JPH08115340A - Document retrieval device and generating device for index file used for the same - Google Patents

Abstract

PURPOSE: To reduce omission of retrieval and improve the speed of retrieval using a retrieval key word as to a document retrieval device using index file. CONSTITUTION: A retrieval character string specification part 709 specifies a retrieval character string. A key word expansion part 705 finds the host word of the retrieval character string specified by the retrieval character string specification part. a retrieval part 711 retrieves the index file 708 with the host word found by the key word expansion part 705. When a coincidence is obtained as a result of the retrieval, a retrieval part retrieves only the range of a key word group classified to be the host word by using the retrieval character string. Thus, when there is the host word, the retrieval character string is only compared with the key words in the index file in the limited range belonging to the host word, so the retrieval speed is greatly improved.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention allows a keyword to be registered in an index together with a key when a document is registered.
The present invention relates to a document search device for efficiently searching for documents and an index file creation device for use in the document search device.

[0002]

2. Description of the Related Art As a conventional document retrieval apparatus, an index file is created in which key words are associated with documents at the time of document registration, and when a document is retrieved, the retrieval keyword is the keyword in this index file. There is an index search method that checks whether they match and, if they match, displays the name and appearance position of the document containing the keyword. In this index search method, if the search keyword input by the user is not found in the index file, there is a method of expanding the search keyword using, for example, a thesaurus dictionary, and searching again using the synonym as a search key. There is also a method of expanding a search keyword from the beginning and searching for a plurality of synonyms. However, when the index file is searched with a plurality of expanded search keywords, the search method becomes complicated as the number of search keywords increases, and the search takes time. Further, if the number of expanded search keywords is limited, the search time is shortened, but the search omission may occur.

In order to shorten the search time and eliminate omission of search, for example, in the "database system" described in Japanese Patent Laid-Open No. 177792/1993, the word used in a document when inputting data is the same as the word. Replace with a representative word of a meaningful group. Further, for example, in the devices described in Japanese Patent Application Laid-Open No. 5-89177 and Japanese Patent Application Laid-Open No. 5-89178, individual target names and upper general names are registered in advance as at least two layers. In the device described in Japanese Patent Laid-Open No. 5-89177, when a target name is searched, not only the target name is searched, but also all names belonging to higher general names of this target name can be expanded according to a user's instruction. It is also possible to extend the search range by going back to the higher-ranked generic name. Japanese Patent Laid-Open No. 5-891
In the apparatus described in Japanese Patent Publication No. 78, when a general name is searched, the target name is decomposed into the search.

[0004]

In the above-mentioned prior art in which the word used in the data at the time of inputting into the above-mentioned database is replaced with the representative word, the original data is changed, and accurate data is saved and retrieved. You cannot do it. In addition, since the user indicates whether to replace with a representative word and which category should be used for words belonging to a plurality of categories, it is very difficult to register a large amount of data at one time.

In the above-mentioned prior art which uses a knowledge base having a tree structure of at least two hierarchies of individual target names and superordinate general names, the knowledge base of the tree structure uses a tree structure for a given search word. Although it corresponds to a related word dictionary for expanding to a synonym or a subordinate word, even if this structure is applied to an index file, it does not improve the search speed.

It is an object of the present invention to improve search speed by a search keyword with less omission of search in a document search device using an index file. Another object of the present invention is to easily and automatically create an index file for document retrieval.

[0007]

A document retrieval apparatus according to the present invention comprises a document storage unit (508) for storing a document and an index file storage unit for classifying keywords for the document by upper terms and storing them together with document identification information. (50
7), a search character string designating means (501) for designating a search character string for searching a document, and an upper word extracting means (504, for obtaining a higher word of the search character string designated by the search character string designating means). 505) and a higher-rank word to find a match with a higher-rank word in the index file, and if they match, a search means (506) that searches the range of the keyword group classified into the higher-rank word with the search character string. ) And.

The index file creating apparatus of the present invention further comprises a document input means (102) for reading a document,
Index file storage means (108) for storing an index file in which keywords classified by higher-order words and document identification information are associated with each other, and keyword extraction means (103) for extracting keywords to be used for search from an input document. A high-order word extraction means (105) for obtaining a high-order word of the keyword extracted by the keyword extraction means,
Registration means (10) for classifying the keywords extracted by the keyword extraction means by the high-order words extracted by the high-order word extraction means and registering the high-order words and the keywords in the index file storage means together with the document identification information of the read document.
7) and are provided.

