JPH08314975A - Information retrieving device - Google Patents

Abstract

PURPOSE: To provide an information retrieving device capable of retrieving data including a key word at a high speed. CONSTITUTION: The information retrieving device retrieving data including the key word is provided with a character information extracting means 33 calculating the number of the appearing times of characters constituting retrieving object data, a retrieving file generation means 35 preparing a retrieving file by coordinating the set of adjacent characters in retrieving object data and the appearing frequency pattern obtained by combining each number of the appearing times of this character and a retrieving means 39 retrieving the appearing frequency pattern corresponding to the set of the adjacent characters in the key word from the retrieving file and collating the obtained appearing frequency data to retrieve retrieving object data including the key word. As this device uses a system utilizing the appearing frequency, the size of the retrieving file is reduced compared with a conventional system which determines the existing place of the pattern by the position of the character. In addition, retrieval is quickened by disposing the appearing frequency in the retrieving file in a rising or falling order.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information retrieval apparatus for retrieving data containing a character string designated by a keyword from a database, and particularly to high-speed retrieval processing and processing for adding, updating, and deleting data in the database. It can be done in.

[0002]

2. Description of the Related Art In recent years, in information retrieval, when a user inputs a word that he / she came up with as a keyword, full-text retrieval of data registered in a database and automatic retrieval of data including a character string of the keyword are attracting attention. Are gathering. Especially, the relational database management system (RDBMS) is being commercialized,
Applications in many fields are widespread.

An information retrieval device of this type is disclosed in, for example, Japanese Patent Laid-Open No.
3-155324, JP-A 1-282635, JP-A-2
-302869, Japanese Patent Laid-Open No. 3-113626. In Japanese Patent Laid-Open No. 63-155324 (information retrieval device),
Information is retrieved by shifting the registered data by one word for each cycle with respect to the retrieval keyword designated by the operator. In addition, Japanese Patent Laid-Open No. 1-282635
In the (index maintenance method), a main index and a comparatively small sub-index are retained as indexes used for searching registered data, and the index record of additional registered data is added to the sub-index to register the registered data. It speeds up the update / addition / deletion of data. Further, in Japanese Patent Laid-Open No. 3-113626 (buffer control system and device), a task search is performed in parallel with the retrieval of data in the buffer and the transfer of the data from the auxiliary storage device to the buffer. We are trying to improve the throughput. Further, in Japanese Patent Laid-Open No. 2-302869 (file editing method), in order to eliminate waste of the memory area and to identify the data to be processed on the memory at high speed, a table holding a free area pointer of the memory is provided. A table is created and the character string data to be processed on the memory is specified by referring to this table.

In these four conventional information retrieval devices,
A method in which a data area and a data update area are secured in a storage area in units of record numbers, data is stored in the data update area at the time of data update, and the data area and the data update area are replaced (Japanese Patent Laid-Open No. Sho 63-63). -1554324, JP-A-1-282635, JP-A-2-302869),
Add / update / update data using the correspondence table (table) between the character string data stored in the search file and its records
A method of deleting (Japanese Patent Application Laid-Open No. Hei 2-302869) is adopted, and the present invention corresponds to the latter.

Next, data registration, change, and deletion processing in the conventional information retrieval apparatus will be described. FIG.
Shows a processing diagram of the conventional information retrieval apparatus, and FIG. 20 shows a table and a data flow in the conventional information retrieval apparatus. As a search target file, the three records in the table of FIG. 20A will be described as an example. Here, in the A record, two characters and two characters are respectively stored as a data structure, and
Records shall store 2 and 5 characters.
At this time, A record “12” and A record “Ai”
Corresponds to the area, and “34” of the A record corresponds to “up” of the A record. As shown in FIG. 19, the data registration, modification, and deletion processes are performed by sequentially reading the table of FIG. 20A of the search target file as a search target record, and step 2: the position of each character (record Whatever number), the character size, and a pointer to the same item (item) are set, and the processing of converting the table of FIG. 20A into the table of FIG. 20B is performed on the memory. For example, the pointer set is “12” of the first record A and the third record A.
Since "Ai" of is a same item, a pointer p1 to "Ai" is set. First record A
“34” of “3” performs the process of setting a pointer to “up” of the third record A.

