JP2003006231A - Method and system for creation of index and retrieval of computer character information - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method and a system for creation of index and/or retrieval of computer character information. SOLUTION: Each character position in a group of retrieval object documents is determined in accordance with the order of all characters in all documents. Position data of same characters are sequentially stored in one or more database blocks corresponding to the character. Each database block is constituted of a plurality of mini blocks, each of which has storing space of a plurality of bytes. Retrieval is carried out based on these index structures. In regard to the retrieved position of a retrieval word, the document in which the retrieval word exists is determined by a document information sharing conversion part.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to indexing and retrieval of information in computer information processing,
In particular, it relates to methods and systems for indexing and retrieving computer character information.

[0002]

2. Description of the Related Art In the current full-text search of computer text information, there are two indexing methods, a character list method and a word list method. In the case of the character list method, a large storage space is required because the index is created by using the characters in the document as a search unit. In the word list method, the index is created by using the words in the document as the search unit, so the storage space used is small and the search speed is improved, but the index creation speed is slow and the search omission rate is high. Becomes higher.

Japanese Unexamined Patent Publication No. 8-235212, 8-10
1848 and 10-307841 disclose full-text search systems in which an index is created using a character list method and the index information of a document is stored by a file system. For each character in the string, the system creates a corresponding file to store the position of that character in each document. To save character position data storage space, when creating a character index, the system stores the first appearance position of each character in the corresponding index file. The first
Based on the position, a difference algorithm (for a definition of the difference algorithm, see the description of terms herein) uses each position following the first position and the preceding position to form a difference value, Sequentially stored after position 1. When performing a full text search, the second position is restored using the first position of each character and the subsequent first difference value stored in the index file. Subsequently, the third position is restored using the restored second position and the subsequent difference value. This restoration is repeated until a matching position of the characters of the search word is found. When collating frequently appearing characters such as "no" with character positions far away and backward, each position of the character is 1 from the first position to the matching target position.
You need to restore each one. For example, when a character appears 1000 times in a document, in order to acquire the 999th position, it is necessary to perform the restoration from the first difference value 998 times. Therefore, in the above-mentioned full-text search system, it takes a lot of time to restore the difference value to the position of each character in the document.

[0004]

Due to the rapid development of computer network technology, conventional full-text search systems cannot meet the ever-increasing demand for data search.

Therefore, an object of the present invention is to create an index of computer character information, to support high-speed storage and retrieval of a large amount of data, and to create an index of character information and retrieval which supports sharing of data by a plurality of users. To provide a new method and system for.

An object of the present invention is to provide a document indexing method capable of reducing the space for storing the document index and ensuring high-speed full-text search.

Another object of the present invention is to provide a method for full text search using the above index.

Another object of the present invention is to provide a method for quickly locating a corresponding document in a large number of documents according to arbitrary positions of characters belonging to the document.

[0009]

In order to achieve the above-mentioned object, the present inventors have developed a new method and system for performing full-text recording and full-text retrieval of character information. The method and system are developed in client / server mode. The features of the SQL server relational database are used to create an index of character information, which can store a large amount of data and improve the sharability, integrity and integrity of the data.
Therefore, web servers on the Internet and intranet can have a large-capacity high-speed full-text search function, and can realize comprehensive sharing of information sources.

According to the present invention, in the method of creating an index of character information, the process of determining the position of each character of a group of documents to be searched according to the order of all the characters in the entire document, and the character position data Sequentially storing in one or more database blocks corresponding to, and obtaining the maximum position and the minimum position stored in each database block, and each database block into a plurality of mini blocks each having a storage space of multiple bytes. Partitioning and obtaining the minimum position stored in each miniblock.

Further, according to the present invention, in order to search the index of each character of the search word, the process of obtaining the relative positional relationship between the characters of the search word and each database block of the character match the relative positional relationship. Each of the mini blocks of the database block has a relative positional relationship with respect to the database block that may have a position matching the relative positional relationship. For each process of determining whether there is a possibility of having a matching position, and for a miniblock that may have a position matching the relative positional relationship,
A method of searching character information based on an index of character information created according to claim 1, including a step of determining whether each position of the mini-block matches a relative positional relationship.

The present invention further provides a method for indexing and searching character information, in which a certain amount of shared memory is reserved and each field of the document information of a group of documents stored in a database is searched. A document information sharing conversion unit that reads a copy from a database into a shared memory is provided, and when performing a full-text search, a related method is directly obtained from the shared memory.

Further, according to the present invention, in a system for performing indexing and searching of character information, the position of each character in a group of documents to be searched is determined according to the order of all characters in the entire document, The position data is sequentially stored in one or more database blocks corresponding to the character, the maximum position and the minimum position stored in each database block are obtained, and each database block has a plurality of storage spaces each having a plurality of bytes. An index generation unit that divides into mini blocks and acquires the minimum position stored in each mini block, and in order to search the index of each character of the search word, acquires the relative positional relationship between each character of the search word, Determine whether each database block of characters may have a position that matches the relative positional relationship, and For a database block that may have a position that matches the positional relationship, determine whether or not each miniblock of that database block may have a position that matches the relative positional relationship. For a miniblock that may have positions that match the relationship, each position of that miniblock is
Provided is a system including a full-text search unit that determines whether or not a relative positional relationship matches, and a storage unit that stores a document and its index information.

Further, according to the present invention, in a system for performing indexing and searching of character information, each of the document information of a group of documents to be searched, which reserves a shared memory of a certain capacity and is stored in a database. There is provided a system characterized by including document information sharing conversion means for directly acquiring related document information from a shared memory when a part of fields is read from a database into the shared memory and a full-text search is performed.

Glossary of Terms In the following discussion, some common terms for databases such as lists, records, fields, etc. are used.
A list is a component of the structure of a database and consists of a number of record items. Each of this record item is
It consists of multiple fields. The following terms are used herein. These terms are explained below.

Document Category: Each document to be recorded is divided into a plurality of document categories, that is, document categories, according to its content, creator, issuance time, recording operator, main computer for recording, or each other element. being classified. Each document category consists of multiple documents.

Documents: editorials, novels, news articles, patent specifications, etc.

Characters: Characters mentioned in this specification include letters (including alphabetic characters, kanji, Japanese characters, various characters, 1-byte or 2-byte characters such as hiragana and katakana), punctuation marks, numbers and special characters. And tabs etc. are included.

Character Internal Code: In different operating systems, the code standards for double-byte characters are different. For example, the Japanese code standard on the WINDOWS (registered trademark) platform is Shift JIS (8-bit Japanese code standard in Macintosh and DOS-V).
Is. The Japanese code standard on the UNIX platform is EUC (extended UNIX code). In order to record documents from different platforms in the system of the present invention, the present inventors have used the internal code system to specify JIS (Japanese Industrial Standard) of the same character, Shift JIS (Macintosh and DOS-V).
Converts various standard codes such as 8-bit Japanese code standard in) or EUC (extended UNIX code) into only one corresponding internal code.

Character position: When a document is recorded according to the recording order of each document in one document category and the order of characters in each document, each character in the document is given an absolute position in the document category. For example, the first in the document category
The character position of the character is 1, and the character positions of the following characters are 2, 3 ,. ． ． Is.

Difference value: In one document category, the difference value between two character positions is calculated by the difference algorithm based on the current character position and the previous character position of the character.

Difference Algorithm: In the document category, the difference between the current position of a character and the previous position of the character is from a base with a small base to a base with a large base such as decimal to 127. To be converted. In order to identify each difference value, the difference value at each character position is obtained by setting the most significant bit of all digits except the final digit (unit digit) of each 127-adic difference value to 1. For example, the decimal value 2048383 is converted to a 127 decimal value as shown below.