[0009]

In the document retrieval apparatus of the present invention, first, the retrieval character string designating means designates the retrieval character string. The high-order word extraction means obtains a high-order word of the search character string designated by the search character string designation means. The search means searches the index file with the high-order word obtained by the high-order word extraction means. That is,
Search whether the high-order word matches the high-order word in the index file. If they match as a result of the search, the search means next searches only the range of the keyword group classified into the superordinate word with the search character string. In this way, when there is a high-order word, the search character string may be compared with the keywords in the index file in a limited range belonging to the high-order word, so that the search speed is significantly improved. It should be noted that although a search process using broader terms is required more than in the conventional technique, the comparison target of the search character string is narrowed down,
Overall, a significant improvement in search speed is achieved. Also,
When narrowing down the comparison target of the search character strings, the target is narrowed down based on the relation of the high-order word, rather than simply reducing the number as in the conventional technique, and therefore there is little possibility of missing the search.

In the index file creating apparatus of the present invention, the keyword extracting means extracts a keyword to be used for retrieval from the input document. The high-order word extraction means finds a high-order word of the keyword extracted by the keyword extraction means. The registration means classifies the keywords extracted by the keyword extraction means by the upper words extracted by the upper word extraction means, and registers the upper words and the keywords in the index file storage means together with the document identification information of the read document.

[0011]

【Example】

(Embodiment 1) FIG. 1 is a block diagram showing the configuration of an embodiment of an index file creation device for creating an index file when registering a document in the document search device of the present invention. The index file creating apparatus according to this embodiment includes a document storage unit 101 that stores a document, a document input unit 102 that reads a target document to extract an index, and a keyword that is input from the document input unit 102. A keyword extraction unit 103 for extracting,
An analysis dictionary 104 that is referred to for analysis when extracting a keyword, a keyword expansion unit 105 that expands the extracted keyword, a thesaurus dictionary 106 that is referred to when expanding the keyword, and a keyword extraction unit 103. It has an index registration unit 107 for registering the keywords and words expanded by the keyword expansion unit 105 as an index, and a file storage unit 108 for holding an index file.

FIG. 2 shows a flow of operation of the embodiment configured as described above. The document input unit 102 stores the document file in the document storage unit 101 according to a user's instruction, and at the same time, reads the document file into the keyword extraction unit 103 (step S201). The keyword extraction unit 103 sequentially analyzes the read document using a well-known analysis method such as morphological analysis, extracts independent words, and sets them as keywords (step S202). The analysis dictionary 104 includes a list of independent words that can be indexes. The keyword extraction unit 103 uses this analysis dictionary 104 to extract independent words. The extracted independent word is a keyword. Further, the keyword extracting unit 103 accesses the thesaurus dictionary 106 through the keyword expanding unit 105 to obtain a keyword upper word (step S203). The index registration unit 107 classifies the keywords by upper words, and stores the upper words and keywords in the index file storage unit 108 together with the position identification information (step S204). This step S201 to step S20
The process of 4 is repeated for the number of designated documents (step S205). The storage positions of the keywords in the index file storage unit 108 are positions arranged in ascending order of high-order words and in ascending order of keywords. At this time, since the items of the dictionary and thesaurus dictionary are numbered so as not to overlap each other, they may be sorted in ascending order by the number corresponding to the word, or may be in the character code order of the word.

FIG. 3 shows an example of the index file.
FIG. 3A shows an example of a conventional general index file, in which a keyword is associated with a list of document files including the keyword. FIG. 3 (b)
Is an example of the index file in the present embodiment, and each keyword has a higher word than the conventional index file in FIG. 7A, and the keywords are classified by the higher word by sorting by the higher word. The difference is. In the present embodiment, these superordinate words and keywords are character strings of word notations, but they may be represented by pointers to storage locations of notations. The file list is represented by a list indicating the appearance position of the keyword or a pointer to the list. The appearance position of the keyword is the file name, file serial number,
Alternatively, it is represented by a physical address or the like. There is a possibility that a word existing as a high-order word and a word existing as a keyword under another high-order word are the same, but in the present embodiment, since there are only two hierarchies, processing for eliminating duplication is not performed. That is, when there are a plurality of upper words for one keyword, the same keyword is hung in each of the plurality of upper words in the present embodiment. You may select a word. FIG. 3C shows a specific example of the index file created. That is, the index file created in this way has the generic word "mountain" as illustrated in FIG. 3 (c).
It corresponds to the keywords "Aso", "Ishizuchi", etc., which are classified by "river", and the list of documents including them.

