JPH05204993A - Retrieving device - Google Patents

Abstract

PURPOSE:To reduce the capacity of a summary file which is used for presearch the whole sentence in particular and to improve the retrieving efficiency for a retrieving device which applies a whole sentence retrieving system that can extract the requested document data out of a secondary storage storing a large quantity of document data with no application of the retrieving index information. CONSTITUTION:The capacity of a summary file 10 can be reduced in such a way where the independent word candidate groups extracted out of a text file 11 with no duplication are arranged and continuously stored in a file for each number of independent words. Then a retrieving range is limited based on the information on the boundary positions among those independent word candidates. Thus a whole sentence is retrieved with high efficiency.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a retrieval device based on a full-text retrieval system capable of extracting requested document data from a secondary storage device storing a large amount of document data.

[0002]

2. Description of the Related Art In recent years, due to the spread of word processors and personal computers, a large amount of document data has been distributed and used at work and at home. In order to organize and effectively use this large amount of document data, large-capacity databases and high-speed search devices have been researched and developed.

However, in the conventional search machine, it is necessary to add index (keyword) information for searching, and as the amount of data increases, this indexing work requires a great deal of labor. On the other hand, as a method that does not require the work of adding the index information, a search based on a full-text search method capable of accurately and quickly finding desired document data from the large amount of document data without the index information. The device has been announced. For example, 1970 Slotnick, D.
L.) proposed the Logic Par Track Disc,
This is an attempt to speed up the search by adding a search-dedicated processor to each head of the disk, which is a type of secondary storage device, and transferring only the information satisfying the search conditions to the host computer. RAP of the University of Toronto has been realized as a device. On the other hand, a text search machine for full-text search (Technical report of IEICE /
Data Engineering 89-38) attempts to speed up the search by using a secondary storage device and using a hierarchical pre-search method that uses two types of summary files: a character component table and condensed text.

[0004]

However, in the logic-per-track disc, in order to perform a full-text search, it is necessary to search the entire disc, which is a kind of secondary storage device. There was a problem that the time required for inputting / outputting data increased and it became difficult to speed up the search.

Further, in the above-mentioned text search machine for full-text search, a plurality of secondary storage devices and a summary file in which the whole text is condensed are used for the purpose of reducing the time required to input / output data to / from the secondary storage device. Although the pre-research method is used, there is a problem that the summary file inevitably becomes large.

In consideration of the above problems of the conventional retrieval apparatus, the present invention provides a retrieval apparatus which has a smaller capacity than a full-text retrieval text machine and can use a summary file with high efficiency in retrieval. To aim.

[0007]

According to a first aspect of the present invention, there is provided a search device for pre-searching a pre-created summary file in response to a search request and using the result to obtain a corresponding text file. The summary file is a group of independent word candidates obtained by extracting independent word candidates from the corresponding text file based on the Japanese grammar rules and classifying the independent word candidate groups obtained by this by grouping them by the number of independent word characters, And a boundary position information for each character number, which is a summary file.

According to the present invention of claim 2, in response to a search request,
In a search device that pre-searches a pre-created summary file and uses the results to obtain a text file, the summary file extracts independent word candidates from the text file based on Japanese grammar rules and When an independent word candidate of a word candidate group comprehensively contains another independent word candidate in itself, the contained independent word candidate is deleted, and the independent word candidate group obtained thereby becomes independent. It is a retrieval device which is a summary file composed of independent word candidate groups obtained by classifying by word character counts and boundary position information for each character count.

[0009]

According to the present invention, as described above, an independent word candidate group in which independent word candidate groups extracted from the whole sentence are collectively recorded for each independent word character and an independent word are used as a condensed file in which the text file is condensed. By using the boundary position information obtained when stopping from the candidate group for each number of characters, it is possible to reduce the storage capacity of the summary file because there is no need to separate independent words, and efficient search based on the boundary position information can be performed. It can be performed. Therefore, an effective full-text search can be performed.

[0010]

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

FIG. 1 is a block diagram of an embodiment of the retrieval apparatus of the present invention.

In FIG. 1, a secondary storage device 1 is a means for storing a large amount of document data, and a recording medium 2 records a body file of document data and a summary file used for retrieval, such as an optical disc. Is something
The character string search circuit 3 is a circuit for performing a search operation, the data memory circuit 4 is a circuit for temporarily recording data, and the control circuit 5 is for the recording medium 2 in response to a search request from a host computer. A control circuit for controlling the data memory circuit 4 and the character string search circuit 3, and the host computer 6 receives the search request from the user, outputs the search result to the user, and sends the search request to the secondary storage device 1. It serves to receive search results from the secondary storage device 1. Reference numeral 7 is a search request sent from the host computer 6 to the secondary storage device 1,
Reference numeral 8 is a search result sent from the secondary storage device 1 to the host computer 6.