(2048383) 10 = (0x01 0x00 0x00 0x00) 127 The resulting 127-ary value has four digits, each of which is 1
It is represented by a hexadecimal value. The remaining 3 most significant bits of the 4 except the last digit 0 are set to 1 (ie 0
Logical sum with x80 is required). Therefore, the obtained difference value is (81808000) 16.

Since the position of the character in the document varies, the difference value has various values, and the number of bytes of each difference value also differs. In the difference value, the most significant bit of the last digit (8th bit of 1 byte of the last digit) is 0, and the most significant bits of the other digits are all set to 1, so each difference value is identified. be able to. For example, two consecutive positions of the letter "day" in the document category are 1390 and 1450.
Suppose The difference between these two positions is ((1450-139
0)% 127) 10 = (60) 10 = (3C) 16. Only one byte is needed to store this difference value. In another example, assume that the two consecutive positions of the character "good" in the document category are 1308 and 9054. The difference value of the characters at this time is calculated as follows: Step 1: When calculating ((9054-1308) / 127) 10, the quotient is (60) 10 and the remainder is (126) 10 = (7E) 16.

Step 2: When (60/127) 10 is calculated,
The quotient is (0) 10 and the remainder is (60) 10 = (3C) 16.

Regarding the 127-ary value obtained in steps 1 and 2, the most significant bit of the unit digit (7E) 16 does not change. The most significant bit of the other digit (3C) 16 is set to 1, ie (3C) 16 to (BC) 16. This difference value (BC7E) 16
Requires 2 bytes to store.

Restoration: When performing a full-text search, it is necessary to restore the difference value and acquire the character position of each character in the corresponding document category. The algorithm for restoring the difference value of the character position is the reverse of the difference algorithm.

Miniblock: A miniblock is a structure defined by the present inventors for storing data, and is an array for storing data items for storing character positions and a plurality of difference values of the character positions. Consists of. For example, a miniblock comprises a long integer type data item and a BYTE type array to store multiple position data of one character. Here, the long integer type item is used to store one character position of a character, and the BYTE type array is each character position of the character following the character position stored in the long integer type item. Used to store the difference value of.

Database block: A database block is a physical area in the database for storing data and is a field in the list structure.
A database block consists of multiple miniblocks and can store multiple position values for a character. Each record item in the list consists of one database block and two fields for storing the minimum and maximum positions of the characters stored in that database block, respectively. These two fields allow a quick determination when performing a search whether the database block of the current record item may have a character position that matches other characters in the search term. be able to. If it is determined that there may be a matching character position in the current record item, further decompression processing is performed on that database block. If it is determined that there is no matching character position in the current record item, the record item next to the character is determined / restored.

Search word: A search word is a character string consisting of one or more characters. The search word is a character string to be searched,
Specified by the operator.

Matching position: Since the index of each document is stored together with each character as a unit, the position of the character in the search word is designated as the position of the entire search word, that is, the matching position and stored in the database. Collation processing is performed on the character position information. For example, in the search word “US America”, the character position of “US” is 10001, the character position of “country” is 10002, the character position of “A” is 10003, and the character position of “ME” is 100
04, the character position of "ri" is 10005,
It is assumed that the character position of "F" is 10006. When the character position of the first character of the search word is used as the matching position of the search word, the matching position of the search word "US America" is
It is the character position 10001 of “rice”.

Displacement: The displacement is the offset amount with respect to the matching position of each character of the search word. For example, if the first character of the search term "US America" is set as the starting point,
The displacement of "rice" is 0, the displacement of "country" is -1,
The displacement of "A" is -2, the displacement of "M" is -3, the displacement of "L" is -4, and the displacement of "F" is -5. Also, the last character of the search term can be set as the starting point, in which case the displacement of "rice" is 0,
The displacement of “Country” is 1, the displacement of “A” is 2,
The displacement of "M" is 3, the displacement of "L" is 4,
The displacement of “F” is 5.

As mentioned above, the present invention records / stores all documents in multiple document categories. Each document category consists of multiple documents. Each character in the document is stored according to the recording order of each document.

From the recorded document, document category information,
Document information, character information and character position information are extracted and stored in the database. The four types of information will be described below.

Document category information: According to the content, each document can be classified into various document categories such as politics, economy, sports, and travel. The recorded documents can also be classified into different document categories according to the recording operator or other factors. For each document category, the present invention specifies a document category number that can uniquely identify the corresponding document category. Further, each document category has a document category name and a final position of each character in the document category. The final position of each character in the document category is the character position of the final character in the document most recently recorded in the current document category. Therefore, the document category information includes the document category number, the document category name, and the final position of each character.

Document information: Since each document category has a plurality of documents, a document number is given to each document to be recorded. The document number is unique in one document category, and may be a long integer type value starting continuously from 1 or an intermittent integer value.

Therefore, the document information includes the unique document number of each document, the document category number to which the document belongs, the character positions of the first character and the last character in the document (in the following description, these two The position includes the start position and the end position of the document).

Character position information: Each document category includes a plurality of documents, and each document includes a plurality of characters. A character set is composed of all unique characters. Each character in the character set has a corresponding internal code. Since each character may appear multiple times in each document in the document category, all character positions of all characters in each document category are stored in the database. The character position information is a document category number to which the character belongs, an internal code corresponding to the character, a database block for storing character position data (including the character position and the difference value),
It contains the first and last character positions (these two positions are hereinafter referred to as the minimum and maximum positions of the database block) stored in this database block.

Character information: In order to speed up the search, the present invention stores the character information of each character in each document category in a database. The character information is stored in the document category number to which the character belongs, the corresponding internal code, the maximum character position of the character in the document category to which the character belongs, the minimum character position in the final database block of the character, and the final database block of the character. Contains the length of the data.

In the present invention, the index of the character position of all the characters in the document category is created together with each character as a unit. That is, each character position of each character of the character set in each document of the document category is stored in the database.

Characters such as the Chinese character "target" and the Japanese character "no" are frequently used. In order to reduce the space for storing the character position information of a character, in the method of the present invention, the position information of each character in each document of the document category is used as the first character position and each subsequent character position in this document category. expressed. Further, this subsequent position is represented as a difference value between two consecutive character positions. Actually, in terms of physical structure, each difference value between the first character position and the succeeding character position is stored in the database block. Since the difference value at each position has a smaller number of bytes used than the actual character position value, the space used by the index of each character can be relatively reduced.

By representing each position following the first character position of the character as a difference value, the space used is reduced,
Search speed will be reduced. For characters that appear very often in the document category, there are many difference values. For example, a character is 100,0 in the document category.
When it appears 00 times, it takes a lot of time to restore the 90,000th character position by using the difference value between the first character position and each subsequent character position.

In order to solve this problem, in the present invention,
Improvements to JetSearch, a full-text search system that uses system files to store character position data. In the present invention, the character position data is stored as a plurality of parts instead of one file. In the database 104, each record item has a field of a database block for storing position data of characters in the document category to which it belongs. A character in a document category may have multiple database blocks to store its position data.

When storing the position data of a frequently used character, there will be many position data corresponding to that character. When a search is performed, character positions satisfying the matching condition of a character that appears many times (for example, when searching a document including NBA in the document category of sports news, the first N is a document of sports news through the search). If it is known that it is the 10,000th character of the category, B and A before the 10,000th character do not meet the matching condition), and the system searches for the corresponding position. There is a need for quick and reliable judgment / search. When performing a search for database blocks that store difference values continuously, character positions are searched one by one from the first difference value of each database block to the character position that satisfies the matching condition of each character of the search word. Needs to be restored. This wastes system time and is not a good idea.