Step 3: The above processing is performed as shown in FIG.
20 is repeated until the end of the table, and the table of FIG. 20B is finally created by setting the table in the empty area of the memory.

Step 5: When data is updated, for example, in FIG. 20A, "34" in the first record A is changed to "56".
To update to, the record A is checked from the table of FIG. 20B, the content “34” of the third line is detected,
Rewrite as "56". No size or pointer changes occur.

Further, in the case of deletion, for example, in the case of deleting the "up" of the third record A in FIG.
The record A is examined from the table in (b), the pointer "p2" in the third line is traced from this table, and the content "up" in the seventh line is detected and deleted. At this time, the pointer to "to p2" on the third line is also deleted.

As described above, in the conventional information retrieval apparatus, registration, update, and deletion are realized by tracing the table with the head address, size, and pointer as clues.

[0010]

However, in the conventional information retrieval apparatus, when the number of data in the database is large, it takes a long time to retrieve the full text of the data including the character string designated by the keyword. Is required.

Further, in the conventional information retrieval apparatus, when updating or deleting data, it is necessary to secure an extended area for the record to be updated or deleted for the processing, and a huge storage area must be prepared. There was a problem.

Further, after updating or deleting the data, it is necessary to perform a sorting process to arrange the order of the records, which causes a problem that it takes time to update or delete the data.

Further, at the time of updating or deleting the data, since the updated information is inserted into the search file in which the data is tightly packed or the data is deleted from this file, the position after the insertion position or the deletion position is deleted. However, there is a problem that it takes time to update and delete all the data in (1).

There is also a method of setting a deletion flag for the data to be deleted so as to make it in time, but in this case, there is the problem that the number of search files increases as the deletion data increases. It was

The present invention solves the above-mentioned conventional problems, and it is possible to search at high speed whether or not data including a character string designated by a keyword exists in a database, and the data can be searched. It is an object of the present invention to provide an information retrieval device that can quickly update and delete.

[0016]

Therefore, in the present invention, in an information retrieval apparatus for retrieving data containing a keyword, character information extraction means for calculating the frequency of appearance of characters constituting the retrieval object data, and retrieval object data. Search file generation means for creating a search file by associating a set of adjacent characters with an appearance frequency pattern combining the respective appearance frequencies of the characters, and an appearance frequency pattern corresponding to a set of adjacent characters in the keyword from the search file Is provided, and the obtained appearance frequency pattern is collated to search for search target data including a keyword.

Further, the search file generating means arranges the appearance frequency patterns on the search file in ascending or descending order,
The search means uses the appearance frequency pattern in the range satisfying the ascending order or the descending order on the search file for matching.

Further, the search file generating means adds the same appearance frequency pattern as the preceding appearance frequency pattern to the end of the array of the appearance frequency pattern when creating the search file, and the searching means adds the added appearance frequency pattern. When the order of matching in the pattern is reached, the matching is stopped.

Further, a blank extension area is added to the end of the array of appearance frequency patterns in the search file, and when it is necessary to add an appearance frequency pattern as the data to be searched changes, the appearance frequency is added to this extension area. A search file correction means for adding a pattern is provided.

Further, in order to specify the range of the effective appearance frequency pattern to be used for the keyword search from the appearance frequency patterns described in the search file, the occurrence of the effective appearance frequency pattern at the end of the array before the appearance occurs. A search file correction means is provided in which the same appearance frequency pattern as the frequency pattern is added and a blank extension area is added before the array of only the invalid appearance frequency patterns.

[0021]

In this information retrieval apparatus, the frequency of appearance of each character type of the retrieval target data, that is, the number of the character that appears in each character type is calculated, and the character string of the data is shifted by one character at a time. , An appearance frequency pattern in which the appearance frequency of each character of a character pattern consisting of the character and an adjacent character is set, and the appearance frequency pattern is collected to form a search file. At the time of search, the keyword is divided into a plurality of character patterns, the appearance frequency pattern corresponding to each character pattern is obtained from this search file, the appearance frequency pattern is collated, and each character pattern is continuous in the search target data. When it is found, the data is detected as containing a keyword.

