JP2519129B2 - Multi-word information retrieval processing method and retrieval file creation device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information search processing system for searching information. INDUSTRIAL APPLICABILITY The present invention is particularly suitable for performing a partial match search using a multi-keyword, and greatly reduces the number of times the input multi-search input character string and a registered keyword are matched to each other, and The present invention relates to an information search processing method capable of performing partial match (intermediate match, forward match, backward match) and perfect match search processes at high speed. INDUSTRIAL APPLICABILITY The present invention is suitable for an information search processing method for performing multi-keyword partial match search processing in a database system.

[0002]

2. Description of the Related Art Conventionally, a character string matching with a search input character string is performed from the beginning to the end of a keyword included in a record to be searched, and an input character string designated by a searcher and a keyword matching a search condition are selected. Sequential search method that searches record numbers, or create a character string that can be searched and input from keywords and store the search file in the index format, and search the search file for keywords that match the input character string specified by the searcher and the search conditions. An index method of searching using an index structure is generally used as a partial match search technique using multiple keywords.

[0003]

However, in the sequential search method, since the search input character string is matched from the beginning to the end of the keyword of the record to be searched, when a large number of keywords are included, many searches are required for the search. It takes time. For this reason, dedicated processors and LSIs that perform high-speed character string collation have been proposed for searching large-scale databases. However, in these methods, the hardware is limited, and the computer and the dedicated processor that perform the search process are also proposed. And LS
It takes a long time to transfer a character string to and from I, and the realization of high speed that is satisfactory as a system is an issue.

The index method can speed up the partial match search, but has a drawback that the search file becomes huge. For this reason, exact match, prefix match, and suffix match searches are used, but intermediate matches are often not supported. This is to make an intermediate match
In addition to the exact match, forward match, and backward match search indexes, a large number of intermediate match indexes are required,
This is mainly because the storage capacity of the search file becomes huge, the search time increases accordingly, and the maintenance of the search file is not easy. Also, depending on the system, due to the limitation of the size of the search file, it may not be possible to support all prefix-matching and suffix-matching searches for keywords. However, searchers often memorize characteristic characters or character strings in keywords, and therefore, it is required that the search can be smoothly performed by supporting partial match search including intermediate matches.

The present inventor has noticed that, in the case of Japanese, there is a characteristic that the same character or the same character string rarely appears in a word that can be a keyword, and the keyword constituent characters are stored for each character type. Create a search file by using the same character as the constituent characters of the search-input arbitrary character string from the search file, and speed up the search by matching the continuity of the arbitrary character string in the search file. I found that I can do it.

From the above viewpoint, the present invention can realize the speedup of partial match search using a multi-keyword for a huge database system, and is not limited to a specific hardware, and the search processing can be performed. Provides a versatile information search processing method that does not require transfer of character strings with a dedicated processor or LSI by performing on main memory, and can search arbitrary character strings by focusing on characters and character positions The purpose is to do.

[0007]

The first feature of the present invention is:
A record composed of multiple keywords is used as the search unit.
However, each keyword is a logical division of that keyword.
The attribute is defined as
The record of the keyword that matches with having a certain input character is searched.
Search for multi-keyword information retrieval processing method
In the device for creating a file, a record identification code assigning means for assigning an ascending code to each record to be searched, and an attribute code indicating a logical division of a keyword to each keyword of this record. a keyword attribute code providing means for, decomposing the keyword for each character, a character position order code providing means for providing a character position sequence code indicating the location of the keyword in each character, the record
All the keywords that the keyword has for each keyword attribute
Therefore, a fixed-length key whose starting position is the specified character position
Keyword strings that are arranged and concatenated in the keyword attribute area
Keyword sequence generation means to be generated, and the key generated above
Create a character position information shall be the Do Ri character position sequence codes the least significant information from the record identification code and keyword attribute code and character position order code and for each character constituting each keyword word sequence, the character position A means for storing information in an area for each character type to create a search file is provided.

The character position information is obtained by arranging each keyword included in the record in the keyword attribute area corresponding to the keyword attribute code, and
It is created by converting all characters of each keyword into a code consisting of an integer with a record identification code, a keyword attribute code, and a character position order code. Record identification code × n + (P _a −1) + character position order Code n: Number of characters in keyword string P _a : It is preferable to be given as a number which is the first character position in the keyword string of the keyword attribute area of the keyword attribute code a.