Next, the search operation in the search device configured as described above will be described with reference to FIG. The search request input from the user to the host computer 6 in step 1 is sent to the secondary storage device 1. In step 2, the control circuit 5 controls the character string search circuit 3 to search using the summary file recorded in the recording medium 2 based on the sent search request, and in step 3, the search result is obtained. The control circuit 5 controls the character string search circuit 3 to search for a text file based on the above, and in step 4, the search result is transferred from the secondary storage device 1 to the host computer 6.
And send the search results to the user.

FIG. 3 is a configuration diagram of a data file recorded on the recording medium 2. In FIG. 3, a data file 9 is the information recorded on the recording medium 2, and a summary file 10 is a file used for pre-search, which is composed of a boundary position information file and an independent word candidate group file. Reference numeral 11 is a body file in which the document data is stored as it is. At this time, the capacity ratio between the summary file 10 and the body file 11 varies depending on the content of the document data, but in the present invention, the newspaper data is about 1: 4, and the text like a patent document is used. The ratio is about 1 to 10 for a special document that has many independent words that appear in it.

How to create the summary file 10 will be described with reference to FIG. In step 1, independent word candidates are extracted from the full-text file 11 based on Japanese grammar rules. As the method for extracting the independent word, it can be considered to extract only kanji and katakana by utilizing the fact that there are almost no independent words composed of hiragana in Japanese. In step 2, the extracted independent word candidates are deleted from overlapping independent words within a certain document range of the full-text file to obtain a first stage independent word candidate group. Step 3
Then, the extracted independent words of the first-stage independent word candidate group are compared with each other, and the independent words comprehensively included in the other are deleted to obtain the second-stage independent word candidate group. In step 4, the second-stage independent word candidate group is classified and arranged according to the number of characters of each independent word to obtain a third-stage independent word candidate group. At this time, since the information about the maximum number of independent word characters and the boundary of the independent word candidate group for each character number is obtained, the information is used as boundary position information. The summary file 10 is a combination of the third-stage independent word candidate group and boundary position information.

The summary file 10 will be described more specifically. FIG. 5 is a configuration diagram of the summary file 10 used in the pre-search before the full-text search.

In FIG. 5, reference numeral 12 denotes the third stage in which the independent word candidate groups of the second stage extracted from the sentence data stored in the secondary storage device are arranged and stored together according to the number of characters of the independent word. 13 is a schematic diagram of a file composed of independent word candidate groups, 13 is an example of an independent word candidate group consisting of one character in the file, and 14 is an independent word candidate group consisting of one character in the file Is the end address of the position where 15 is stored, 15 is an example of an independent word candidate group consisting of 2 characters in the file, and 16 is 2 in the file.
It is the end address of the position where the independent word candidate group consisting of characters is stored, 17 is an example of an independent word candidate group consisting of 3 characters in the file, and 18 is an independent word consisting of 3 characters in the file. This is the end address of the position where the word candidate group is stored, and the above 13 and 14, 15 and 16,
The entire structure of the independent word candidate group file is obtained by repeating the structures 17 and 18 up to N, which is the maximum number of characters in the independent word candidate group of the third stage extracted in the sentence data.

Reference numeral 19 is a boundary address when the independent word group for each number of characters obtained in the step of arranging the independent word candidate groups in the second stage collectively for each number of characters of the independent word is stored in the independent word candidate group file. Boundary position information that is
FIG. 20 is a schematic diagram of a boundary position information file composed of information of the maximum number of characters N in the independent word candidate group of the stage, and 20 stores information of the maximum number of characters N in the independent word candidate group of the third stage. 21 stores the end address of the position where the independent word candidate group consisting of one character is stored in the independent word candidate group file, and 22 consists of two characters in the independent word candidate group file. The end address of the position where the independent word candidate group is stored is stored, and 23 is the end address of the position where the independent word candidate group consisting of three characters is stored in the independent word candidate group file. Reference numeral 24 stores the end address of the position where the independent word candidate group having the maximum number N of characters is stored in the independent word candidate group file. As described above, the summary file 10 is composed of the independent word candidate group file 12 and the boundary position information file 19.

In this embodiment, the independent word candidate group file 12 and the boundary position information file 1 are used as the summary file 10.
Although it is divided into two, it is made into one file by connecting the position information file 19 before and after the independent word candidate group file 12 or divided into a plurality of files by the number of independent word character groups for each candidate group. You may.

