresponding to each document and searching for a desired document based on the summary sentence Etc. are known.

[0004] In addition, as a method using an index, there is a method using a character component type index, a method using automatic keyword extraction by morphological analysis, and the like.

[0005]

However, in such a conventional document search method and apparatus, a storage device for storing registered documents and the like requires a large storage capacity, and in particular, a storage device for storing registered documents and the like requires a large amount of storage space. However, there is a problem that the storage capacity for storing the index becomes large in the case where the method of searching for the document is adopted.

For example, when a character component type index is used, full-text search is possible, but the size of the index file is 1.5 times the size of the original document.

[0007] Also, in a full-text search by automatic keyword extraction by morphological analysis, the file size is almost equal to that of the original document.

Accordingly, the present invention is capable of significantly reducing the required capacity of a storage device for storing registered documents and the like while suppressing the influence on document search accuracy, and providing more search information to a user. It is an object of the present invention to provide a document search method and apparatus which can be used.

[0009]

According to a first aspect of the present invention, there is provided a document search method for searching a document corresponding to a desired search request from a plurality of registered documents. A summary of the document is automatically generated from the document, an index of the document is created from the automatically generated summary, and a document corresponding to the desired search request is searched based on the index.

According to a second aspect of the present invention, in the first aspect of the invention, if the number of past accesses of the registered document exceeds a prescribed value within a predetermined period, the registered document An index of the document is created from the original text of the document, and a document corresponding to the desired search request is searched based on the index.

[0011] In the invention according to claim 3, in the invention according to claim 1 or 2, when a word included in the abstract includes a synonym or abbreviated word having a smaller number of words than the word, the word is replaced with the word. It is characterized in that the above-mentioned summary sentence is reduced in size by replacing it with a synonym or a shortened word.

Further, in the invention according to claim 4, in the invention according to claim 3, when there are a plurality of words having synonyms or abbreviated words having a smaller number of words than the words included in the summary sentence, the words are replaced with the words. It is characterized in that the word that appears first in the abstract is replaced with the above-mentioned synonym or shortened word.

According to a fifth aspect of the present invention, in the document search method for searching for a document corresponding to a desired search request from a plurality of registered documents, the number of past accesses of the registered document is within a predetermined period. If the value exceeds the specified value, a first index of the document is created from the original text of the registered document, and a document corresponding to the desired search request is searched based on the first index, If the number of past accesses of the registered document is equal to or less than a prescribed value within a certain period, a summary sentence of the document is automatically generated from the registered document, and the document summary is generated from the automatically generated summary sentence. A second index is created, and a document corresponding to the desired search request is searched based on the second index.

According to a sixth aspect of the present invention, there is provided a document search method for searching a plurality of registered documents for a document corresponding to a desired search request, wherein the registered document is compressed to automatically generate a compressed document. The compressed document is searched in response to the desired search request, and the searched compressed document is expanded and output.

According to a seventh aspect of the present invention, in the sixth aspect of the present invention, if the number of past accesses of the registered document exceeds a prescribed value within a predetermined period, the registered document is And outputs the retrieved document original text. If the past access count of the registered document is equal to or less than a prescribed value within a certain period, the compressed document is retrieved and the retrieved compressed The document is decompressed and output.

The invention according to claim 8 is a document retrieval method for retrieving a document corresponding to a desired retrieval request from a plurality of registered documents, wherein the registered document has been accessed in the past for a first period. If the value exceeds the first prescribed value, the first document of the document
Is created, the original text of the registered document is searched based on the first index, and the searched document original text is output. The number of past accesses of the registered document is within the first period. If the value is equal to or less than the first prescribed value, a first index of the registered document is created from the original text of the registered document, and the registered document is compressed to automatically generate a compressed document. 1, the compressed document is retrieved based on the index, and the retrieved compressed document is decompressed and output. The registered document is accessed in the second period within the second period longer than the first period. In the following cases, a summary sentence of the document is automatically generated from the registered document, and a second index of the document is created from the automatically generated summary sentence based on the second index. When the contracted document is retrieved and the retrieved compressed document is decompressed and output, and the past access count of the registered document is zero within a third period longer than the second period, the registered document is registered. The document is deleted and only the summary sentence is retained.

According to a ninth aspect of the present invention, in the invention of the eighth aspect, when a word included in the summary sentence includes a synonym or abbreviated word having a smaller number of words than the word, the word is included in the summary sentence. The present invention is characterized by reducing the size of the summary sentence by replacing a word having a synonym or abbreviated word having a small number of words with the synonym or abbreviated word.