A second feature of the present invention is to include a search file created in the first feature, and means for extracting character position information of the same character as a constituent character of a search input character string from the search file, Character position information in which the record identification code and the keyword attribute code are common between the extracted character position information, the character position order code is in the same order as the search input character string, and the keyword attribute code is the same as the search input And a means for outputting a record identification code corresponding to the search input character string as a search result based on the combination of the extracted character position information.

Further, it is preferable to extract the combination of the character position information which can form the character string of the search input, in order from the character having the lowest appearance frequency in all the keywords of the search input.

Further, the extraction of the combination of character position information that can form the character string of the search input is performed by extracting the character position sequence code of the character having a low appearance frequency in all the keywords of the search input character string as i, and the character of the character having a high appearance frequency. When the position sequence code is j, (character position information of character with character position sequence code i)-(character position information of character with character position sequence code j) =
It is preferable to use the formula i-j.

[0012]

[Function] In a Japanese character string, the frequency of appearance of the same character is lower than that in English. Especially with regard to kanji, the frequency with which the same kanji appears repeatedly is low. For example, Kojien (published by Iwanami Shoten) has about 9 million explanatory texts for headwords. Of these, the JIS 1st level kanji appearance frequency is 1,155 average. Therefore, JIS first level 29
For 65 types of kanji, if the search input is n characters, the average of the collation targets extracted from the explanation document of the entry word of Kojien is n × 1155 characters. In general, since the search input is less than several tens of characters, even a character string including a character having a high frequency of appearance has a much smaller number of matching times than a character string in which all characters are sequentially matched.

Furthermore, the frequency of occurrence of the same character string is extremely low in Japanese character strings, especially in kanji character strings. For example, even though there are many terms that use two character strings "communication", the character string "communication ..." is the frequency of occurrence of the same four characters such as "communication line" and "communication device". Will be very low. For this reason, if you proceed with matching with the search file using the constituent characters of the search input character string, from among the search target candidate keywords that have been obtained so far,
Keywords that are different from the search input character string are deleted, and the search target is narrowed down by the constituent characters that are matched. Especially,
If the collation is performed in order from the character having the lowest appearance frequency in all the keywords in the search input, the number of collation matches can be further narrowed down and the number of collation matches can be reduced.

Therefore, the record identification code that can specify the record, the keyword attribute code that can specify the attribute of the keyword, and the character position in the keyword are specified corresponding to each character that constitutes the keyword of the record to be searched. Character position information is created from the character position sequence codes that can be created, this character position information is stored for each character type, a search file is created, and this search file is collated and matched with the search input character string. It is possible to significantly reduce the number of collation matching processes in the column search.

The search file is created as follows.

First, a record identification code is assigned to each record to be searched in ascending order according to the registration order. Next, the logical type of the keyword included in each record is used as an attribute, and a keyword attribute code indicating the attribute is assigned. For example, in the case of a book search system, a keyword attribute code indicating attributes such as a book name, an author name, a publisher name, a publication year, and an abstract is added. Furthermore, the keyword is decomposed character by character, and a character position sequence code indicating the character position from the beginning of the keyword is given to each character. Character position information of each character of the keyword is generated from these record identification code, keyword attribute code, and character position order code. At this time, the leading character position of the keyword preset for each keyword attribute code is added to the character position information as a constant so that the keyword attribute can be represented by the character position. This character position information is grouped for each character type, and these groups are collected to create a search file. Therefore, this search file has a file structure in which character position information is stored for each character type and is stored in a known recording medium.

In the search process, one or more pairs of a search input character string and a search input character string attribute are input. For each search input character string, the search input character string is decomposed character by character, character position information of the same character as the decomposed character is extracted from the search file, the record identification code is common, and the character position sequence code is the search input character string. A combination of character position information that is in the same order as the character string and whose keyword attribute code is the same as the search input is collated and extracted. A record identification code corresponding to the search input character string is extracted as a search result from the extracted character position information corresponding to each search input.

This collation processing is to check the continuity of the character string between the search input and the search file and the matching of the attributes, and the record identification code is common from the character position information in the search file. This is done by extracting keywords with the same input character position order and keyword attribute code.