Since this information retrieval apparatus uses the appearance frequency, the size of the retrieval file can be reduced as compared with the conventional method in which the location of the pattern is held at the character position. Also, when searching for keywords,
From the character string of the search target data, only the character pattern that matches the character pattern of the keyword is read from the search file, and the inclusion or non-existence of the keyword is checked based on the appearance frequency pattern. Since it is not necessary to match the character string of, the search can be performed at high speed.

Further, the order of the appearance frequency patterns in the search file is set in ascending or descending order, and at the time of retrieval, the matching of the appearance frequency patterns is proceeded within the range satisfying this ascending or descending order, so that the other party who does not need to be verified You can check at high speed. Therefore, the number of collations can be reduced and the search can be speeded up.

At the time of updating the data, a new appearance frequency pattern can be overwritten on the appearance frequency pattern of the search file or the extended area, so that it is not necessary to shift the data of the search file one by one. Files can be updated at high speed. Also, by providing an extension area to which an appearance frequency pattern can be added after each array of the appearance frequency pattern, the allocation ratio of the extension area in the search file is made constant, and the update processing time is averaged. You can set a standard for the update time.

When deleting data, instead of deleting all the related data of the search file, an appearance frequency pattern that breaks the ascending or descending order or a blank diffusion area is provided before the array of invalid appearance frequency patterns. By
It can be distinguished from the effective appearance frequency pattern. Therefore, since it is not necessary to delete the search file, the deletion process can be realized in a short time.

In the present specification, the term "character" is used to include not only character data but all patterned or symbolized data. The information retrieval device of the present invention can realize registration, update, deletion, and retrieval of not only character data but also all patterned or symbolized data.

[0027]

【Example】

(First Embodiment) In the information retrieval apparatus of the first embodiment, when the text character string is, for example, "Ahhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh"
"N") Another appearance frequency, that is, what number "a" is each "a" in the character string, what number "i" is each "i", and A character pattern frequency table (FIG. 3) is created by determining what number "n" is, and then a set of two consecutive characters in the character string is set as a character pattern, and the character pattern is corresponded to. Find the appearance frequency pattern. The character pattern is (a, a) (a, i) (i, a) with respect to “aaaa” in the character string, and the appearance frequency pattern of each character pattern is (1, 2)
It becomes (2,1) (1,3). Next, the appearance frequency patterns are collected for each character pattern to create an appearance frequency list (index file) for each text shown in FIG.

On the other hand, when the character string of the keyword is "Aia", the character pattern of this character string is (A, I)
(I, ah). In matching the keyword with the text, first, an appearance frequency pattern corresponding to the character pattern of the keyword is obtained from the text appearance frequency list. The appearance frequency pattern of (A, I) obtained from the appearance frequency list of FIG. 5 is (2, 1) (4, 2) (8, 3),
The appearance frequency pattern of (i, a) is (1, 3) (2, 5)
Is. When the second element of the appearance frequency pattern of (A, A) and the first element of the appearance frequency pattern of (A, A) match, the text includes the character string "Aia" of the keyword. . In the case of this example, the second element (1) of (2,1) and the first element (1) of (1,3) are the same, and the second element (2) of (4,2) is Since the first element (2) of (2,5) matches, this text is "Aia".
Is included.