The invention of claim 10 is the invention of claim 9 in which, when there are a plurality of words having synonyms or abbreviated words having a smaller number of words than the words included in the summary sentence, the words are replaced with the words. It is characterized in that the word that appears first in the abstract is replaced with the above-mentioned synonym or shortened word.

The invention according to claim 11 is a document retrieval apparatus for retrieving a document corresponding to a desired retrieval request from a plurality of registered documents, wherein the document storage means for storing the registered document; Search request input means for inputting a request, automatic document summarization means for automatically generating a summary sentence of the document from the document stored in the document storage means, and the above-mentioned document from the summary sentence generated by the automatic document summarization means. Index creating means for creating an index, an index storing means for storing the index created by the index creating means, and a document stored in the document storing means based on the index stored in the index storing means. Document search means for searching for a document corresponding to the desired search request.

According to a twelfth aspect of the present invention, in the invention of the eleventh aspect, when a word included in the summary includes a synonym or abbreviated word having a smaller number of words than the word,
The present invention is further characterized by further comprising a summary sentence reducing means for reducing the size of the summary sentence by replacing the word with the synonym or the abbreviation.

According to a thirteenth aspect of the present invention, in the twelfth aspect of the invention, the abstract sentence miniaturizing means includes a plurality of synonyms or abbreviated words having a smaller number of words than the words included in the abstract sentence. In this case, the word is replaced with the above-mentioned synonym or abbreviated word except for the word that appears first in the summary sentence.

According to a fourteenth aspect of the present invention, there is provided a document retrieval apparatus for retrieving a document corresponding to a desired retrieval request from a plurality of registered documents, wherein the document storage means for storing the registered document; Input / output means for inputting a request and outputting a search result corresponding to the search request; and automatically generating a summary sentence of the document from the original document stored in the document storage means and storing the summary in the sentence storage means. Automatic document summarization means, index creation means for creating an index of the document from the document original or summary sentence stored in the document storage means, and index storage means for storing the index created by the index creation means Compressing the document original stored in the document storage means, storing the compressed document in the document storage means as a compressed document, A document compression / decompression unit for decompressing the compressed document stored in the document storage unit, and the document storage unit based on the index stored in the index storage unit in response to the search request input from the input unit. A document search unit for executing a document search process for searching the stored document original text or the compressed document or the summary sentence and outputting the document to the input / output unit; and a management unit for managing the document search process by the document search unit. , Is provided.

According to a fifteenth aspect of the present invention, in the invention of the fourteenth aspect, when a word included in the summary includes a synonym or abbreviated word having a smaller number of words than the word,
The present invention is further characterized by further comprising a summary sentence reducing means for reducing the size of the summary sentence by replacing the word with the synonym or the abbreviation.

According to a sixteenth aspect of the present invention, in the invention of the fifteenth aspect, the abstract sentence miniaturizing means includes a plurality of words including synonyms or abbreviated words having a smaller number of words than the words included in the abstract sentence. In this case, the word is replaced with the above-mentioned synonym or abbreviated word except for the word that appears first in the summary sentence.

According to a seventeenth aspect of the present invention, in the fourteenth aspect of the present invention, the search means determines that a past access count of the document stored in the document storage means is within a first period within a first period. If the value exceeds the value, the document original stored in the document storage unit is searched based on the first index created from the document original stored in the index storage unit, and the searched document original is output. If the past access count of the document stored in the document storage means is equal to or less than the first prescribed value within the first period, the compressed document is searched based on the first index. If the searched compressed document is expanded and output, and the number of past accesses of the document stored in the document storage means is equal to or less than a second rule within a second period longer than the first period,
The compressed document is searched based on the second index created from the summary sentence stored in the index storage means, and the searched compressed document is decompressed and output. When the number of times of access in the past is zero within a third period longer than the second period, the document stored in the document storage unit is deleted and only the summary sentence is retained.

[0026]

Embodiments of the present invention will be described below in detail with reference to the accompanying drawings.

FIG. 1 is a schematic block diagram showing an embodiment of a document search system configured by applying the document search method and apparatus according to the present invention.