As a result, collation with all keywords in the search file becomes unnecessary, and only the character position information of the same characters as the constituent characters of the search input in the search file need to be collated and matched, so the number of collation is sequentially performed. This can be extremely reduced as compared with collation. Further, in Japanese words, the frequency of occurrence of the same character string is low, and the search target is narrowed down each time character matching is performed, so the number of times matching is reduced.

Further, when collating the character position information extracted from the retrieval file, if the letters appearing in low frequency in all keywords in the retrieval input are performed in order, the retrieval object is further narrowed down, and the number of collation matching is further reduced. it can.

When the character position information matching the search input is found in this way, the record corresponding to the record identification code corresponding to the search input is extracted and output to the searcher as the search result.

[0022]

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

FIG. 1 shows the configuration of an information retrieval processing device according to an embodiment of the present invention.

The information retrieval processing apparatus of the present embodiment has a CPU 1 for performing various arithmetic processing or judgment processing, a program for retrieval processing, retrieval file creation, etc., a retrieval file created or for performing retrieval processing, retrieval input, etc. A memory 2, a keyboard 4, a display 5 for connecting an input / output unit 3, an external storage device 7 for storing various information, an external storage device control unit 6, a CPU 1, a memory 2, an input / output unit 3, an external unit Common bus 8 for connecting the storage device controller 6
With.

In the information search process of this embodiment, a search string is created by storing the constituent characters of each keyword for each character type in a keyword string created from multiple keywords included in a record to be searched for use in the search process. The search file creation process is performed, and the search process is performed to perform matching and matching with the search file to extract a record of a keyword that matches the search input.

First, the search file creation process will be described.

This search file creation processing is roughly divided into three areas: securing a search file area, adding character position information to each keyword constituent character, and storing character position information grouped by character type in a file. Can be divided into Each of these processes will be described.

Secure search file area JI
Classification is performed according to the S code table, and the appearance frequency is counted in the character type described in the JIS code table. As a result, the number of character position information registered in each character type group forming the search file can be known, so that the area of the search file composed of all character type groups can be secured. At the same time,
From the number of pieces of character position information registered in each character type group, the start address of the character type group continuously stored in the search file can be known. The first address of this character type group is JI
The character column address table shown in FIG. 2 is arranged in the order described in the S code table.

Addition of Character Position Information to Characters Constituting Each Keyword The character position information described here constitutes each keyword in a keyword string created by arranging each keyword included in a record in a keyword attribute area corresponding to a keyword attribute number. The record number indicating the registration order of the records to which all the characters belong, the character position number indicating the position where the character appears in the keyword, and the keyword attribute number indicating the logical type of the keyword.

First, the record number will be described. For example, in a general book search system, books are searched for by keywords such as book name, author name, publisher name, publication year, and abstract. At this time, the record is the book name, author name, publisher name, publication year,
It is a search target composed of keywords of the abstract, and numbers are added in ascending order from 1 to the order in which this record is registered as the record number.

Next, the keyword attribute number will be described. Generally, a searcher uses a keyword that makes it easy to specify a book to be searched as a search input or inputs a stored keyword as a search input. Therefore, in the book search system, for example, a keyword attribute is added to each keyword of a book name, an author name, a publisher name, a publication year, and an abstract, and the search input and the keyword of the book search system are logically linked. . Here, as the keyword attribute number, "1" is assigned to the book name, "2" to the author name, "3" to the publisher name, "4" to the publication year, and "5" to the abstract.

Next, the character position number will be described. Each character constituting each keyword is numbered in ascending order from the first character of each keyword in the order of 1, 2, 3, ... As a character position number.

Next, the keyword string will be described. In order to realize a partial match search for a keyword included in a record by matching a character string with a search input character string, all keywords included in the record are combined to form a character string, which is used as a keyword string. That is, each keyword is arranged in the fixed-length keyword attribute area corresponding to the keyword attribute number to create a keyword string. Thus, the attribute of the keyword to which the character belongs can be known from the character position in the keyword string. Following each keyword attribute area, the END symbol indicating the delimiter of the keyword attribute area is stored in the keyword string. This END symbol assigns an unused code in the JIS code table, and here is code A3B.
Let B.