The information retrieving apparatus of the embodiment basically retrieves the text containing the keyword in this manner. As shown in FIG. 1, the apparatus includes a text database 31 that stores text data to be searched, a text generation unit 32 that reads texts from the text database 31 one by one and generates a text record for each text. A keyword reading unit 38 that reads the keyword you want to search, a correction text reading unit 312 that reads the corrected text data for text correction, a character pattern frequency table from the text record and the correction text record, and the character pattern from the keyword Continuous character information extraction unit 33
And an index generation unit 35 that creates index file data including a character pattern and its appearance frequency pattern from a character pattern frequency table, and a continuous character information extraction unit 33.
Character pattern frequency table and index generator created by
The index file storage unit 34 that stores the index file created in 35, the expansion area generation unit 36 that adds an expansion area to the end of the appearance frequency pattern list of the index file, and the index file data and the character pattern of the keyword An index search unit 39 for matching and searching for whether or not the character pattern of the keyword is included in the text in a continuous state, and an index search unit when there is no occurrence frequency pattern to be matched
39, the character frequency existence search unit 310 that instructs the collation stop,
A search result output unit 311 for displaying and outputting the search result of the index search unit 39, and an update data generation unit 37 for updating the appearance frequency pattern of the index file at the time of text correction are provided.

In the first embodiment, the operation of generating an index file from text and the operation of searching for text containing a keyword using this index file will be described. In this case, the modified text reading unit 312 and the update data generating unit 37 of the information retrieval device of FIG. 1 are not used.

The index file is generated according to the processing flow of FIG.

Step 1: First, the text generator 32
The text is read from the text database 31 and the record number Ni is assigned to the text record of the text. Now, suppose this text record is "Ahhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh" Step 2: The continuous character information extraction unit 33 sets the pointer p to the first character of the text record and initializes the count value of the appearance frequency N (A) of the character type, and then Step 3: the step of the character at the pointer position. 4: Count the appearance frequency for each character type, Step 5: Create a character pattern frequency table (Fig. 3),
It is stored in the index file storage unit 34. In this table, a set of appearance frequency and record number for each character type is described.

Step 6: When the pointer position has not reached the last character of the text record, Step 7: The pointer is moved to the next character, and the procedure returns to Step 3 to check the appearance frequency of each character for each character type.

When the creation of the character pattern frequency table for the text record of record number Ni is completed by repeating this procedure, Step 8: The index generation unit 35 adds the character pattern ( A pointer i designating a character set of two characters each) is set, and step 9: the appearance frequency of each character of the character pattern pointed to by the pointer i is obtained, and step 10: the appearance frequency pattern composed of the set of the appearance frequencies is set to It is recorded in the index file (FIG. 5) in association with the character pattern. The record number Ni is also recorded in the index file.

Step 11: If the pointer i has not reached the end of the text record, Step 12: Move the pointer i by one character and return to Step 9, and the appearance frequency of the character pattern pointed to by the pointer from the character pattern frequency table. A pattern is obtained, and when the character pattern is not yet recorded in the index file, the character pattern and the appearance frequency pattern are recorded in the index file in association with each other, and the record number Ni is also recorded. If the character pattern is already recorded in the index file, the obtained appearance frequency pattern is recorded at the end of the appearance frequency pattern list corresponding to the character pattern.

When the pointer i reaches the end of the text record by repeating the above procedure, step 13: the area expansion generating unit 36 provides an expansion area at the end of all the appearance frequency pattern lists of the index file, and sets it there. It is used as a blank area, or the same appearance frequency pattern as the last appearance frequency pattern is recorded therein.

FIG. 6 shows the data structure of the index file added with the extension area. In the appearance frequency pattern list of this file, the individual elements forming the appearance frequency pattern are arranged in ascending order. That is, if the elements of the appearance frequency list are (Ai, Bi), A1 <A2 <... <Am, B1
The relationship <B2 <... <Bm holds.

Step 14: The above processing is repeated for the text records sequentially read from the text database 31, and Step 15: When the processing for all the texts stored in the text database 31 is completed, the index file is generated. finish.

Next, the operation of searching the text including the keyword using the index file will be described. This is performed according to the procedure shown in FIG. Here, it is assumed that the keyword is “Aia”.

Step 1: First, the keyword reading unit 38 reads a keyword, and Step 2: the continuous character information extracting unit 33 uses this keyword as a set of two character patterns (A [p], B [p + 1). ]) (B
[p + 1], C [p + 2]). Here, A [p] indicates the p-th character A of the keyword. In the case of the keyword "Aia", it is decomposed into (A, I) and (I, A).