Further, in this embodiment, the independent word candidate group files are arranged and stored in the ascending order of the number of characters, but they may be arranged and stored in the descending order of the number of characters.

The independent word candidate group file 12 has a number of characters of each independent word for the independent word candidate group selected from the full text file based on the Japanese grammar rules so as not to overlap. Categorized into
It is configured such that the sorted candidate word groups for each number of independent words are continuously stored without a break between each independent word. In this way, since the independent word candidate group file 12 continuously stores the candidate groups for each number of independent word characters in the file, information such as spaces for separating independent words or extra information is not entered. The capacity can be reduced as compared with the conventional method.

Further, the storage address information at the end of the candidate group for each independent word character number obtained at the time of sorting the independent word candidate group is stored as a boundary position information file. If, for example, a word consisting of three characters is given as a search request based on this boundary position information, the above-mentioned 12 characters are used only between the end address of two characters and the end address of three characters of the boundary position information. Since it suffices to collate the independent word candidate group file with the required word, unnecessary collation can be omitted, and pre-search can be performed more efficiently.

In the present embodiment, the search efficiency is improved by collating the specific range of the independent word candidate group file that matches the number of characters of the search request based on the boundary position information file. It is also possible to collate a range within the number of characters in the search request in the group file. In this case, for example, if the search request is "Seto", compound words such as "Setomono" and "Seto Ohashi" can be dealt with, and "Seto" is replaced with "Setomono" or "Seto Ohashi" in the independent word candidate group file.
It can be deleted as the same as, and the capacity can be reduced. At this time, the boundary position information can be used to determine whether the position of the search word in the independent word candidate group file does not extend across the boundary position or each independent word.

Further, the present invention may be realized by software using a computer or by using a dedicated hardware circuit.

[0025]

As described above in detail, according to the present invention, the following effects can be obtained.

(1) The independent word candidate group file, which is a summary file used as a pre-search for the search, extracts independent word candidates from the main text file, and arranges the independent word candidate groups collectively by the number of characters of each independent word. Since the candidate groups for each number of independent word characters obtained in this way are stored consecutively, there is no need for delimiters such as spaces required between independent words in the conventional summary file, and the summary file can be stored with a small capacity. Can be created.

If the search area is not specified, a summary file can be created with a smaller capacity by combining the compound words.

(2) Boundary position information file obtained when an independent word candidate group file is created collectively for each number of independent word characters in order to know the search position of the independent word candidate group file in the pre-search stage of the search By using, it is only necessary to search only the specific region of the independent word candidate group file for each character number of the word given as the search request, so that the pre-search can be efficiently performed.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a search device according to an embodiment of the present invention.

FIG. 2 is a flowchart showing a search procedure in the search device.

FIG. 3 is a configuration diagram of a file in the search device.

FIG. 4 is a flowchart showing a procedure of creating a summary file in the search device.

FIG. 5 is a schematic diagram of a summary file in the search device.

[Explanation of symbols]

1 Secondary Recording Device 2 Recording Medium (eg Optical Disk) 3 Character String Search Circuit 4 Data Memory Circuit 5 Control Circuit 6 Host Computer 7 Search Request 8 Search Result 9 File Structure Recorded in Secondary Recording Device 10 Summary File 11 Body File 12 Independent word candidate group file 13 1 character Independent word candidate group example 14 1 character Independent word candidate group file storage end address 15 2 characters Independent word candidate group example 16 2 character Independent word candidate group file storage end address 17 3 characters Independent word candidate group example 18 3 characters Independent word candidate group file storage end address 19 Boundary position information file 20 Independent word maximum number of characters information 21 1 character Independent word candidate group end address 22 2 characters Independent word candidate group end address 23 3 characters Independence Ending word group address 24 Maximum number of characters Independent word group ending address

Claims (2)

[Claims] 1. A search device for pre-searching a pre-created summary file in response to a search request and using the result to obtain a corresponding text file, wherein the summary file complies with Japanese grammar rules. Independent word candidates are extracted from the corresponding text file based on the above, and independent word candidate groups obtained thereby are collectively classified by the number of independent word characters, and composed of boundary position information for each number of characters. Retrieval device characterized by being a summarized file. 2. A search device for presearching a preliminarily created summary file in response to a search request and using the result to obtain a text file, wherein the summary file is based on Japanese grammar rules. When an independent word candidate is extracted from the text file and the independent word candidate of the independent word candidate group comprehensively includes another independent word candidate in itself, the independent word candidate that is included is deleted. It is characterized by being a summary file composed of the independent word candidate group obtained by classifying the independent word candidate group thus obtained by the number of independent word characters and the boundary position information for each character number. Search device.