Therefore, the inventor has divided the database block into a number of mini-blocks, each of these mini-blocks having a certain number of bytes for storing position data. The start byte of each miniblock stores the character position of the character in the document category. The character position of this character (referred to as the minimum position of the mini-block in the following description) is determined by the order of all the characters in the document category, not by the difference value, and differs from the maximum position and the minimum position of the database block. ing. Each subsequent byte of the start byte of the miniblock stores the difference value between two consecutive character positions. The character positions of the characters are different from each other, and the difference values are also different values. The number of bytes used for each difference value also varies. If the remaining bytes of the miniblock are not enough for the new difference value, the system fills the remaining bytes of the miniblock with 0x00 and uses another new miniblock to generate the minimum of this new miniblock. Store the new difference value in the start byte as a position. Subsequently, each succeeding character position is stored as a difference value in a new mini block. In other words, after representing each character position of a character with a certain number of difference values, a new miniblock is needed, the minimum position is defined for this new miniblock, and the subsequent character positions are the difference values. expressed. If each miniblock in the database block is full of character position data, create a new record item for that character and continue storing position data for the new database block. There is a need.

If the set mini-block length is too long, the restoration will take a lot of time. If the set miniblock length is too short, a relatively large space is used. Therefore, for miniblocks, it is preferable to be hundreds of bytes long. In other words,
It is appropriate to use a new miniblock for every 100 to 200 occurrences of the character.

However, commonly used characters can occur more than one million times in a document category. Therefore, according to the present invention, the number of occurrences of characters is 400,000 to 8 in one document category.
It is configured to create a new record item and store the subsequent position data of the character every 0,000 times (e.g., 4000 miniblocks are full).
The record item should also contain the minimum and maximum positions of the database block for this item.

As described above, when constructing a database of character position data, the record item for storing the character position data is the record item for storing the character position data from the viewpoint of shortening the time used for searching characters. It has two overlapping fields for storing the minimum and maximum character positions of the characters in the document category stored in the database block, namely the minimum and maximum positions of the database block. When performing a full-text search, if all the records for storing the character position data of all the characters of the search word are found from the database, each has a database block for storing the character position data. It is certain that there will be many record items. Each database block consists of thousands of miniblocks, each having a capacity of hundreds of bytes for storing character position data (including the miniblock's minimum position and each difference value). If the record item does not have a minimum and maximum position for a database block, the minimum positions of each miniblock are compared one by one in each database block of multiple record items until a matching character position for each character of the search term is found. Takes a very long time. However, when two fields are added to store the minimum position and the maximum position of the database block in one record item, the character position of the character of the search word in the document category can be known.
The character position of this character can be used to compare the minimum and maximum positions of each database block of position data for other characters of the search term. Therefore, it is determined whether or not there is a possibility that the character position to be searched exists in a certain database block. After the database block is determined, it is possible to quickly determine which miniblock may have the character position to be searched based on the minimum position of each miniblock in the determined database block. Therefore, the use time is reduced as compared with the method of restoring the character position of each character one by one.

In summary, each character index of the present invention is divided into three levels.

First, each character may have multiple record items. Each record item includes a database block for storing character position data and the minimum and maximum positions of the characters stored in this database block.

Next, in order to store the character position data, each database block which stores the character position data,
Contains thousands of miniblocks. Each mini-block consists of two parts, the first part stores the first character position of each character stored in the mini-block, i.e. the minimum character position of the mini-block, and the second part follows. Each character position of, that is, the difference value is stored.

Third, each subsequent character position of a character in the document category is represented by the difference value between the character position and the preceding character position.

[0053]

Embodiments of the present invention will be described below with reference to the drawings. Obviously, the spirit of the present invention is not limited to the following examples. In the following examples, a client / server network structure is used as an example as a hardware platform for implementing the method of the present invention. That is, the client is an example of the terminal operating means, and the server is an example of the information storing / processing means. Further, the networks described in the following embodiments should be understood as an example of connecting means. It will be apparent to those skilled in the art that the following teachings of the present invention allow the method of the present invention to be implemented on other computer systems such as a single computer, browser / browser server, etc.

The client (terminal operating means) 101 is
It is a platform for operators, and requests for recording, updating or retrieving documents can be made from the server 10 from here.
3 is sent.

The network (connecting means) 102 is for transmitting client / server information.

The server (information storage / processing means) 103 is
It is for receiving a request for recording, updating or retrieving a document transmitted by the client 101 via the network 102. Document index generator 10
At the time of the processing of 5, the position information of all the characters in all the documents is stored in the database 104. Further, the full-text search engine 106 searches for document information that satisfies the search request designated by the operator.

FIG. 1B shows a hardware block diagram of the server 103. In FIG. 1B, the CPU 1 is a RAM
The program stored in 3 is executed to control each server. The ROM 2 stores a program for implementing the processing shown in each flowchart, and this program is executed by the CPU 1. RAM3 is CPU
1 provides space for executing the program. CR
T4 displays under the control of the CPU 1. The keyboard 5 is used to enter information. External storage device 6
Is a hard disk or a soft disk for storing the document to be searched and the character index information generated from this document. The bus 7 connects the above-mentioned parts,
Realizes data transmission between each unit.

The database 104 is for storing full-text index data of all document categories and other types of document information data, and the external storage device 6
It is provided in.

The document index generator 105 is for recording a document in the database 104 according to the method of the character list, and is performed by the CPU 1.

The full-text search engine 106 is for carrying out a full-text search, and is executed by the CPU 1.

The document information sharing conversion unit 107 stores the information of the recorded document in the memory RAM 3 of the server.
In order to provide a block of shared memory in, the CPU 1 executes. When generating the document index, the document information in the shared memory is updated by the document index generating unit 105 at a good timing. When performing a full-text search, the full-text search engine 10
6 directly acquires information related to the document from the shared memory.

In the client / server network structure, the system of the present invention is executed on the server, and the document index generating unit 105, the full-text search engine 106,
The document information sharing conversion unit 107 and the database 104 are mainly included.

The database 104 is used to store a plurality of document categories. One or more documents can be recorded in the designated document category by the document index generation unit 105, and the full-text index of the document category can be created or updated. When recording a document, the document index generation unit 105 converts each character in the document into a corresponding internal code, and stores the position information of each character in the document category to which it belongs.

Full-text search according to search conditions and fuzzy values (a value of 0 is a strict search, a value of more than 0 indicates a fuzzy search, and a high fuzzy value indicates that the matching accuracy is low and there are many search results). When performing the search, the full-text search engine 106 searches for related record items in the database 104, compares the positions of the characters in the search word, finds a character string that matches the search condition, and searches for the search word in all documents. Returns the number of the document containing and the position of the search term in each document.

Since a large amount of text information is stored in the database 104, it takes a very long database processing time to perform a full text search. In order to reduce the input / output of the database 104, shorten the search time, and improve the system performance, the present inventors have
A document information sharing conversion unit 107 that functions so as to leave a shared memory block in the memory RAM 3 is designed. continue,
A list of document information in the database 104 is searched. The data items (document number, start position and end position of the document, and deletion flag) of all documents of the specified document category that represent the document number and the position range of the document are read from the database 104 into the shared memory according to the recording in order. Will be resident in it. When performing a full-text search, a binary algorithm is used to convert the document position determined according to the matching position of the search word into the corresponding document number. When recording a new document, information about the document is added to the shared memory. When the document is deleted, the deletion flag of the document is set to 1 (that is, deleted).

The features of the present invention are mainly in the following three points.