Referring to FIG. 1, the document search system includes:
An input / output unit 10 for performing document registration and document search requests and displaying a document search result, an automatic document summarization unit 20 for automatically generating an abstract from a registered document, and an original text of a registered document (hereinafter referred to as a registered original text) or an abstract sentence. An indexing unit 30 for creating an index of the document, a document compression / decompression unit 40 for compressing or decompressing the registered original text, a document search unit 50 for executing a desired search request from the document index,
A document storage unit 60 for storing a registered document or its digest or compressed document, an index storage unit 70 for storing an index of the registered document or digest, a document management file storage unit 80 for storing a document management file, A control unit 90 is provided to control the overall operation.

In this document search system, first, at the time of document registration, the control unit 90 detects a document registration request from the input / output unit 10, whereby the control unit 90 causes the input / output unit 1 to register.
0 instructs the designated document (registered document) to be stored in the document storage unit 60.

At the same time, the control unit 90 instructs the indexing unit 30 to execute indexing for creating an index of the document.

The indexing unit 30 includes the control unit 90
An index of the document is generated in accordance with an instruction to execute indexing from, and the generated index is stored in the index storage unit 70.

At this time, the control unit 90 creates a new entry in the document management file storage unit 80 and records the registration date.

At the time of document search, the control unit 90 detects a document search request from the input / output unit 10, and thereby, the control unit 90
Instructs the document search unit 50 to search for a document corresponding to the document search request.

Thus, the document search unit 50 performs a search process on the index storage unit 70 and the document storage unit 60 according to a keyword or a natural language sentence serving as a search command.

When the search processing by the document search section 50 is completed, the control section 90 instructs the input / output section 10 to display a search result list indicating the search results.

As a result, the search result by the document search unit 50 is output to the input / output unit 10, and a search result list is displayed on the input / output unit 10.

When the user finds a desired document from the search result list displayed on the input / output unit 10, the user issues a request to open the desired document from the input / output unit 10. Is detected by the control unit 90, whereby the control unit 90 instructs the document search unit 50 to output the document to the input / output unit 10.

Here, when the desired document is stored in its original state in the document storage unit 60, the document search unit 50 controls the document stored in the document storage unit 60 under the control of the control unit 90. Is output to the input / output unit 10 as it is,
The input / output unit 10 displays the document.

When the desired document is compressed and stored in the document storage unit 60, the document search unit 50
Under the control of the control unit 90, the document stored in the document storage unit 60 is decompressed by the document compression / decompression unit 40, and then output to the input / output unit 10, and the input / output unit 10 displays the document. .

At this time, the document search unit 50 updates the document management file of the opened document stored in the document management file storage unit 80, and records the access date and time of the document in the document management file. .

By the way, in the document search system according to this embodiment, in order to reduce the size of the registered document stored in the document storage unit 60 and the index stored in the index storage unit 70, for example, on Friday, Afternoon 1
At 0 o'clock or the like, the batch processing is performed to reduce the size of the registered document stored in the document storage unit 60 and the index stored in the index storage unit 70.

Here, the document management batch process for reducing the size of the registered document stored in the document storage unit 60 and the process of reducing the size of the index stored in the index storage unit 70 is based on the number of accesses to the document. Done.

That is, as shown in FIG. 2, in this batch processing, if six months have passed since the registration of the document, and if the number of accesses to the document has been two or less in the past six months, the registered document is deleted. The compressed document is compressed and stored in the document storage unit 60, and the index storage unit 70 stores an index based on the registered original text before compression.

If one year has passed since the registration of the document, and if the number of accesses to the document has been one or less in the past one year, the document storage unit 60 stores a compressed document obtained by compressing the registered document. The index storage unit 7 is stored.
0 stores an index based on the summary sentence of the document.

If two years have passed since the registration of the document, and the number of accesses to the document has been zero in the past two years, the compressed document of the document is also deleted. In this case, the index stored in the index storage unit 70 is also deleted.

Note that the number of accesses used in the document management batch processing is counted as one access for registration of the document.

The document management batch process is performed based on a document management file stored in the document management file storage unit 80.

FIG. 3 shows the document management file storage section 80.
1 shows an example of a document management file stored in the file management server.

This document management file is, for example, 1996
This is a document management file stored in the document management file storage unit 80 as of January 31, and in the document management file, the history of the last three document opens for each document is recorded. Here, document registration is also recorded as one access.

In FIG. 3, "FLAG" indicates the state of each document stored in the document storage unit 60.

That is, when “FLAG” is “N”, the document stored in the document storage unit 60 is an uncompressed document, and the index stored in the index storage unit 70 is a normal index whose source is a registered original text. It is shown that.