Step 3: The continuous character information extraction unit 33 sets the record number Ni, Step 4: sets the pointer p to the first character position of the keyword, and Step 5: the first two character patterns of the keyword,
In the example, (a, i) and (i, a) are set as the target of collation.

Step 6: The index search unit 39 uses the record number stored in the index file storage unit 34.
From the Ni index file (Fig. 6), the appearance frequency list (M [p] [j], N [p] of the first character pattern (A, I))
[j]) first (j = 1) appearance frequency pattern (2, 1)
Step 7: Also, the head (k = k =) of the appearance frequency list (M [p + 1] [k], N [p + 1] [k]) of the second character pattern (i, a) is read.
The appearance frequency pattern (1, 3) of 1) is read. Here, an index file corresponding to the keyword "Aia" is shown in FIG. Character pattern (A [p], B [p + 1]) = (Oh,
I) and (B [p + 1], C [p + 2]) = (i, a), the appearance frequency patterns are (MA [p] [1], MB [p + 1] [ 1])
= (2, 1) (MA [p] [n], MB [p + 1] [n]) = (8, 3)
And (MB [p + 1] [1], MC [p + 2] [1]) = (1,3) (MB [p + 1]
[n], MC [p + 2] [n]) = (2,5) correspond. This MA
[q] [j] means the letter A, the jth appearance frequency pattern with respect to the qth character pattern of the keyword character string. Further, at the end of the appearance frequency pattern for each character pattern, a blank or the same appearance frequency pattern as the last appearance frequency pattern is stored as an extension area.

Step 8: As shown in FIG. 9, the index search unit 39 causes the second element of the appearance frequency pattern (2,1) in the first character pattern and the appearance frequency pattern (1 in the second character pattern). , 3) and the first element of (3) are matched, and whether the first character pattern and the second character pattern are continuous is checked. If they match, step 9: The consecutive characters "Aia" and the record number Ni of the index file are output to the search result output unit 311.

Step 10: The index checking unit 39 compares the second element (N [p] [j]) of the appearance frequency pattern with the second character pattern in the first character pattern for the next matching of continuity. The first element (M [p + 1] of the appearance frequency pattern in
[k]) and the first element (M [p + 1] [k]) is larger, or the second element (N [p] [j]) and the first element When (M [p + 1] [k]) is equal, step 11: j is incremented, and step 12: of the second element (N [p] [j]) of the appearance frequency pattern in the first character pattern. When it is larger, or when the second element (N [p] [j]) and the first element (M [p + 1] [k]) are equal, step 13: k is incremented.

Step 14: The appearance frequency existence search unit 310
The second element (N [p] [j]) of the appearance frequency pattern in the first character pattern and the second element (N [p] [j + 1]) of the next appearance frequency pattern in the character pattern The size is compared, and the first element (M [p + 1] [k]) of the appearance frequency pattern in the second character pattern is compared with the first element (M [p [p] of the next appearance frequency pattern in the character pattern. +1] [k + 1])
If they are in ascending order, the process returns to step 15: step 8 to check the second element and the first element between the updated appearance frequency patterns.

In step 14, when there is no ascending order relation, that is, the next appearance frequency pattern is the same as the blank area or the last appearance frequency pattern generated at the end of the appearance frequency list by the area expansion generating unit 36. When the area in which the appearance frequency pattern is recorded is reached, the collation for the character pattern ends.