The first point is the structure of the index of the character position information in the database 104. Since a character can appear multiple times in a document category, all positions of the character in the document category need to be stored.
If each position of a character is stored in the database as a field of integer type or long integer type (each position uses 4 bytes), a huge storage space is required. In order to reduce wasted storage space and to allow full-text search at a desired speed, the inventor uses a difference algorithm to determine the current character position and the previous character position of a character in a document category. The difference value between the character position and the character position is calculated. Each subsequent position of that character in the document category is shown as a difference value and stored in the database as a binary (image type) field.

In the present invention, the field in each record that stores the position of a character in the document category to which it belongs is called a database block. Each database block is divided into 4,000 miniblocks. Each miniblock has 260 bytes for storing character position data. In FIG. 2, the first 4 bytes of each miniblock store the character position of the character in the document category to which it belongs. Each byte after the fifth byte (including the fifth byte) stores the difference value calculated by the difference algorithm between two consecutive character positions of the character (the current character position and the previous character position). If the remaining bytes of the miniblock are not enough for the new difference value, the system fills the remaining bytes with 0x00. the system,
Use the new miniblock to store the current character position of a character in the document category to which it belongs as the minimum position of this miniblock in its first 4 bytes, and represent each subsequent position of the character with a difference value. . In other words, the difference value is used to position the character several times (for example, 100 times to 200 times).
Times), a new miniblock for storage is needed and a minimum position is given for this miniblock.
Each subsequent position of the character is then represented as a difference value and stored after the minimum position in the new miniblock. If the 4,000 miniblocks of a database block are all filled with character position data, then a new record entry needs to be created for that character.

The second point is as follows. With respect to the character position information stored in the character index structure of the database 104, in order to obtain the full-text search result quickly and accurately, the present inventors design a method for searching / matching the character position. With respect to the features of the character index structure of the present invention, the method of the present invention uses a query language to quickly locate match words. Then, the document information sharing conversion unit acquires the document number of the document in which each retrieved result exists.

Thirdly, in the case of storing a large amount of data,
Frequent database input / output, database execution performance degradation, and
There are problems that the access time to the database increases and the full-text search speed decreases. In order to reduce the execution load of the database, reduce the time for searching the database, and increase the speed of full-text search, the present inventors have
Design a method to quickly convert character position information to document information. Regarding the features of high-speed access of cache memory,
The data item of the document information for the designated document category in the database 104, which includes the document number, the document position range, and the document number, the document start position, the document end position, and the deletion flag, is read into the memory RAM 3 at once. Be done. By using the dichotomy while following the position range of each document for any character position, the matching document data in memory is quickly located and the document number is obtained. When recording and deleting a document, the document information sharing conversion unit also provides a corresponding interface for updating related document information in a timely manner.

Description of Each Embodiment First, an execution procedure of the system of the present invention on the network 102 will be exemplarily described.

During the procedure in which the full-text search system on the server 103 processes a character search request from a user via the client 101, the following processing is performed: The character or word to be searched is input by the user via the keyboard of the client 101. OR, AND, or NO when two or more characters or words are searched
A logical relationship between letters or words such as T should be given.

2. The given search word and logical relation are transmitted via the network 102 to the full-text search system running on the server 103.

3. On the server 103, the full-text search engine 106 of the full-text search system searches the document index according to each search word to process the received search result and acquire all relevant matching positions from the database 104.

4. The document information sharing conversion unit 107 obtains the document number (document number, document start position, document end position) by the document information sharing conversion unit 107 so as to obtain the document number and the deletion flag of the document in which each of the obtained match positions exists. And the deletion flag) and the start position and the end position of the sentence, respectively. A valid document number is determined according to the delete flag of each document, and this document number forms the result set as the output of the full-text search system.

5. This search result set is sent to the client 1 by the full-text search system via the network 102.
01, and is displayed on the screen by the client 101.

Document Index Creation Process In the present invention,
A method for creating a document index is provided in which a difference value is calculated from two consecutive character positions of the characters of the document category to which it belongs by a difference algorithm and the difference value is stored in the image type database block of the database 104. Such a database block is composed of, for example, 4,000 mini blocks each of which is composed of a long integer type numerical value (4 bytes) and 256 bytes. See FIG. 2 for the structure of the database block. The database block that stores the character position data is therefore
1,040,000 bytes ((256 + 4) * 4000 = 1,040,
000).

When defining a database block, the number of mini blocks included in this can be changed, and the size of each mini block can also be changed.
The document index generation unit 105 for the recorded document
About the process of creating a document index by
This will be described below with reference to FIG. This process
When the database 104 is first created, a document is added to the database 104, or a document is deleted from the database 104, the document index generation unit 105 shown in FIG. 1 is used.

First, according to the contents of each document, the operator who records the document, or each other element, the document to be recorded is classified into the corresponding document category in advance. A document category name is given to each document category. In step 402, for example, a dialog box having an input box is displayed to prompt the operator to perform an input process. The operator inputs the document category name to which the document to be recorded belongs.

In step 404, step 402
The database 104 is searched for related document category information in accordance with the document category name designated by.
If the specified document category name does not exist in the database 104, the system assigns a document category number for that document category and sets an initial value for the final position of all characters in the document category (eg, document The category number is 1 and the final position of each character is 0). Then, the information of this document category is inserted into the database 104,
The document category number and final position are returned. When the specified document category name exists in the database 104, the document category number and the final positions of all characters in the document category are acquired from the database 104. According to the acquired document category number, the character information of each character in the database 104, that is, the maximum character position of the character in the document category, the minimum position of the final database block of the character, and the character position data stored in the final database block. The length etc. is searched.

In step 406, the document information sharing conversion unit 107 is started based on the high speed access function of the cache memory. It is determined whether the document information of the designated document category exists in the shared memory of the RAM 3.
If the document information of the specified document category does not exist in the shared memory, a certain amount of shared memory in the memory RAM3
used. Data items related to the document number of each document and the document position range recorded in the specified document category (document number, start position and end position of document of document category, deletion flag, etc.) are read from the database 104 into the shared memory. Be done. Multiple users can use it at the same time to search for a specified document category.

In step 408, the reserved memory space of RAM3 is applied to the operating system and initialized.

In step 410, it is judged whether or not the document to be recorded exists. If the document to be recorded exists, the process proceeds to step 412. If the document to be recorded does not exist, steps 420, 430 and 432
Is performed and the related information in the database 104 is updated.

In step 412, the information of the document to be recorded is read. The document information in the database 104 is searched according to the document number. If the document number is not found in the database 104, the document information is stored in the document information in the database 104. The document information includes the document category number to which the document belongs, the document number,
It includes the character positions of the first and last characters in the document (start position and end position of the document). If the document number is found in the document information of the database 104, it means that the document number is stored in the document information of the database 104, and an error code is returned.

At step 414, it is checked whether there are unprocessed characters in the document being recorded. If there are unprocessed characters in the document, then the process of step 416 is performed to index the characters.
Upon completion of processing the last character in the document, step 426.
Is executed, and the end position of the document stored in the document information of the database 104 is updated with the character position of the final character.

In step 416, the characters in the document are sequentially read and converted to the corresponding internal code (eg, Shift JI used in the WINDOWS system).
The S code is converted to the internal code of the system. When converting "America" to the internal code, "A",
The internal codes of "me,""ri," and "ka" are:
283, 341, 347 and 288). The character position of the character is obtained from the final position of the document category to which it belongs. Subsequently, the difference value between the current character position and the previous character position is calculated by the difference algorithm.