When the "FLAG" is "C", the document stored in the document storage unit 60 is a compressed document, and the index stored in the index storage unit 70 is a normal index having a registered original as a creation source. Indicates that there is.

When the “FLAG” is “S”, the document stored in the document storage unit 60 is a compressed document, and the index stored in the index storage unit 70 is a summary sentence index whose source is a summary sentence. It is shown that.

When "FLAG" is "X", it indicates that the document has been deleted.

"DN" is the document number of each document,
“LA1” to “LA3” indicate the past three access dates.

FIG. 4 to FIG. 6 are flowcharts showing the document management batch processing.

4 to 6, when the document management batch process starts (step 101), an initialization process is first performed (step 102). In this initialization processing, 1) the document number “DN” to be processed is set to “0” (DN = 0) 2) The date six months ago is set as “DATE0” (D
ATE0 = date six months ago) 3) Set date one year ago as “DATE1” (D
ATE1 = date one year ago) 4) Set date two years ago as "DATE2" (D
ATE2 = date two years ago).

Next, the document information of the document number “DN”, ie, “FLAG”,
“LA1”, “LA2”, and “LA3” are acquired (step 103).

Then, it is checked whether "FLAG" is "X", that is, "FLAG == X" (step 10).
4).

Here, if "FLAG == X" (YES in step 104), the document number "DN" is incremented by "1", "DN = DN + 1" (step 108), and then the document management file Is the last one, "F
ILE END ”(step 109),“ FILE E
ND "(NO in step 109), step 1
Return to 03.

In step 109, "FILE E
ND ”(YE at step 109)
S), the document management batch process ends (step 1).
10).

In step 104, “FLAG ==
X ”(YE at step 104)
S) Then, it is checked whether “LA1” acquired in step 103 is older than “DATE2” (step 10).
5).

If it is determined that "LA1" is older than "DATE2" (YES in step 105),
The document with the document number DN is deleted (step 106), the "FLAG" of the document number "DN" is updated to "X" (step 107), and the process proceeds to step 108.

If it is determined in step 105 that "LA1" is newer than "DATE2" (NO in step 105), the process proceeds to step 111 in FIG.

In step 111, it is checked whether or not there is data in "LA2". Here, if it is determined that data exists in “LA2” (YES in step 111), then “L2”
It is checked whether "A2" is older than "DATE1" (step 112).

If it is determined that "LA2" is older than "DATE1" (YES in step 112), then "FLAG" is "S", that is, "FLAG =
= S ”(step 113). Here, "FL
If it is determined that AG == S ”(YES in step 113), the process proceeds to step 108 in FIG.

In step 113, "FLAG ==
S ”(NO in step 113),
Next, “FLAG” is “C”, that is, “FLAG”
== C ”(step 114).

If it is determined that "FLAG == C" is not satisfied (NO in step 114), the document number "D
The document with the document number "DN" is read from the document storage unit 60, the document with the document number "DN" is compressed by the document compression / decompression unit 40, and the compressed document with the document number "DN" stored in the document storage unit 60 is read. The document is compressed and stored for updating (step 115).

The new index is created based on the summary sentence corresponding to the document with the document number “DN”, and the newly created index is stored in the index storage unit 70 with the document with the document number “DN”. The corresponding stored index is updated (step 116).
Thereafter, the "FLAG" of the document number "DN" is updated to "S" (step 117), and the process proceeds to step 108 in FIG.

In step 114, “FLAG ==
C ”(YE at step 114).
S) The process proceeds to step 116 without performing the process of step 115.

If it is determined in step 111 that there is no data in "LA2" (N in step 111
O) Then, it is checked whether "LA1" is older than "DATE1" (step 118). Here, "LA1" is "DA1".
If it is determined that the date is older than “TE1” (Y in step 118).
ES), the process proceeds to a step 113, but “LA1” is changed to “DA
If it is determined that it is newer than “TE1” (NO in step 118), the process proceeds to step 126 in FIG.

If it is determined in step 112 that "LA2" is newer than "DATE1" (step 1)
(NO at 12), proceed to step 119 in FIG.

In step 119, it is checked whether or not there is data in "LA3". If it is determined that there is data in “LA3” (YES in step 119), then “L3”
It is checked whether "A3" is older than "DATE0" (step 120).

If it is determined that "LA3" is older than "DATE0" (YES in step 120), then "FLAG" is "C", that is, "FLAG =
= C ”(step 121). Here, "FL
When it is determined that AG == C ”(YES in step 121), the process proceeds to step 108 in FIG.