Then, for this keyword string, all the characters that make up the keyword are converted from the record number, the keyword attribute number, and the character position number into a code consisting of integers to create character position information. This character position information is an integer code given by the following equation (1). Character position information code = record number × n + (P _a -1)
+ P (1) n: Number of characters in keyword string P _a : Position of first character in keyword string in keyword attribute area of keyword attribute number a p: Character position number

For example, the keyword attribute area size of the keyword string is as follows: book name = 64 bytes 32 characters, author name = 32 bytes 16 characters, publisher = 64 bytes 32 characters, publication year = 8 bytes 4 characters, abstract = 400 bytes In the 200-character book retrieval system, the record with the record number 100 is "Book name = Structure of correspondence", "Author name = Ichiro Tanaka", "Publisher = Taiheiyo Publishing", "Year of publication = 1990", " In case of "abstract = first person", the keyword string is as shown in FIG. At this time, since the keyword string is 578 bytes (289 characters), the character position information of each character is constructed as shown in FIG.

If the character position information is composed of 4-byte codes in this way, it is possible to handle 2 ³² ÷ 289 ≈14.86 million keyword strings each having 289 characters.

Registration of Character Position Information in Search File Next, the character position information given for each character is registered in the search file. As mentioned above, the character type group is J
The search files are stored in the order listed in the IS code table. Then, the character position information of each character is registered in each character type group. The registration of the character position information is performed by storing the character position information at the beginning of the unstored area of the character type group. Therefore, if record numbers are given in the order of registration, the character position information will be registered in the character type group in ascending numerical order.

FIG. 5 shows an example in which the character position information of the book name "Structure of communication document" is registered in the search file. At this time, the character position information in each group is stored in ascending order. If the character position information is 4 bytes,

[Equation 1] become.

The additional registration of the character position information is performed by adding a new code to the head of the unstored area of the group corresponding to each character of each keyword included in the additional record. The deletion is performed by changing the corresponding character position information in the group corresponding to each character of each keyword included in the deleted record to a special symbol. Thereby, additional registration and deletion can be performed in a short time.

As described above, the character position information stored for each character type group of the search file can be taken out as a directory at the leading address of the character field of the character field address table.

The flow of the above-mentioned search file creation processing is shown in FIGS. 6 and 7.

That is, the frequency of occurrence of character types is counted to create a character column address table (S11, 12), and a search file area is secured (S13). Next, the record registration rank counter is initialized to k = 1, the record number is set to “1”,
The number of characters in the keyword string is n = 289, the first character position of the keyword attribute area is P ₁ = 1 for the book name, P ₂ = 34 for the author name, P ₃ = 51 for the publisher name, P for the year of publication. Set ₄ = 84 and P ₅ = 89 for abstract (S1
Four). Then, the first record is taken out (S15). The processing up to this point is the pre-processing of registration. From this point, the registration process is performed for each record. First, the keyword attribute number is set to a = 1 (S16), and the keyword having the keyword attribute number a is extracted from the record (S17). Further, the number of constituent characters of the keyword is set to m, the character position number is set to p = 1, and the first character position of the keyword attribute area corresponding to the keyword attribute number _a is set to Pa (S18). Next, character position information corresponding to the character position number p is created using the formula (2) of D = k × n + (P _a −1) + p (2) in order from the first character of the extracted keyword ( S19).

Then, the character field directory (character field start address) indicating the character field of the search file in which the character type group of the character at the character position number p is stored is taken out from the character field address table (S20), and the character field is retrieved. Character position information is stored at the beginning of an unstored line in the character field of the search file indicated by the directory (S21). Then, p = p + 1 and m = m
When all the characters in the keyword are processed, the keyword attribute number is incremented by 1 with a = a + 1, and the next keyword processing is performed (S24, S25). Moreover, if all the keywords that the record has are processed, k = k + 1
Then, the record registration rank counter is incremented by 1 and the process moves to the next record (S26, S27, S28). When the processing of all records ends, the registration processing ends (S26).

Next, a search process using the search file thus created will be described.

In this embodiment, a keyword including the same character string as the search input character string is collated for the character position information extracted from the search file, and after confirming that the keyword has the same attribute as the search input, An example of searching a record common to all search input character strings will be described.