In step 418, the memory RAM3
Is checked to see if the remaining portion of the reserved space of is sufficient to store the difference value of step 416. If the reserved space of memory is full, steps 420, 422.
And 424 are executed so that all data of the characters in the reserved space of the memory are written to the character position information of the database 104. If the memory reserved space is not full, the flow chart proceeds to step 424.

In step 420, the memory RAM3
All of the document category number to which the character belongs, the corresponding internal code, the database block for storing multiple difference values, the minimum position and the maximum position stored in the character database block, etc. in the reserved space of The character position information is stored in the character position information of the database 104.

In step 422, the database 1
When the storage of all character position information in the reserved space of the memory RAM 3 in the character position information 04 is successfully completed, the position information of each character in the document being recorded can be continuously stored. , The reserved space of memory is reinitialized.

In step 424, step 416
The character position or the difference value acquired in step 3 is stored in the reserved space of the memory RAM 3. Then, returning to step 414, the next character in the document being recorded is extracted.

When all the characters in the document being recorded are processed in step 426, the character position of the last character in the document is stored in the document information of the database 104. That is, the end position of the document in the document information of the database 104 is updated with the character position of the last character of the document.

In step 428, the document information sharing converter 107 is used to store the information of the document to be recorded in the shared memory.

When the recording of the new document is completed in step 430, the character information of the database 104 is changed to the number of the document category to which the character belongs, the corresponding internal code, the maximum character position of the character in the document category to which the character belongs, and the character. Is updated with the newest character information of each character including the minimum position of the final database block that stores the position data and the number of bytes of the character position data stored in the final database block.

In step 432, the final position of the document category stored in the document category information of the database 104 is updated using the character position of the final character in the final document just recorded.

Embodiment 1: In the WINDOWS platform, the document (text file) America1. The txt is recorded in the database 104 and an index is created for each character in the document. The contents of the document are as follows: United States America United States This document consists of 14 characters, 5 Chinese characters (actually 4 Chinese characters excluding duplications), 8 Katakana (4 actual Katakanas without duplication) , Including one blank. "Rice", "Go",
Each of the characters "Zou" and white space appears once in the document. Each of "country", "a", "me", "ri" and "mosquito" appears twice in the document.

The document category name to which this document belongs is a news category, the document category number is 1, and this document category is a new document category in which no document is recorded. In this document category, the document number of the above document is 1, the document name is America 1, and the document name is 1999.8.10. This document is supplied to and recorded in the document index generation unit 105 as the first document in the news document category.

1.14 characters: The entire contents of the document including "US America USA" are read into the memory RAM3. Document category number 1, document number 1, document name A
merica1, at the time of publication 1998.8.10, and the start character position of this document in the news document category,
The database 10 by the document index generation unit 105
It is stored in the document information of the news document category in 4.

2. Document America1. 8 in txt
The three different characters are converted one by one from the shift JIS code to the system internal code. For a blank in the document, the document index generation unit 105 can determine whether to process the blank in the document according to a predetermined parameter. The default value of the parameter is
No index is created for white space. Document Amer
ica1. In txt, the seventh character is blank.
In the first embodiment, it is assumed that the parameter is set to a default value, that is, it is set not to create an index for a blank space. Therefore, the document index generation unit does not process the blank. As a result, the value of the character position of the eighth character and each subsequent character is decreased by one (see Table 1). 3. The start position of each document category is 1, and 1 is used as one stage. The character position of each character is determined according to the order of the characters in the document category. In this embodiment, the document America1. The character positions of all the characters of txt are shown in Table 1. Table 2 shows the correspondence relationship between each character and the first 4 bytes in each mini-block after the processing by the document index generation unit 105 is completed. 4. If a character appears twice in the news document category, the difference value between the current character position and the previous character position is calculated by the difference algorithm, and the fifth byte of the character's miniblock and each subsequent character block are calculated. Sequentially stored in bytes. For example, the document America1. In txt, the eighth character, the ninth character, the tenth character, and the eleventh character “America” are duplicated. In the following, the character "a" in the document will be taken as an example for detailed description. The character "a" appears twice in the document at character positions 3 and 7 in the news document category. As can be seen in column 5 of Table 3, the position data stored in the miniblock is
It is 0x000003004. Byte order in miniblock: 1 2 3 4 5 Position data: 0x00 00 00 00 03 04 The position data stored in the first 4 bytes of the miniblock is 0x00000003, which is the hexadecimal number, which is the news document. The first of this letter in the category
The character position, (03) 10 = (03) 16. The data of the 5th byte of the mini block is 0x04. This is the difference value between the second character position and the first character position of the character "A", and is (07-03) 10 = (4) 16. The calculation of the difference value between the three characters "Merika" is similar to that for "A". The difference value between the second character position and the first character position of each of these three characters is also 4. The first character position and the second character position of the kanji "country" are 2 and 13, and the difference value is (13-
It is calculated as 2) 10 = (11) 10 = (0B) 16. The difference values of the four katakana “America” and the Chinese character “country” are shown in column 5 of Table 3. The bold letters indicate the difference between Table 3 and Table 2.

If a character frequently appears in the record being processed, the size of the position data of the character may exceed the size of the miniblock. In this case, multiple miniblocks are required to store the character position data.
Since the number of characters of the document of this embodiment is small, the document index generation unit 105 stores each character “US”, “country”, “A” to store each position data (minimum position and difference value of mini block). , "Me", "ri", "mosquito",
Supply only one miniblock to "Go" and "Pop." 5. After the processing of each character is completed, the related information of all the characters is written in the character information and the character position information of the database 104, respectively. Each database block in the database 104 stores the character position of each character in the news document category. Data is stored in each miniblock according to its order in the database block. When a miniblock is full of data, the next miniblock is used to store data, and this process is done until all 4,000 miniblocks in the database block are filled with data.

Taking the character "a" as an example, the document category number 1, the internal code 283, the position data 0x000000304 stored in the mini block,
The minimum position 3 and the maximum position 7 of the position data stored in the database block are stored in the character position information of the database 104. Document category number 1, internal code 283, and final position 7 in the news document category
The minimum position 3 and the maximum position 7 of the position data stored in the database block, and the number of bytes in the mini block occupied by the position data stored in the mini block are stored in the character information of the database 104.

6. The final character position in the document category information and the end position of the recorded document in the document information in the database 104 are updated. In the first embodiment, the final character position in the document category information in the database 104 is updated to 13, and the document America1. txt
The end position in the document information of is 13.

7. Information related to the news document category
It is stored in the document category information of the database 104.
Document America1. Related information of txt is stored in the document information. The data in columns 2, 3, 4 and 5 of Table 3 are
It is stored in the character position information. The data in columns 2, 3 and 4 of Table 3 and the length (in bytes) of the position data in column 5 stored in the database block are stored in the character information.

Example 2 In this example, it is assumed that a document is recorded after recording the document of Example 1 and a plurality of other documents.

In the WINDOWS platform, in the present embodiment, the document (text file) America2. t
The xt is recorded in the database 104 and an index is created for each character in the document. The contents of the document are as follows: United States United States America This document consists of 14 characters, 5 Chinese characters (actually 4 Chinese characters excluding duplications), 8 Katakana (4 actual Katakanas without duplication) , Including one blank.
The characters "U.S.A.", "Go", "Popular" and the space are
Occurs once in a document. "Country", "A", "Me",
Each of "li" and "mosquito" appears twice in the document.