In step 121, “FLAG ==
C ”(NO in step 121),
Next, “FLAG” is “N”, that is, “FLAG”
== N ”(step 122).

If it is determined that "FLAG == N" is not satisfied (NO in step 122), the document number "D
N ”is created from the text (step 12).
3) After that, "FLAG" of the document number "DN" is updated to "C" (step 125), and the process proceeds to step 108 of FIG.

Also, in step 122, "FLAG ==
N ”(YE in step 122).
S) The document with the document number "DN" is compressed and stored in the document storage unit 60 (step 124), and the process proceeds to step 125 without performing the process of step 123.
9, if it is determined that there is no data in "LA2" (NO in step 119), then "LA2"
It is checked whether it is older than “TE0” (step 127). If it is determined that “LA2” is older than “DATE0” (YES in step 127), the process proceeds to step 121, but “LA2” is older than “DATE0”. If it is determined that it is new (NO in step 127), the process proceeds to step 128.

If it is determined in step 120 that "LA3" is newer than "DATE0" (NO in step 120), the flow advances to step 128.

In step 126, it is checked whether "LA1" is older than "DATE0". Here, "LA
If it is determined that “1” is older than “DATE0” (YES in step 126), the process proceeds to step 121.

If it is determined in step 126 that "LA1" is newer than "DATE0" (step 1)
Then, it is checked whether "FLAG" is "N", that is, "FLAG == N" (step 12).
8). Here, if it is determined that “FLAG == N” (YES in step 128), step 108 in FIG.
Proceed to.

In step 128, “FLAG ==
N ”(NO in step 128),
Next, “FLAG” is “C”, that is, “FLAG”
== C ”(step 129).

If it is determined that "FLAG == C" is not satisfied (NO in step 129), the document number "D
N ”is created from the text (step 13
0), decompresses and saves the document with the document number "DN" (step 131), and then updates "FLAG" of the document number "DN" with "N" (step 132), and proceeds to step 108 in FIG. .

Also, at step 129, "FLAG ==
C ”(YE in step 129).
S), the process proceeds to step 131 without performing the process of step 130.

FIG. 7 shows an example of a document management file stored in the document management file storage unit 80, for example, on November 1, 1996 after the above-described document management batch processing has been performed. It is a thing.

As is clear from comparison with the document management file shown in FIG. 3, the document number “DN” is changed to “0000”.
Since the document “0002” has been accessed 0 times in the past two years, “FLAG” has become “X”, indicating that it has been deleted, and the document number “DN” has been changed to “000000”.
Since the document “03” has been accessed twice in the past six months, “FLAG” becomes “C” and the document storage unit 60
The document stored in the index storage unit 70 is a compressed document.
Indicates that the index is a normal index created from the registered original text, and the document whose document number “DN” is “0000043” is
Since the number of accesses per year is one, “FLAG” becomes “S”, the document stored in the document storage unit 60 is a compressed document, and the index stored in the index storage unit 70 is And the document number “DN” is “0001”.
Since the document “2014” has been accessed twice in the past six months, “FLAG” is changed to “C”, and the document stored in the document storage unit 60 is stored in the compressed document and index storage unit 70. The index indicates that it is a normal index whose origin is the registered original text.

FIG. 8 shows the automatic document summarizing section 20 shown in FIG.
This is an example of the configuration.

In FIG. 8, the automatic document summarizing section 20
Are a document structure analysis unit 21, a morphological analysis unit 22, a syntax analysis unit 23, an important sentence determination unit 24, a synonym / abbreviated word replacement unit 2
5. It comprises a morphological analysis dictionary 26 and a synonym / abbreviated word dictionary 27.

Here, the document structure analysis unit 21 analyzes the digitized input document and analyzes the structure of titles, chapter and section headings, paragraphs, sentences, and the like.

The morphological analysis unit 22 uses the morphological analysis dictionary 26 to divide the title, body, and the like into words based on part-of-speech connection rules, cuts out phrases, and identifies the parts of speech of each word.

In the parsing unit 23, the nominative character ("wa"
Cases, "ga" cases, etc.) are analyzed, and the structure of the clauses in each sentence is analyzed.

The important sentence judging section 24 finds each sentence included in the sentence from the results of the document structure analysis, the morphological analysis, and the syntax analysis, as a title, a chapter heading, a paragraph position relative to the entire sentence, and a sentence in the paragraph. Location, extraction and importance of important words in the entire sentence, importance of important words included in the sentence, number of important words included in the sentence, usage of the nominative and objective cases of important words in the sentence, etc. Then, the importance of each sentence is determined, and a sentence to be an abstract sentence is determined based on the summarization rate of sentences and the state of anaphor between sentences.