Here, FIG. 10 shows a flow of search for the keyword "Aia". First character pattern (a,
Frequency) list (2, 1) (4, 2) (8, 3)
First appearance frequency pattern (2,1) of (8,3) and appearance frequency list (1,3) of the second character pattern (i, a)
(2,5) The first appearance frequency pattern of (2,5) (1,
3) and the second element “1” of the first appearance frequency pattern matches the first element “1” of the second appearance frequency pattern, so the result is stored in the search result output unit 311. . Then, the second appearance frequency pattern (4, 2) in the first character pattern is compared with the second appearance frequency pattern (2, 5) in the second character pattern, and the second appearance frequency pattern of the first appearance frequency pattern is compared. Since the element “2” matches the first element “2” of the second appearance frequency pattern, the result is also stored in the search result output unit 311. Subsequently, the third appearance frequency pattern (8, 3) in the first character pattern and the third appearance frequency pattern (2, 3 in the second character pattern
5), the third appearance frequency pattern (2, 5) in the second character pattern is the second appearance frequency pattern (2, 5) in the same character pattern.
The relation of 5) is not satisfied in ascending order. for that reason,
The matching process of this character pattern ends here.

Step 16: When the matching process for the character pattern is completed in Step 14, it is checked whether or not the matching with other character patterns of the keyword has been completed. If not, Step 17: Step 5 The following processing is performed.

Step 18: When the collation of the continuity of all the character patterns of the keyword is completed, it is checked whether the collation result of Step 9 is obtained for all the character patterns of the keyword, and the result is obtained. , The text of record number Ni is displayed as including the keyword.

As described above, in the information retrieving apparatus of the embodiment, the text character pattern (combination of consecutive character types)
Create a search file by associating the appearance frequency list in which the appearance frequency patterns of the character patterns are arranged in ascending order,
Each appearance frequency list corresponding to the character pattern of the keyword is read from this search file, and the appearance frequency pattern between the lists is collated to search whether or not the character string of the keyword is present in the text data. In the matching of the appearance frequency patterns, since the appearance frequency patterns are arranged in ascending order, it is not necessary to comprehensively match the appearance frequency patterns with each other, and the search can be performed with a small number of matching times. Therefore, high-speed search is possible.

The appearance frequency patterns in the appearance frequency list may be in descending order. In addition, this information retrieval device can retrieve not only character string data, but also pattern strings of other patterned data.

(Second Embodiment) In the second embodiment, the data update of the index file due to the change of the text stored in the database will be described.

In the information retrieving apparatus of FIG. 1, the modified text reading unit 312 reads the modified text data for updating the text with the text data number, and the continuous character information extraction unit 33 extracts the character pattern frequency from the modified text data. Create a table. The created character pattern frequency table is stored in the index file storage unit 34. The index generating unit 35 creates an appearance frequency pattern for each character pattern from the character pattern frequency table, and the update data generating unit 37
Overwrites the appearance frequency pattern in the existing appearance frequency pattern list of the index file. Further, the extension area generation unit 36 adds an extension area for the appearance frequency pattern when the number of appearance frequency patterns exceeds the existing appearance frequency pattern list.

The operation of updating the index file data due to this text change will be described with reference to the processing flow of FIG. Here, the case where the text record “Ahhhhhhhhhhh” is changed to “Ahhhhhhhhhhhhhhhhhhhhhhhhhhh”

Step 1: First, the modified text reading unit 31
2 reads the changed text record and sets the record number.

Step 2: The continuous character information extraction unit 33 sets the pointer p to the first character "a" of the text record, initializes the count value of the appearance frequency N (A) of the character type, and then the step 3: pointer Step 4: Count appearance frequency for each character type, and Step 5: Create character pattern frequency table (FIG. 12) and store it in the index file storage unit 34. In this table, the appearance frequency for each character type of the updated text record is described together with the record number.

Step 6: When the pointer position has not reached the last character of the text record after updating, Step 7: The pointer is moved to the next character, and the procedure returns to Step 3 and the appearance frequency for each character type is determined. Find out.

When this procedure is repeated and the creation of the character pattern frequency table relating to the updated text record of the record number Ni is completed, step 8: the index generating unit 35 stores the index file corresponding to the text record number Ni in the index file. After searching from the section 34 and setting the pointer at the beginning of this file, Step 9: Specify the character pattern (set of two characters each) at the beginning of the updated text record in the character pattern frequency table (FIG. 12). Step 10: The appearance frequency of each character of the character pattern pointed to by the pointer i is obtained.