First, the search processing is roughly divided into the following configurations. The start address of the character field in the character field address table corresponding to each character of the search input character string is calculated. The search input character string is rearranged in order from the character with the lowest appearance frequency in all keywords. A corresponding character type group is extracted from the search file in order from the beginning of the rearranged character string, and a combination of character position information that matches the order of the character string of the search input is extracted from the character position information stored therein. From the extracted character position information, character position information having the same attribute as the search input is extracted. After repeating ~ for the number of search inputs, the record number common to all search input strings is output as the search result.

Next, each processing will be specifically described.

Calculation of the starting address of the character field in the character field address table corresponding to each character of the search input character string As in the case of creating the search file, the JIS of the search input character
The code table entry order is calculated and used as the address pointer of each search input character in the character field address table.

Rearrangement in the order of appearance frequency Then, the appearance frequency of the search input character is checked by referring to the start address of the character field in the character field address table indicating the start address of each character type group of the search file, and the character string of the search input is searched. Are sorted in ascending order of appearance frequency in all keywords. As described above, the starting address of the character field in the character field address table indicates the starting address of each character type group stored in the search file, and if the difference from the succeeding character field starting address is taken, From the number of character position information stored in the character type group, the character type appearance frequency in all keywords can be known.

This is because the number of collations between the character position information of each character stored in the search file can be extremely reduced by performing the collation matching from the character having the low appearance frequency. That is, when collating character position information to check character continuity, the record number, character position number, and keyword attribute number in the character position information in the two character type groups are collated. If the number of stored character position information is small, the number of collations can be reduced accordingly. Therefore, when collating character position information, collating from a character having a low appearance frequency reduces the number of collations. In particular, the larger the number of search input characters is, the higher the ratio of characters having a low appearance frequency is included.

Collation of Character Strings Character position information stored in each character type group is extracted from a character having a low appearance frequency by referring to the character field address table. And based on the extracted character position information,
A combination of character position information having the same record number and the same character position number difference between the character type groups and the character position difference of the search input character string is extracted in order from the character type group having the lowest appearance frequency.

The matching of the character position information is performed by using the following expression (where i is the character position number of a character having a low appearance frequency in all keywords in the search input character string and j is a character position number of a character having a high appearance frequency). In (3), extract character position information such that (character position information in character type group of character of character position number i)-(character position information in character type group of character of character position number j) = ij (3) Good.

In the collation processing of the character position information between the character type groups, the character continuity of the character type group having a low frequency of appearance and the character position information of the character type group having a higher frequency of occurrence are taken to determine the character continuity. Collate.

If arbitrary constituent characters in the search input character string are A and B, the character position number difference between A and B is L in order to extract the corresponding character position information. information a _x, delete delete a _x + L> B _y if B _y a _x + L <B _y if a _x when the character position information of the group B was B _y a _x + L = B _y if a _x, B _The number of collations is reduced by deleting the discontinuous character position information from the collation target such as deleting _y as a match.

For example, when the character position information of the group A is 571, 1140, 10230, 11365, 11928, 13069 and the character position information of the group B is 569, 1706, 10231, 10235, 11362, the character position number difference L = 1, The number of times of matching between these two groups is only 7 in total, and it is not necessary to match all the character position information within the group.

Collation of keyword attribute number The keyword attribute is collated with the character position number of the character position information obtained from the character string collation. That is, if the character position number is 1 to 32, the keyword attribute of the character position information is the book name, and if the character position number is 34 to 49, the keyword attribute of the character position information is the author name, and the character position number is 51 to
If it is 82, the keyword attribute of the character position information is the issuer name, if the character position number is 84 to 87, the keyword attribute of the character position information is the year of publication, and if the character position number is 89 to 288, it is the character position information. It can be seen that the keyword attribute is abstract. Therefore, from the character position information obtained by the character string matching, only the character position information having the same attribute as that specified at the time of search input is extracted.

Extracting Record Numbers By repeating (1) to (3) as many times as the number of search inputs, a record number common to all search input character strings is extracted from the obtained character position information corresponding to each search input character string. In the above embodiment, the case where the number of search inputs is one or more has been described.
In addition, when there are multiple search inputs, the explanation was given as an example of performing a logical product operation between each search input, but in the case of multiple search inputs involving logical operations other than logical product, the character position information that remains as the matching result is A designated logical operation is performed in association with each search input, and a record number satisfying the logic of the search input is output as a search result.