Note that, out of the automatic document summarizing section 20, the important sentence judging section 24 does not constitute the gist of the present invention, but can be realized by using a technique already proposed in a document summarizing system or the like. For example, a context analysis and a semantic analysis can be included in addition to the above configuration.

The synonym / abbreviated word replacement unit 25 searches the head sentence for the summary sentence obtained by the important sentence determination unit 24 for words described in the synonym / abbreviated word dictionary 27, If there is a synonym or abbreviated word with fewer characters than the summary sentence obtained in step 24, and two or more identical words are searched, the words after two are replaced with synonyms or abbreviated words with fewer characters. I do.

For example, as an example in which a shortened word dictionary as shown in FIG. 9 is used, a synonym / abbreviated word replacement unit when the summary sentence obtained by the important sentence determination unit 24 is as shown in FIG. FIG. 11 shows an example of the reduced summary sentence replaced by 25.

That is, in the abbreviated summary sentence shown in FIG. 11, the word "Kansai International Airport" is replaced with the abbreviation "Kanku International Airport" from the second appearance, and the word "New Tokyo International Airport" is From the second appearance, it is replaced with the abbreviation "Narita", and as a result, the summary sentence shown in FIG. 10 is shortened as shown in FIG.

When the first word is detected, the following abbreviations can be added, for example, in parentheses, or the first word can be replaced with an abbreviation.
When the first word is detected, if the following abbreviations are additionally described with, for example, parentheses, the readability of the abbreviated summary sentence can be improved.

In the synonym / abbreviated word substitution processing performed by the synonym / abbreviated word substitution unit 25, all the words described in the synonym / abbreviated word dictionary 27 are also registered in the morphological analysis dictionary 26. I do.

Using the result of morphological analysis by the morphological analysis unit 22 performed at the time of document summarization by the automatic document summarization unit 20, when the independent words of each clause are parts of speech (such as nouns or proper nouns) that can be synonyms or shortened words Is detected whether the word is registered in the synonym / abbreviated word dictionary 27.

Each time the word is detected, a flag indicating whether or not the word appears for the first time is prepared. Then, when this word is detected, the flag is checked, and words appearing for the second time or later are replaced with words having a smaller number of words registered in the synonym / abbreviated word dictionary 27.

If the words described in the synonym / abbreviated word dictionary 27 are not included in the morphological analysis dictionary 26, the part of the part of speech (noun or proper noun) that can be a synonym / abbreviated word is Suppose it could be one word,
It is possible to cope by combining and inspecting continuous ones.

The document compression / decompression unit 40 shown in FIG.
The indexing unit 30 and the document search unit 50 do not constitute the gist of the present invention, but can be realized by using a technique already implemented in this type of document search system or file compression / decompression system.

The index creation method in the indexing unit 30 includes a character component method, a keyword automatic extraction method by morphological analysis, and the like. In the present invention, any method that employs any method can be used.

In a full-text search system, the index size is usually about 100% to 150% of the registered document size, though it differs depending on the method. When compressed, the original text can be compressed to a size of 10%.

It is said that the abstract sentence can roughly grasp the subject of the original sentence in an amount of about 10% to 20% of the original sentence.

Therefore, for example, when a 10% digest is performed for a registered document of 100 Mbytes, the registered document size is 100 Mbytes in the initial state according to the above embodiment.
e, index size is 100 Mbytes, total 20
In the original text compression state, the registered document size is 10 Mbytes and the index size is 10 Mbytes.
The document amount can be reduced to 55% of 0 Mbytes, a total of 110 Mbytes, and in a state using an index based on a summary sentence, the registered document size is 10 Mbytes, the index size is 10 Mbytes, and a total of
% Document size can be reduced.

In the above embodiment, the document search system for Japanese is shown. However, the present invention can of course be applied to a document search system for other languages as well. is there.

[0107]

As described above, according to the present invention,
A summary of the document is automatically generated from the registered document, an index of the document is created from the automatically generated summary, and a document corresponding to the desired search request is searched based on the index. Therefore, it is possible to significantly reduce the required capacity of the storage device for storing the registered documents and the like while suppressing the influence on the document search accuracy, and to provide more search information to the user.