(Embodiment 2) This embodiment 2 is an embodiment in which an index file is created when a document is registered as in the case of the embodiment 1, but a part of the processing procedure of extracting a keyword and registering it in the index file is Different from the first embodiment.
That is, as described above, in the first embodiment, a process of obtaining a high-order word every time a keyword is extracted, classifying the high-order word by the high-order word, and saving the high-order word and the key together with the position identification information in the index file is performed. In other words, so to speak, high-order words corresponding to keywords are extracted in parallel and registered in the index file. On the other hand, in the keyword extraction and the registration to the index file of the second embodiment, once the keyword is extracted, it is once stored in the index file together with the position identification information, and after that, the keyword in the index file is completely extracted after the extraction of the keyword is completed. The upper word is extracted, the keywords are classified by the upper word, and the upper word and the key are saved again in the index file together with the position identification information. The block diagram showing the configuration of the apparatus is the same as that of the first embodiment and is shown in FIG.

FIG. 4 is a flow chart relating to keyword extraction and index file registration in the second embodiment. The document input unit 102 starts reading from the document storage unit 101 the document file for which the keyword is to be extracted by the keyword extraction unit 103 according to a user's instruction (step S401). The keyword extracting unit 103 analyzes the read documents in order and refers to the analysis dictionary 104 to extract an independent word (step S402). Index registration unit 10
7 stores the extracted word as data of the index file in the index file storage unit 108 (step S403). After extracting the keywords for all the registered documents (step S404), the data in the index file is supplemented. The thesaurus dictionary 106 is accessed through the keyword expansion unit 105 to obtain a broader term for each keyword (steps S405 and S406). The index registration unit 107 classifies the keywords according to the higher-rank words, and stores the higher-rank words and the keywords together with the position identification information and the like in the index file storage unit 10.
It is stored again in step 8 (step S407). The index file created in this way is similar to the first embodiment, and the keywords “Aso”, “Ishizuchi”, etc., which are classified by the generic terms “mountain”, “river”, etc. as illustrated in FIG. It corresponds to the list of included documents.

(Embodiment 3) Embodiment 3 is an example of a search device for searching using the index file created in Embodiment 1 or 2, and FIG. 5 is a block diagram thereof. This search device includes a search character string designating unit 501 for designating a search character string for searching a document, and a search keyword for analyzing a search character string designated by the search character string designating unit 501 and extracting a search key. An extraction unit 502, an analysis dictionary 503 that is referred to by the search keyword extraction unit 502 to extract a search keyword, a keyword expansion unit 504 that obtains an upper word of the keyword extracted by the search keyword extraction unit 502, and an upper word. For this purpose, a thesaurus dictionary 505 referred to by the keyword expansion unit 504, an index file storage unit 507 that stores an index file that associates keywords classified by upper terms with the appearance positions of documents including the keywords, and the index file are used. A search unit 506 for searching the appearance position of the search key by
A document storage unit 508 that stores documents and a display unit 509 that displays search results are provided.

FIG. 6 shows a third embodiment constructed as described above.
3 is a flowchart showing the operation of the search device of FIG. The user designates a search character string to be searched by the search character string designating unit 501 (step S601). Search keyword extraction unit 502
Analyzes the search character string and extracts a search keyword (step S602). For example, if the search keyword is "search system", two words "search" and "system" are extracted. The search keyword extraction unit 502 uses the analysis dictionary 503 to extract the search keyword. Also, the search keyword extraction unit 502 is a keyword expansion unit 504.
The thesaurus dictionary 505 is accessed through to obtain the high-order word of the search keyword and send the result to the search unit 506 (step S603). The search unit 506 searches the index file in the index file storage unit 507. That is, first, it searches for a higher-rank word obtained by the keyword expansion unit 504 that matches the higher-rank word in the index file (step S604). As a result of that search,
If there is a matching high-ranking word, a search keyword or a synonym is further searched (steps S605 and S60).
6, step S607, step S608), if there is no match in the upper word, the search for the search keyword is ended (when the determination in step S605 is no).