When one search input is designated by a plurality of character strings, it is often the case that an item having a large number of characters such as an abstract is targeted, but after the keyword attribute matching of the first character string is completed, For the second and subsequent strings,
The character position information having the record number and keyword attribute number obtained up to that point is extracted from the character type group of the first matching character of the character string, and the obtained character position information is the same as the character type group of the first character of the character string matching. Perform collation processing for other characters in the character string.

The above-mentioned operations (1) to (5) will be described with a specific example.

It is assumed that a book name is specified as the search target and "communication document" is specified as the search input character string. In this case, the attribute number of the keyword attribute of the book name is "1".

For example, the frequency of appearance of each character in all keywords is in the order of “writing” <“text” <“belt” <“communication”, and if collation is performed in this order, first, “writing” in the search file. The character position information extracted from the character column of “Sentence” and the character position information extracted from the character column of “Sentence” are expressed by the above formula (3).
When the character position information whose difference is "-1" is extracted using, the character position information "28"
"904" and "28903" in the "sentence" can be extracted as continuous character position information.

Next, the character position information remaining as a collation result in the "writing" and the character position information extracted from the character column of the search file corresponding to "shin" are similarly calculated using the equation (3). , When the character position information with the difference of “−2” is extracted, the character position information “28904” in the “call” and the character position information “28902” in the “shin” are regarded as continuous character position information. Can be extracted. Similarly, the character position information "28904" in "sho" and the character position information "289" in "tsu"
01 ”can be extracted as continuous character position information. Furthermore, since the keyword attribute is "book name", from the character position information that remains in the character string collation so far, character position information with character position numbers 1-32,
"28901" to "28904" can be extracted.

Since the number of characters in the keyword string is "289", the character position number is 28901 ÷ 289 = 100.
It turns out that it is ~ 4. You can also see that the record number to which this string belongs is 100.

This search processing operation is shown as a flowchart in FIGS. 8 and 9.

That is, the number of characters in the keyword string is n = 289.
In addition, the character position range P _a of the keyword attribute area is P ₁ = 1 to 32 for the book name, P ₂ = 34 to 49 for the author name, and P _{3 for the} publisher name.
= 51-82, year of publication is P ₄ = 84~87, abstracts P ₅ = 89~28
8, the keyword attribute number is set to a = 1 (S31),
If there is a search input character string of the keyword attribute number a, it is taken out (S32, S33). From here, the search input character string collation processing starts. Therefore, (the number of characters in the search input character string) -1 is set as the number of times of collation q (S34), and the characters are rearranged in the order of low appearance frequency of the search input characters in all the keywords (S35). Then, the character position information stored in the character type group corresponding to each rearranged character is retrieved from the search file (S36). Next, when the character position number of a character having a low appearance frequency in all keywords of the search input character string is i and the character position number of a character having a high appearance frequency is j, (character position information of the character having the character position number i)- A combination of character position information (character position information of character with character position number j) = i-j is taken out (S37). The same process is performed for the remaining characters in the search input character string (S38,
S39), the character position number from the remaining character position information retrieve only record numbers within the character position range P _a keyword attribute number a. The following equation (4) is used to extract the character position number from the character position information. (Character position information) ÷ (Number of characters in keyword string) = Record number surplus character position number (4) Records that have the search input character string as a character string and have the same keyword as the search input attribute in the processing up to this point Know the number (S40). The same process is performed up to the abstract, and the record number having the same keyword as the retrieved and input attribute is taken out (S41, S42). When all the search input character strings have been collated, the record numbers common to all the search input character strings are output as the search results for the remaining record numbers (S43).

[0066]

As described above, according to the present invention, the record identification code, the character position sequence code, and the attribute number indicating the type of the keyword to which the character belongs for each character type of the constituent characters of the keyword included in the record to be searched. A search file that stores the character position information is created, and the character position information is extracted from this search file for each character type that constitutes the character string of the search input, and the character string that matches the search input is searched.