It is assumed that the document category name to which this document belongs is a news category and the document category number is 1, and a plurality of documents are recorded in this document category.
Also, the final position of each character in this document category is
39491, the document number of the above document is 130011, the document name is America2, and at the time of publication, 1999.8.
It is assumed to be 11. This document is supplied to and recorded in the document index generation unit 105 as the first document in the news document category. In the database 104, the maximum character position of katakana “ri” in the news document category is 8237, and the final database block for storing the position data of this character stores 258-byte position data. The maximum character position of the character “Zou” in the news document category is 1320, and the final database block for storing the position data of this character stores 25 bytes of position data. Next, this document is supplied to the document index generation unit 105 and is recorded during processing.

1. To find the current final position in the news document category and the maximum character position for each character in this category, the document index generator 105 searches the document category information in the database 104. In the second embodiment, as a result of searching the database 104, the final position of each character in the news document category is 30491,
The maximum character position of katakana “ri” in the news document category is 8237. In addition, the database block of the last record item that stores the position data of "ri" is 2
It stores position data of 58 bytes, the maximum character position of "Public" in the news document category is 1320, and the database block of the last record item that stores the position data of "Li" is a position of 25 bytes. Stores data.

2. The entire content of the document is read into the memory RAM3. The document index generation unit 105 stores the document number, the document name, the creator, and the issuing time in the document information of the database 104. In this example, the system is 14
Read the characters "United States United States America". The document number 130011, the document name America2, the time of publication 1999.8.11, and the start character position 30492 of this document in the news document category are stored in the document information of the news document category of the database 104 by the document index generation unit 105. .

3. Document America2. 1 in txt
The three characters are converted one by one from the shift JIS code to the corresponding internal code of the system. For a blank in the document, the document index generation unit 105 can determine whether to process the blank in the document according to a predetermined parameter. In the first embodiment, the parameters are set to default values. That is, no index is created for blanks. Document Am
erica2. In txt, the eighth character is blank. The document index generator does not process white space. As a result, the value of the character position of each of the ninth character and the following characters is decreased by one (see Table 4). 4. The final position of the document category + 1 is used as the starting position of the new document to be recorded, and the step increment is 1. The character position of each character is determined according to the order of the characters in the document category. In this example, the document America2.
The start position of txt is 30492. American
a2. The character positions of all the characters in txt are shown in Table 4.

5. Based on the maximum position of a certain character (described in item 1 of this embodiment) and the current character position of that character in the document category before the recording of a new document, the difference value of the character is
Calculated by the difference algorithm. If the current difference value of a character requires multiple bytes and the remaining bytes of the current miniblock's 260 bytes (4 + 256 bytes) are not enough for the new difference value, the system will use 0x00 to Fill the remaining bytes of the miniblock. Subsequently, the new miniblock is used. The current position of the character in the document category is stored in the first 4 bytes as the minimum position of the new miniblock. If this character appears thereafter, the difference value between each character position and the previous character position is stored in the fifth byte and each subsequent byte. The current character position is stored as the maximum position for that character in that document category. If the miniblock is filled with position data, a new miniblock is used and the current character position of the character is
It is stored in it as the minimum position of this miniblock. The above process is repeated until all miniblocks in the database block are filled with data. In the second embodiment, katakana "ri" and kanji "shu" are taken as examples. As mentioned above, database 1
In 04, the maximum position of "ri" in the news document category is 8237, and the final database block for storing the position data of this katakana stores the position data of 258 bytes. As can be seen in Table 4, the document America2. When txt is recorded, the character position of the first character "ri" in this document is 3049.
4, and the maximum character position of Katakana “ri” in the news document category stored in the character position information of the database 104 is 8237. According to the difference algorithm, the difference value between 30494 and 8237 is 0x81.
B020, and this storage requires 3 bytes.
To maintain the integrity of each miniblock in the database block, the 259th and 260th bytes of the first miniblock are filled with 0x00, and the position data of the characters appearing in this document are stored in the second miniblock. Stored in. Character position of "ri" in news document category 30
494 (0x771E) is stored in the first to fourth bytes of this mini block as the minimum position of the second mini block. The character position of the second appearance of "ri" in the document is 30503. The corresponding difference value at this time is 0x09, and this value is stored in the fifth byte of the second mini block. Please refer to the bold face of the katakana "ri" line in Table 5. Further, as described above, the maximum position of the character “Population” in the news document category is 1320,
The first mini-block for storing the character position data stores the 25-byte position data. The character "Zou" appears once in this document and its character position is 30497 in the News Documents category. The difference value is calculated as 0x81E65E based on the current character position 30497 of this character and the maximum character position 1320 of this character in the news document category retrieved from the character position information of the database 104. This value requires 3 bytes to store. This difference value is stored in the 26th byte, the 27th byte and the 28th byte of the mini-block of the character "Popular". See the row for the letters "Population" in Table 5. The calculation and storage of the difference values of the other characters are the same as those of the characters “ri” and “go”. See Table 5 for location information for the eight different characters of Example 2. 6. After the above processing of each character is completed, a record of each character is obtained according to the order of the internal code of the character, and the database 10
4 is retrieved. When the record of the specified character is found, the database 1 is used by using the information of the character currently recorded.
The character information 04 and the character position information 04 are updated. When the record of the designated character is not found, the information of the currently recorded character is stored in the character information and the character position information of the database 104. In the second embodiment, the document America
2. The letters "li" and "people" in txt are taken as an example. The character information of "ri" and "shoku" are respectively
It is found in the database 104. Then, the character information and the character position information of the database 104 are updated. When updating the character position information, for the character “ri”, the document index generation unit 105 causes the document index generation unit 105 to determine the last 2 bytes of the first mini-block of the final database block in the database 104 and the first byte of the second mini-block. 5 bytes are updated and the database 10
In the character information of No. 4, the maximum character position of the character in the news document category is updated to 30503. In the case of writing the position information of “Public” in the database 104, the document index generating unit 105 sets the second position of the field storing the character position in the character position information of the database 104.
Update the data of 6th byte, 27th byte and 28th byte. At the same time, the maximum character position of the character “Population” in the news document category is updated to 30497 in the character information of the database 104. The other 6-character data updating process is the same as that of "ri" and "go". 7. In the database 104, the final position in the document category information and the end position in the document information of the recorded document are updated. In the second embodiment, in the document category information of the news document category of the database 104, the final positions of all characters in the news document category are 3050.
Updated to 4. In the document information, the document Ame
rica2. The ending position of txt is 30504.

Third Embodiment In this embodiment, it is assumed that a document is recorded after recording the document of the second embodiment and a plurality of other documents.

In the WINDOWS platform, in the present embodiment, the document (text file) America2. t
Record xt in database 104 and create an index for each character in the document, including 7 characters. The contents of the document are as follows: United States The document category name to which this document belongs is the news category, the document category number is 1, and a plurality of documents are recorded in this document category. In this document category, the final position of characters is 303842975, the document number of the above document is 290370, and the document name is Ameri.
It is assumed that ca3 is 2000.5.1 at the time of issuance. This document is supplied to the document index generation unit 105,
It is recorded with the steps shown below. In this embodiment,
The letter "ri" is taken as an example. Database 1
In 04, Katakana “ri” in the news document category
The maximum character position of is 1016947. The character position data is stored in a plurality of database blocks. The final database block for storing the character position data stores 1039997 bytes of position data.

1. In order to find the current final position in the news document category, the maximum character position of each character in this category, and the length of the position data stored in the final database block corresponding to each character, the document index generation unit 105 , The document category information of the news document category in the database 104 is searched. In the third embodiment, the final character position in the news document category is 303842975, and the maximum character position of katakana “ri” in the news document category is 1016947.
Is.