[Brief description of the drawings]

FIG. 1 is a schematic block diagram showing an embodiment of a document search system configured by applying a document search method and apparatus according to the present invention.

FIG. 2 is a view for explaining an outline of a document management batch processing transformer for performing a process of reducing the size of a registered document stored in a document storage unit and a process of reducing the size of an index stored in an index storage unit shown in FIG. 1;

FIG. 3 is a diagram showing an example of a document management file stored in a document management file storage unit shown in FIG.

FIG. 4 is a flowchart illustrating details of a document management batch process described in FIG. 2;

FIG. 5 is a flowchart illustrating details of a document management batch process described in FIG. 2;

FIG. 6 is a flowchart illustrating details of a document management batch process described in FIG. 2;

FIG. 7 is a diagram showing an example of a document management file stored in a document management file storage unit after the document management batch processing of FIGS. 4 to 6 has been performed;

FIG. 8 is a block diagram illustrating a configuration example of an automatic document summarizing unit illustrated in FIG. 1;

FIG. 9 is a diagram showing an example of a contracted word dictionary used in a synonymous / abbreviated word replacing unit of the automatic document summarizing unit shown in FIG. 8;

FIG. 10 is a view showing an example of a summary sentence obtained by an important sentence determination unit of the automatic document summarization unit shown in FIG. 8;

11 is a diagram showing an example of how the summary sentence shown in FIG. 10 is replaced by a synonym / abbreviated word replacement unit of the automatic document summary unit shown in FIG. 8;

[Explanation of symbols]

 Reference Signs List 10 input / output unit 20 automatic document summarization unit 30 indexing unit 40 document compression / decompression unit 50 document search unit 60 document storage unit 70 index storage unit 80 document management file storage unit 90 control unit 21 document structure analysis unit 22 morphological analysis unit 23 syntax analysis Section 24 important sentence determination section 25 synonym / abbreviated word replacement section 26 morphological analysis dictionary 27 synonym / abbreviated word dictionary

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Shinichi Mukakawa 10 Odron-cho, Hanazono-cho, Ukyo-ku, Kyoto-shi, Kyoto

Claims (17)