Therefore, (1) it is possible to reduce the number of times the character string collation is performed for the search processing, and therefore high-speed search can be performed. (2) Because the search processing is performed by focusing on the characters and the character positions. Since it is possible to search by any character string and (3) high-speed search can be realized only by software without using dedicated hardware, it is possible to efficiently perform partial match search using multi-keywords on a general-purpose information processing device. (4) When used in a database system, it is not necessary to create a partial match search character string that can be used to create that search file, and it is automatically selected from the keywords of the record to be searched. Since it is possible to create a search file, it is possible to construct a database system economically and efficiently, which is an excellent effect.

[Brief description of drawings]

FIG. 1 is a configuration example of an information search processing device used in an embodiment of the present invention.

FIG. 2 is a character field address table according to the embodiment.

FIG. 3 is a keyword string of an embodiment.

FIG. 4 is an example of creating character position information according to the embodiment.

FIG. 5 is an example of registration of a search file according to the embodiment.

FIG. 6 is a flowchart illustrating a search file creation processing procedure according to the embodiment.

FIG. 7 is a flowchart illustrating a search file creation processing procedure according to the embodiment.

FIG. 8 is a flowchart illustrating a search processing procedure according to the embodiment.

FIG. 9 is a flowchart illustrating a search processing procedure according to the embodiment.

[Explanation of symbols]

 1 ... CPU 2 ... Memory 3 ... Input / output unit 4 ... Keyboard 5 ... Display 6 ... External storage device control unit 7 ... External storage device 8 ... Common bus

Claims (5)

(57) [Claims] 1. A recorder composed of a plurality of keywords
The keyword is the search unit, and each keyword is the argument for that keyword.
Attributes are set according to the
Search for keywords that match the specified input character string as the search target.
Multi-keyword information retrieval processing method to retrieve records
In the search file creation device of the above, record identification code assigning means for assigning an ascending code each time a record to be searched appears , and a keyword for assigning an attribute code indicating a logical division of the keyword to each keyword of this record Attribute code assigning means, character position order code assigning means that assigns a character position order code that indicates the position in the keyword to each character by decomposing this keyword for each character, and all keywords that the above records have Genus
The start position is the character position determined by gender.
Keyers arranged and connected in a fixed-length keyword attribute area
And a keyword string generating means for generating a keyword string and each keyword of the generated keyword string
Above each character record identification code and keyword attribute code and character position order code Do from the Ri character position sequence codes lowermost
Position by creating a character position information shall be the information, the character position information searching file creating apparatus and means for creating a search file stored in the area for each character type a. 2. The character position information is a record identification code × n + (P) for all characters forming each keyword of a keyword string in which each keyword forming a record is arranged in a keyword attribute area corresponding to the keyword attribute code. The search file creation apparatus according to claim 1, wherein _a -1) + character position sequence code n: number of characters in keyword string P _a : given as a numeric code which is the first character position in the keyword string of the keyword attribute area of the keyword attribute code a. 3. A record composed of a plurality of keywords
The keyword is the search unit, and each keyword is the argument for that keyword.
Attributes are set according to the
Search for keywords that match the specified input character string as the search target.
Multi-keyword information retrieval processing method to retrieve records
, The multiple keywords of the record to be searched are
Keyword that has a fixed character position as the start position according to the classification
For the keyword string arranged in the attribute area, the record identification code assigned to each character in ascending order for each character that constitutes each keyword, the keyword attribute code indicating the logical division of the keyword of this record, and the keyword includes a search file that the Ri character position order code Do and a character position sequence code indicating the position of a character that contains the character position information shall be the lowest information for each character type, the same as that character in the search input character string Between the character position information of the character retrieved from the search file and the character position information of each of the retrieved characters, the record identification code and the keyword attribute code are common, and the character position sequence code is in the same order as the search input character string. Means for extracting a combination of character position information that has the same keyword attribute code as the search input, and this extraction Multi keyword information retrieval processing method and means for outputting the search result record identification code corresponding to the combination search input based on the character position information. 4. The extraction of a combination of character position information that can form a character string of search input is performed in order from a character having a low appearance frequency in all keywords of search input.
Multi-keyword information retrieval processing method. 5. Extraction of a combination of character position information that can form a search input character string is performed by setting the character position sequence code of a character having a low appearance frequency in all keywords of the search input character string to i, and the character of a character having a high appearance frequency. When the position sequence code is j, character position information that matches (character position information of character of character position sequence code i)-(character position information of character of character position sequence code j) = i-j is extracted. The multi-keyword information search processing method according to claim 3 or 4.