The case where there is a matching upper word will be further described. First, whether or not the thesaurus search is to be performed is designated in advance, or the user is notified that the higher term is found, and the user is designated. wait. When the designation is not for the thesaurus search, the range of the index file belonging to the range of the upper word is searched by the search keyword obtained in step S602 (step S607). For example, if the index file includes the information illustrated in FIG. 3C, and the search keyword is “ASO”, the index file is searched for by the superordinate word “mountain”. If you do so, it will be correct, so if you decide whether to perform the thesaurus search next, and if you do not perform the thesaurus search, compare the keyword belonging to "mountain" in the index file with the keyword specified as "Aso" Search by. Therefore, the range of the search is limited to the ranges of "Aso", "Ishizuchi", "Akagi", and "Fuji", and the search speed is increased.

When a search keyword is expanded into a thesaurus for a search, the specified search keyword is expanded into a thesaurus to obtain synonyms. Then, the index file is searched for using the synonyms within the range in which the upper terms match (steps S606 and S608). Also in this case, the range of the search is the high-ranking word "mountain" in the index file as in the case of the above-mentioned search keyword "Aso".
The search speed is fast because it is limited to the range. When there are a plurality of search keywords, steps S603 to S608 are repeated by the number corresponding to the number of search keywords (step S609).

The search result is displayed on the display unit 509 (step S610), and if requested by the user (step S6).
11), the document file storage unit 508 is accessed and the appearance position in the document is also displayed (step S612).

In this embodiment, an index file having keywords classified by high-order words is first searched for by the high-order word of the search keyword, and if there is a high-order word, the index file with a limited range belonging to the high-order word is searched. Since it suffices to perform a search by comparing with the keyword of, the extra search by the high-ranking word is required more than before, but the search speed is greatly improved as a whole. In addition, in the present embodiment, when searching for items classified by high-ranking words, presenting to the user whether high-ranking words have been found, and prompting whether or not to search by synonyms, the user is It is possible to know in advance whether or not there is a word, and it is possible to judge whether it is effective to search with a synonym, so it is sufficient to specify whether to search with a synonym based on that judgment, so that the desired search result can be obtained efficiently. To be When all synonyms are searched, the result can be obtained only by searching the high-ranking words without expanding the search key in the thesaurus.

(Embodiment 4) This embodiment is an embodiment in which the index file creating apparatus shown in Embodiment 1 or 2 and the search apparatus shown in Embodiment 3 are combined.
When creating the index file, the document storage unit 70
1, the document input unit 702, the keyword extraction unit 703, the analysis dictionary 704, the index file storage unit 708, the keyword expansion unit 705, the index registration unit 707, and the thesaurus dictionary 706, the same configuration as that of the first or second embodiment. The index file creation device that has operations functions. Further, when performing a search using the index file, a search character string designating unit 709, a search keyword extracting unit 710, a search unit 711, a display unit 712, a keyword expanding unit 705, a thesaurus dictionary 706, an analysis dictionary 704, and an index file storage unit 708. The search device having the same configuration and operation as the third embodiment functions by the document storage unit 701. According to the apparatus of the fourth embodiment, one apparatus can perform both the registration registration of the index file and the search using the index file.

(Fifth Embodiment) FIG. 8 is a block diagram showing the structure of the fifth embodiment, which is different from the fourth embodiment shown in FIG. 7 only in the structure of the dictionary section. As an analysis dictionary for keyword extraction, the extraction of keywords can be unified into one dictionary by setting the range of words registered in the thesaurus. In FIG. 8, the dictionary 804 is such that the thesaurus dictionary and the analysis dictionary are unified in this way, and the other configuration is the same as that of the fourth embodiment of FIG. 7. Further, the operations of creating and registering an index file and performing a search using the index file are almost the same.

(Modification of Embodiment) In the fourth and fifth embodiments, the index creation and registration is performed as shown in FIG.
The above procedure can be used, but in that case, step S405, step S406, and step S407 in FIG. 4 may be processed in the background. In this case, when a search command is issued during the background processing, the search is performed according to the procedure of FIG.