[Claims] 1. A document retrieval method for retrieving a document corresponding to a desired retrieval request from a plurality of registered documents, wherein a summary of the document is automatically generated from the registered document, and the automatically generated summary is generated. A document index corresponding to the desired search request based on the index. 2. If the number of past accesses of the registered document exceeds a prescribed value within a predetermined period, an index of the document is created from the original text of the registered document, and the index is stored in the index. 2. The document search method according to claim 1, wherein a document corresponding to the desired search request is searched based on the search request. 3. When a synonym or abbreviated word having a smaller number of words than the word included in the summary sentence exists,
3. The document search method according to claim 1, wherein the abstract is reduced in size by replacing the word with the synonym or abbreviation. 4. When there are a plurality of words having a synonym or abbreviated word having a smaller number of words than the words included in the summary sentence, the word is replaced with the synonym or the abbreviation except for the first word appearing in the summary sentence. 4. The document search method according to claim 3, wherein replacement is performed with a shortened word. 5. A document retrieval method for retrieving a document corresponding to a desired retrieval request from a plurality of registered documents, wherein a past access count of the registered document exceeds a prescribed value within a predetermined period. If there is, the first index of the document is created from the original text of the registered document, and a document corresponding to the desired search request is searched based on the first index. If the number of accesses of the document is equal to or less than a prescribed value within a certain period, a summary of the document is automatically generated from the registered document, and a second index of the document is created from the automatically generated summary. A document corresponding to the desired search request based on the second index. 6. A document retrieval method for retrieving a document corresponding to a desired retrieval request from a plurality of registered documents, wherein the registered document is compressed to automatically generate a compressed document. A document search method characterized by correspondingly searching the compressed document, and expanding and outputting the searched compressed document. 7. If the number of past accesses of the registered document exceeds a prescribed value within a certain period, search the original of the registered document and output the searched document original. If the number of past accesses of the registered document is equal to or less than a prescribed value within a certain period, the compressed document is searched, and the searched compressed document is expanded and output. 6. The document search method described in 6. 8. A document retrieval method for retrieving a document corresponding to a desired retrieval request from a plurality of registered documents, wherein a past access count of the registered document is a first prescribed number within a first period. If the value exceeds the value, a first index of the document is created from the original text of the registered document, and the original text of the registered document is searched based on the first index. If the number of past accesses of the registered document is equal to or less than the first prescribed value within the first period, the first index of the document is obtained from the original text of the registered document. Creating and automatically generating a compressed document by compressing the registered document; searching the compressed document based on the first index; expanding and outputting the searched compressed document; Past Is less than or equal to the second rule within a second period longer than the first period,
A summary of the document is automatically generated from the registered document, a second index of the document is created from the automatically generated summary, and the compressed document is searched based on the second index. If the number of past accesses of the registered document is zero within a third period longer than the second period, the registered document is deleted and the summary is deleted. A document search method characterized by retaining only sentences. 9. When a word included in the summary sentence includes a synonym or abbreviated word having a smaller number of words than the word,
9. The document retrieval method according to claim 8, wherein the abstract sentence is miniaturized by replacing a word having a small number of synonyms or abbreviated words included in the abstract sentence with the synonym or abbreviated word. Method. 10. When there are a plurality of words having a synonym or abbreviated word having a smaller number of words than the words included in the summary sentence, the word is replaced with the synonym or the abbreviation except for the first word appearing in the summary sentence. 10. The document search method according to claim 9, wherein replacement is performed with a shortened word. 11. A document retrieval apparatus for retrieving a document corresponding to a desired retrieval request from a plurality of registered documents, a document storage means for storing the registered document, and a retrieval request input for inputting the retrieval request Means, automatic document summarization means for automatically generating a summary sentence of the document from the document stored in the document storage means, and index creation for creating an index of the document from the summary sentence generated by the automatic document summarization means Means; an index storage means for storing an index created by the index creation means; and a response to the desired search request from documents stored in the document storage means based on the index stored in the index storage means. A document search device comprising: a document search unit that searches for a document to be searched. 12. When a synonym or abbreviated word having a smaller number of words than the word included in the abstract exists, the word is replaced with the synonym or abbreviated word to reduce the size of the abstract. 12. The document search apparatus according to claim 11, further comprising: 13. The summary sentence miniaturizing means, wherein, when there are a plurality of words having a synonym or abbreviated word having a smaller number of words than the words included in the summary sentence, the word appears first in the summary sentence 13. The document search apparatus according to claim 12, wherein a word is replaced with the synonym or abbreviation. 14. A document search device for searching a document corresponding to a desired search request from a plurality of registered documents, a document storage means for storing the registered document, and inputting the search request and performing the search. An input / output unit that outputs a search result corresponding to the request; an automatic document summarization unit that automatically generates a summary sentence of the document from the original document stored in the document storage unit and stores the summary sentence in the sentence storage unit; Index creation means for creating an index of the document from the document original or the summary sentence stored in the document storage means; index storage means for storing the index created by the index creation means; and storage in the document storage means Compressed document original text stored in the document storage means as a compressed document and stored in the document storage means A document compression / decompression unit for decompressing the compressed document, and a document original stored in the document storage unit based on an index stored in the index storage unit in response to the search request input from the input unit. Or a document search unit for executing a document search process for searching for the compressed document or the summary sentence and outputting it to the input / output unit; and a management unit for managing the document search process by the document search unit. A document search device characterized by the following. 15. When a word included in the summary includes a synonym or abbreviated word having a smaller number of words than the word, the word is replaced with the synonym or abbreviated word to reduce the size of the summary sentence. 15. The document search apparatus according to claim 14, further comprising: a summary sentence reducing unit. 16. The summary sentence minimizing means, wherein, when there are a plurality of words having a synonym or abbreviated word having a smaller number of words than the words included in the summary sentence, the word appears first in the summary sentence 16. The document search device according to claim 15, wherein a word is replaced with the synonym or the abbreviation. 17. The method according to claim 17, wherein the past access count of the document stored in the document storage means exceeds a first prescribed value within a first period, and stores the document in the index storage means. A document original stored in the document storage unit is searched based on the first index created from the obtained document original, and the searched document original is output. If the access count is equal to or less than the first prescribed value within the first period, the compressed document is searched based on the first index, and the searched compressed document is expanded and output. If the number of past accesses of the document stored in the storage unit is equal to or less than a second rule within a second period longer than the first period, the document is created from the summary sentence stored in the index storage unit. Second The compressed document is retrieved based on the index of the above, and the retrieved compressed document is decompressed and output. The number of past accesses of the document stored in the document storage means is within a third period longer than the second period. 15. The document search apparatus according to claim 14, wherein when the number is zero, the document stored in the document storage unit is deleted and only the summary sentence is retained.