2. The entire content of the document is read into the memory RAM3. The document index generation unit 105 stores the document number, the document name, the creator, and the issuing time in the document information of the database 104. In this example, the system reads 7 characters: "United States". Document number 2903
70, document name America3, at the time of publication 2000.5.
11, and the start character position 303842976 of this document in the news document category is stored in the document information of the news document category in the database 104 by the document index generation unit 105.

3. Document America3. 7 in txt
The characters are converted one by one from the Shift JIS code to the corresponding internal code of the system (see columns 2 and 3 of Table 7). 4. The final position of the news document category + 1 is used as the starting position of the new document to be recorded, and the incrementing stage is 1.
Is. The character position of each character is determined according to the order of the characters in the document category. In this example, the document Amer
ica3. The start position of txt is 303842976. America3. The character positions of all the characters in txt are shown in Table 7.

5. The difference value of each character is calculated by the difference algorithm based on the maximum position of a certain character (described in item 1 of this embodiment) in the document category before the recording of a new document and the current character position of that character. . If the current difference value of a character requires multiple bytes and the remaining bytes of the current miniblock's 260 bytes (4 + 256 bytes) are not enough for the new difference value, the system will use 0x00 to Fill the remaining bytes of the miniblock. Subsequently, the new miniblock is used. The current position of the character in the document category is stored in the first 4 bytes as the minimum position of the new miniblock. If this character appears after this,
The difference value between each character position and the previous character position is stored in the fifth byte and each subsequent byte of the new miniblock. If the miniblock is filled with position data, a new miniblock is used and the current character position of the character is
It is stored in it as the minimum position of this miniblock. The above process is repeated until all miniblocks in the database block are filled with data. In the third embodiment, katakana “ri” is taken as an example. As described above, in the database 104,
The maximum position of katakana “ri” in the news document category is 1016947, and the final database block for storing the position data of this katakana is 103
Stores 9997 bytes of position data. 1039997
Calculating / 260, this katakana fills 3999 miniblocks in the final database block. Further, when calculating 1039997% 260, the 4000th mini-block of the database block stores 257 bytes of the Katakana position data, leaving 3 bytes. Document America3. In txt, the character position of "ri" in the news document category is 3
The maximum character position of Katakana "ri" in the news document category stored in the character position information of the database 104 is 1016947. According to the difference algorithm, 303842978 and 1016
The difference value with 947 is 0x8194EA9F77, and this storage requires 5 bytes. In order to maintain the integrity of each mini-block in the database block, the second to the second 258th bytes of the 4000th mini-block are maintained.
60 bytes are filled with 0x00. Subsequently, the new database block is used. The character position 0x121C46A2 of "ri" in the news document category is stored in the first mini block of the new database block.

6. After processing all the characters in the document, the database 104 is searched for a record for each character according to the order of the character's internal code. When the record of the specified character is found, the character information and the character position information of the database 104 are updated by using the information of the currently recorded character. When the record of the designated character is not found, the information of the currently recorded character is stored in the character information and the character position information of the database 104. In the third embodiment, the document America3. The character "li" in txt is taken as an example. The character information of “ri” is searched for in the database 104. Then, the character information and the character position information of the database 104 are updated. Document America3. Prior to recording txt, the 103999 byte position data for the character "ri" is stored in the database block of the last record entry for the character in database 104. When the character position information is updated, for the character “ri”, the document index generation unit 105 updates the database block of the last record item of the character position information of the character in the database 104. Subsequently, the new database block is stored in the database 104 as a new record item in the character position information of the character. This new record item becomes the last record item in the character position information of the character in the database 104. 7. In the database 104, the final position in the document category information and the end position in the document information of the recorded document are updated. In the third embodiment, in the document category information of the news document category in the database 104, the final positions of all characters in the news document category are 303
Updated to 842982. In the document information,
Document America3. The end position of txt is 3038
42982.

Full-Text Search Process Further, the present invention provides a full-text search method. In this method, a full-text search for a search word specified by an operator is performed using the character position information of the document index created in the present invention.

The full-text search processing using the created document index in the flowcharts of FIGS. 4A and 4B will be described below.

In step 502, the document category name is input. For example, a dialog box with an input box is displayed prompting the operator to enter the document category name.

In step 504, the full-text search engine 106 searches the database 104 for document category information according to the input document category name.
When the document category information of the document category designated by the operator is found, the full-text search engine 106 acquires the document category number of the document category from the document category information.

At step 506, a search term is entered. For example, a dialog box with an input box is displayed prompting the operator to enter a search term.

In step 508, a data initialization process prior to full text search is performed. This process is as follows: The search word matching position is defined. For example, the initial value is set to 1, that is, the position of the first appearance of the first character of the search word in the designated document category is set to 1. , The step of obtaining the number of characters of the search word, the step of converting each character of the search word into an internal code, and the step of giving a displacement to each character according to the order of each character of the search word. , If the first character of the search word "US America" is set as the starting point, the displacement of "US" is 0, the displacement of "Country" is -1, the displacement of "A" is -2, and the displacement of "M" is −
3. The displacement of "li" is -4, the displacement of "f" is -5, the process of constructing a database query statement, and the process of initializing a result set are included.

In step 510, step 506
The database inquiry statement composed of is given to the database 104 and a database search is performed. All record items of the position information of each character of the search word are searched. These records form a record set. The record set includes a position information record for each character in the database 104. Each record item is
Contains the fields of the database block for storing character position data.

At step 512, it is determined whether or not a record of each character of the search word is in the record set. If there are unrecorded characters in the recordset, the search ends. If there is no character without a record, the process proceeds to step 514.

In step 514, the found record for each character of the search term is aligned character by character according to the minimum position of the database block. Then, for each character, the database block of the first record item, the first miniblock in this database block and the first character position in this first miniblock are respectively the current database block, the current Is set as the miniblock and the current character position.

In step 516, the counter (I)
Is set. This indicates that the I-th character of the search word is being restored / matched. Step 51
At 8, the counter functions as a loop control variable for controlling the restoration / matching process of the character position of the I-th character of the search word. The initial value of I is 1, which indicates that the restore / match process begins with the first character of the search term.

If I is less than or equal to the number of characters in the search word in step 518, the process proceeds to step 520 to restore / collate the I-th character. If it exceeds the number of characters, it indicates that the search result has been acquired and stored, and the process proceeds to step 544.

In step 520, the search word match position is compared to the sum of the maximum position of the database block of the current record entry for character I of the search word and the displacement of character I. 1. If the search word matching position is larger, it means that there is no character position matching the search word matching position in the database block, and the process proceeds to step 538. Here, it is determined whether the letter I has more records in the record set. If character I has more records in the record set, proceed to step 540 and character I
Next record item of the record item, the database of the record item as the current database, the first miniblock therein as the current miniblock, and the first
Sets the minimum position of the miniblock of as the current character position. If there is no record, the search for the current search term ends. 2. If the search term match position is smaller, it means that the miniblock in the database block of the current record item may contain the matching character position. If the search word matching position is equal to the sum, it means that the maximum position of the current database block is the matching character position, and the process proceeds to step 542. Here, 1 is added to I to determine whether or not the next character matches the current matching position.

At step 522, first, some explanation will be given. In the present invention, the database block may have a plurality of mini blocks. Each miniblock includes a minimum position and a plurality of difference values. Position data is stored in ascending order. Therefore, the minimum position of the second mini block of the two consecutive mini blocks can be regarded as the maximum position of the first mini block. For example, the letters "day"
The minimum position of the fifth mini-block in the database block is 1000 and the minimum position of the sixth mini-block is 1.
It is assumed to be 500. It can be determined that the maximum character position stored in the fourth mini-block of the database block is less than 1000 and the maximum character position stored in the fifth mini-block of the database block is less than 1500. For the last miniblock in the database block, its maximum position can be determined to be the maximum position of the database block.