Further, in each of the above-mentioned embodiments, the key stored in the index, the broader word, and the search key are assumed to be character strings as shown in FIG. 3C, but instead of the character strings shown in FIG. The serial numbers in the analysis dictionary and thesaurus dictionary may be used as shown in FIG. In this case, since the search can be performed with a fixed length, the address calculation at the time of the search is simplified, the search speed is improved, and the size of the index file is reduced.

[0026]

According to the document search apparatus of the present invention, an index file having keywords classified by higher-order words is first searched for by a higher-order word of a search character string, and if there is a higher-order word, it belongs to the higher-order word. Since it suffices to perform a search by comparing with a keyword having a limited range, the search speed can be significantly improved as a whole even if an extra search process using an upper word is required. Further, when narrowing down the comparison target of the search character strings, the number is narrowed down based on the relation of the high-order word instead of simply reducing the number as in the conventional technique, so that there is little possibility of missing the search.

According to the index creating device of the present invention,
The specified document is analyzed by the analysis dictionary to extract the keywords, the high-order words of the extracted keywords are obtained, and the extracted keywords are classified by the high-order words to create an index file. Its creation can be done automatically,
And it can be done easily.

[Brief description of drawings]

FIG. 1 is a schematic block diagram of a first embodiment (and a second embodiment).

FIG. 2 is a diagram showing a flow of index creation processing according to the first embodiment.

FIG. 3 is a diagram showing an example of an index file, (a)
Is a conventional example, (b) to (d) are diagrams showing an example of the present invention.

FIG. 4 is a diagram showing a flow of creating an index according to the second embodiment.

FIG. 5 is a schematic block diagram of a third embodiment.

FIG. 6 is a diagram showing a flow of search processing according to the third embodiment.

FIG. 7 is a schematic block diagram of a fourth embodiment.

FIG. 8 is a schematic block diagram of a fifth embodiment.

[Explanation of symbols]

101 ... Document storage unit, 102 ... Document input unit, 103 ... Keyword extraction unit, 104 ... Analysis dictionary, 108 ... Index file storage unit, 105 ... Keyword expansion unit, 1
07 ... Index registration unit, 106 ... Thesaurus dictionary,
501 ... Search character string designation unit, 502 ... Search keyword extraction unit, 503 ... Analysis dictionary, 504 ... Keyword expansion unit, 505 ... Thesaurus dictionary, 506 ... Search unit, 507
... index file storage unit, 508 ... document storage unit,
509 ... Display unit.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Akio Yamashita Akio Yamashita 3-2-1 Sakado, Takatsu-ku, Kawasaki City, Kanagawa KSP R & D Business Park Building Fuji Xerox Co., Ltd. (72) Kazuo Aihara Takatsu-ku, Kawasaki City, Kanagawa Prefecture 3-2-1 Sakado KSP R & D Business Park Building Fuji Xerox Co., Ltd. (72) Inventor Tatsuomi Kita 3-2-1 Sakado, Takatsu-ku, Kawasaki City, Kanagawa KSP R & D Business Park Building Fuji Xerox Co., Ltd. (72 ) Inventor Shinji Kawamoto 3-2-1 Sakado, Takatsu-ku, Kanagawa Prefecture KSP R & D Business Park Building Fuji Xerox Co., Ltd. (72) Inventor Hiroshi Yamaguchi 3-2-1 Sakado, Takatsu-ku, Kawasaki City, Kanagawa Prefecture KSP R & D Business park Le Fuji Xerox Co., Ltd. in

Claims (2)

[Claims] 1. A document storage unit for storing a document, an index file storage unit for classifying keywords for the document by upper terms and storing them together with document identification information, and a search for designating a search character string for searching a document. The character string specifying means, the upper word extracting means for obtaining the upper word of the search character string specified by the search character string specifying means, and the upper word are searched for matching with the upper word in the index file. If so, a document search device provided with a search means for searching a range of a keyword group classified into the superordinate word with a search character string. 2. A document input means for reading a document, an index file storage means for storing an index file in which keywords classified by broader terms and document identification information are associated with each other, and keywords used for retrieval from the input document. A keyword extracting means for extracting a keyword, an upper word extracting means for obtaining an upper word of the keyword extracted by the keyword extracting means, and a keyword extracted by the keyword extracting means classified by the upper word extracted by the upper word extracting means An index file creating apparatus comprising: a registration unit that registers a keyword and a keyword together with the document identification information of the read document in an index file storage unit.