Step 11: The update data generation unit 37 overwrites the appearance frequency pattern consisting of the set of the appearance frequencies on the appearance frequency pattern and the extension area described in the appearance frequency pattern list of the index file pointed by the pointer. To do.

Step 12: When the appearance frequency pattern list is full, the extension area generator 36 adds an extension area to the end of the list.

Step 13: If the pointer i does not reach the end of the updated text record, the index generator 35 moves the pointer i by one character,
Returning to 10, the operation of obtaining the appearance frequency pattern of the character pattern is repeated.

Step 15: When the pointer i reaches the end of the text record, the update data generator 37 generates the same appearance frequency pattern as the last appearance frequency pattern in the list at the end of all appearance frequency pattern lists, Also,
The extension area generation unit 36 adds the extension area to the end of all the appearance frequency pattern lists. In addition, the update data generation unit 37
Updates all the appearance frequency patterns relating to the pre-updated text record that has not been overwritten in the extended area.

FIG. 13 shows the relationship between the appearance frequency pattern list in the updating text record and the index file, and FIG. 14 shows the structure of the updated index file. The same appearance frequency pattern and extension area are added to the end of the appearance frequency pattern list. Further, like the character pattern (n, n), the appearance frequency pattern that exists before the update but does not appear after the update is replaced by the extended area.

The processing procedure of the search performed using the updated index file is the same as that shown in FIG.

As described above, in the information retrieval apparatus of the embodiment,
Since a blank extension area is provided at the end of the appearance frequency pattern list of the index file, even if the number of updated appearance frequency patterns exceeds the number of registered appearance frequency patterns, the extension area Since it can be overwritten immediately, the data can be updated at high speed. Further, since the extension area is configured such that the extension area is added each time to the extent that the area for recording the appearance frequency pattern is not insufficient, the addition work can be done with little effort. In addition, when a huge index file is prepared in advance, not only the memory capacity becomes unnecessarily large, but it takes time to check the appearance frequency pattern to be erased or updated. In this case, there is little possibility of such a situation, and the index file can be updated at high speed when the text data is changed.

(Third Embodiment) In the third embodiment, the correction of the index file when a part of the text is deleted will be described.

The operation will be described by taking as an example the case where the latter half of the text record "Ahhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh" is deleted at the end of the text record "Ahhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh""are deleted. FIG. 15 shows the operation procedure of the information retrieval device at this time. In this operation procedure, steps 1 to 10 are the same as the operation when updating the text record shown in the second embodiment (FIG. 11). The deleted text record constitutes the character pattern frequency table shown in FIG. 16, and the index generation unit 35 determines the appearance frequency of each character of the character pattern from the character pattern frequency table.

Step 11: The update data generator 37 overwrites the appearance frequency pattern list of the index file having the same record number with the appearance frequency pattern consisting of the set of the appearance frequencies.

At this time, since the number of appearance frequency patterns is reduced by deleting a part of the text record, the appearance frequency patterns are not overwritten on the extension area, and the extension area is not added. There is no need for it.

When the operation for obtaining the appearance frequency pattern corresponding to the character pattern is performed up to the end of the text record using the character pattern frequency table of FIG. 16, the appearance frequency pattern list shown in FIG. 17 is obtained.

Step 14: The update data generator 37 adds the same appearance frequency pattern as the appearance frequency pattern at the end of the list to the end of all the appearance frequency pattern lists, and
For a character pattern that does not appear in the text record after deletion, an extension area is added to the beginning of the appearance frequency pattern list.

FIG. 18 shows the index file after deletion. Here, the valid appearance frequency pattern is a portion surrounded by a thick line, and the others are invalid portions that are not used for the search.

The processing procedure for searching the deleted index file is the same as in FIG. Since the index file is searched only in the range in which the elements of the appearance frequency pattern maintain the ascending order, the index file shown in FIG. 18 is not searched in the invalid part.

As described above, in the information retrieval apparatus of the third embodiment, when deleting an index file, all unnecessary index files are not deleted, but only the minimum range of the index file is updated. . Therefore, the trouble of deletion is reduced and the deletion can be performed at high speed.

[0075]

As is apparent from the above description of the embodiments, the information retrieval apparatus of the present invention can retrieve a designated keyword from a database at high speed.
Data can be added, updated, and deleted from the database at high speed.

When adding, updating, or deleting data in this database, the expansion of the existing index file or re-recording of the last appearance frequency pattern minimizes the modification of the index file. You can organize data (remove garbage) while keeping it.

[Brief description of drawings]

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

FIG. 2 is a processing flow of index file generation in the first embodiment,

FIG. 3 is a character pattern frequency table for a text record in the first embodiment,

FIG. 4 is an explanatory diagram of an appearance frequency list according to the first embodiment,

FIG. 5 is a character pattern and appearance frequency pattern according to the first embodiment,

FIG. 6 is an index file according to the first embodiment,

FIG. 7 is a processing flow of index file search in the first embodiment;

FIG. 8 is a character pattern and appearance frequency pattern list at the time of search in the first embodiment,

FIG. 9 is an explanatory diagram of continuity matching in the first embodiment,

FIG. 10 is an explanatory diagram of a search collation order in the first embodiment,

FIG. 11 is a processing flow of updating an index file according to the second embodiment;

FIG. 12 is a character pattern appearance frequency table for updated text records in the second embodiment,

FIG. 13 is a character pattern and appearance frequency pattern according to the second embodiment,

FIG. 14 is an index file according to the second embodiment,

FIG. 15 is a processing flow of deleting an index file according to the third embodiment;

FIG. 16 is a character pattern appearance frequency table for a deleted text record according to the third embodiment.

FIG. 17 is a character pattern and appearance frequency pattern according to the third embodiment,

FIG. 18 is an index file according to the third embodiment,

FIG. 19 is a diagram showing a processing flow of a conventional information retrieval device,

FIG. 20 is a data table of a conventional information search device.

[Explanation of symbols]

 31 Text database 32 Text generation unit 33 Continuous character information extraction unit 34 Index file storage unit 35 Index generation unit 36 Extended area generation unit 37 Update data generation unit 38 Keyword reading unit 39 Index search unit 310 Character frequency existence search unit 311 Search result output Part 312 Modified text reading part

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Tadakazu Kikuchi 1006 Kadoma, Kadoma City, Osaka Prefecture Matsushita Electric Industrial Co., Ltd.

Claims (5)

[Claims] 1. An information retrieval apparatus for retrieving data including a keyword, character information extraction means for calculating an appearance frequency of characters constituting search target data, a set of adjacent characters of search target data, A search file generating unit that creates a search file by associating an appearance frequency pattern in which each appearance frequency is combined with each other, and an appearance frequency pattern corresponding to a set of adjacent characters in a keyword is searched and obtained from the search file. An information retrieving apparatus, comprising: retrieving means for retrieving retrieval target data including a keyword by collating an appearance frequency pattern. 2. The search file generation means arranges the appearance frequency patterns on the search file in ascending or descending order, and the search means has the appearance frequency in a range satisfying the ascending or descending order on the search file. The information search apparatus according to claim 1, wherein a pattern is used for the matching. 3. The search file generating means adds the same appearance frequency pattern as the preceding appearance frequency pattern to the end of the array of the appearance frequency pattern when creating the search file, and the search means The information retrieval apparatus according to claim 2, wherein the collation is stopped when the order of the collation has reached the added appearance frequency pattern. 4. A blank extension area is added to the end of the array of the appearance frequency pattern in the search file, and when the appearance frequency pattern needs to be added due to the change of the search target data, the extension area is added to the extension area. 4. The information search device according to claim 1, further comprising search file correction means for adding the appearance frequency pattern. 5. In order to specify the range of the effective appearance frequency pattern used for the keyword search from the appearance frequency patterns described in the search file, the effective appearance frequency pattern exists at the end of the array before it. 4. The search file correction means for adding the same appearance frequency pattern as that of the above-mentioned appearance frequency pattern, and adding a blank extension area in front of the array of only the invalid appearance frequency pattern is provided. Information retrieval device.