In step 522, the search word match position is compared to the sum of the current miniblock maximum position of the search word letter I and the displacement of the letter I. 1. If the search word matching position is larger, it means that there is no character position matching the search word matching position in the current miniblock, and the process proceeds to step 534. Here, it is determined whether or not there is a next miniblock in the current record item. If there is a block, proceed to step 536 to obtain the next miniblock, set this miniblock as the current miniblock, and set the miniblock's minimum position as the current character position. If there is no record, go to step 538.
2. If the search word matching position is smaller, it means that the current miniblock may store the matching character position, and the process proceeds to step 524. Here, the current mini-block position data is determined. If the search word matching position is equal to the sum, it indicates that the maximum position of the current mini-block is the matching character position, and the process proceeds to step 542.
Here, 1 is added to I to determine whether or not the next character matches the current matching position.

In step 524, the search word matching position is the current restored character position of the character I of the search word and the character I.
Is compared with the displacement of. 1. If the search word matching position is larger, the process proceeds to step 530, and it is determined whether or not the current miniblock has the next difference value. If there is a difference value, the process proceeds to step 532, and the next difference value is acquired. The difference value and the current character position of the character are calculated by the difference algorithm, and the new current character position of the character in the document category to which the character belongs is acquired. If there is no difference value, the process proceeds to step 534. 2. If the search word matching position is smaller, the process proceeds to step 526. Here, the search word matching position is the current restored character position of the character I and the character I.
Is reset as the displacement of No. and the process proceeds to step 528. In step 528, I is set to 1 and step 5
Return to 18. Here, the first character of the search word is searched for a character position that matches the new search word matching position. If the search word matching position is equal to the sum, it indicates that the current restored character position is the matching character position, and the process proceeds to step 542. Here, 1 is added to I, and it is determined whether or not the next character matches the current search word matching position.

In Step 544, Step 518
If I is determined to be greater than the number of characters in the search term, the current restored character position of each character in the search term matches the current search term match position, that is, the search result of the search term is the current Means found in document category. Then, the document information sharing conversion unit determines in which document the current search word matching position is located. Further, it is determined by the deletion flag whether the document has been deleted. When a document is deleted, the search term that appears in the deleted document is
It cannot be a search result. If the document has not been deleted, the document number of the document is obtained.

In step 546, the obtained document number is stored in the search result set.

At step 548, the search word matching position is updated. 1 is added to the current search word matching position to form a new search word matching position. Returning to step 516, I is set to 1. Next, in step 518, a character position that matches the new search word matching position is searched from the first character of the search word.

Fourth Embodiment The full-text search engine 106 performs full-text search for each document stored in the database 104.

In step 502, the document category name is input. For example, a dialog box with an input box is displayed prompting the operator to enter the document category name "News".

In step 504, the full-text search engine 106 searches the database 104 for document category information according to the input document category name.
When the document category information of the news document category is found, the full-text search engine 106 acquires the document category number 1 of the news document category from the document category information.

At step 506, a search term is entered. For example, a dialog box with an input box is displayed prompting the operator to enter the search term "US America".

In step 508, a data initialization process prior to full text search is performed. This process is as follows: defining the initial value of the search word matching position as 1, that is, setting the first appearance position of the first character of the search word in the news document category to 1.
The process of obtaining the number of characters 6 in the search word "US America",
According to the process of converting each character of the search word into an internal code, for example, converting 6 characters from the shift JIS code into the corresponding system internal code (see columns 1 and 2 of Table 9) and the order of each character of the search word. , Is the process of giving displacement to each character, the first character of the search word "US America" is set as the starting point, and the displacement is given to each of the six characters,
The displacement of "rice" is 0, the displacement of "country" is -1, the displacement of "a" is -2, the displacement of "me" is -3, the displacement of "li" is -4,
The displacement of "F" is -5, the input document category name "News" and the 6-character internal code are described in the database SQL query statement, and the result set is emptied. including.

In step 510, step 506
The database inquiry statement composed of is given to the database 104 and a database search is performed. All record items of the position information of each character of the search word are searched. These records form a record set. Each record item includes multiple fields. A database block that stores character position data is included as a field in each record item. That is, one record item corresponds to one database block. The record set includes position information of each character of the search term stored in the database 104. Table 9 shows
6 shows some record items of character position information of 6 characters “US America” in the news document category in the database 104. At step 512, it is determined whether there is a record for each character of the search term in the record set. If there are unrecorded characters in the recordset, the search ends. If there is no character without a record, the process proceeds to step 514.

In step 514, the found record for each character of the search term is aligned character by character according to the minimum position of the database block. For example,
The minimum positions of the database blocks of the three record items of "ri" are 5, 30494 and 30384, respectively.
2978.

In step 516, the counter (I)
The initial value of is set to 1. This means that the restoration / matching process of the I-th character “US” of the search word is started from the first character “US”.

If I = 1 <6 in step 518, the process proceeds to step 520. The search word matching position initialized to 1 is the maximum position 30499 of the database block of the character "rice" and the displacement 0 of the character "rice" (columns 1 and 4 of Table 10).
And 5)). Since 1 <30499 + 0, there is a possibility that there is a character position of "US" matching the current search word matching position 1 in the database block, that is, the sum of the character position and the displacement 0 is the current search. It is equal to the word match position. It is assumed that the minimum position of the second mini block in the database block is X.
Search word matching position 1 is the maximum position X of the first mini-block of the database block (minimum position of the second mini-block)
Is compared with the displacement of the character "rice" and zero. It is determined that the character position of “US” equal to the current search word matching position 1 may exist in the first miniblock. Subsequently, the current search word matching position 1 is compared with the sum of the minimum position 1 and the displacement 0 of the first miniblock. The comparison results are equal and the counter is incremented by 1 to I = 2. Returning to step 518, it is determined whether the database block for the character "Country" has a character position whose sum with the "Country" displacement equals the current search term match position 1.

Currently, I = 2, and the displacement of "country" is -1. The maximum position of the database block for the first record item of characters is 303842982 (see columns 9, 4 and 5 of Table 9). The minimum position of the first mini block is 2
Is. Let Y be the maximum position of the first mini-block. First, the sum of the maximum position of the database block of the record item and the displacement of the character is 303842981 (303842982-
1 = 303842981) is calculated. This sum 3038429
81 is compared to the current search term match position 1, which determines that the database block may have a character position that matches the current search term match position. The search word matching position 1 is compared with the sum (Y-1) of the maximum position of the first miniblock (minimum position of the second miniblock) and the displacement of the database block, and thereby the current search word matching position is obtained. It is determined that the matching character position may be in the first miniblock of the database block. The search word matching position 1 is compared with the sum 1 (2-1 = 1) of the minimum position 2 of the first miniblock and the displacement -1. The comparison results are equal, which means that the character position of the second character "country" of the search word that matches the current search word matching position 1 has been found. The counter is incremented by 1.

The matching process for the remaining four characters "America" is similar to the character "America", only the displacements are different. The counter is 7 when the matching process of “F” is completed. Therefore, the numerical value of the counter is larger than the number of characters of the search word, which means that the first search result of the search word is found at the matching position 1. Then, it progresses to step 544.

In steps 544 to 548, the document information sharing conversion unit 107 searches for the corresponding document number 1 by using the matching position 1 described above. The search result is
Stored in the result set. Then, the search word matching position is 1
By 1 to obtain a new matching position 2 as the current search word matching position. The counter is reset to 1. A search is started for results that match the new search word matching position from the current character position in the current miniblock of the current database block for each character.

When the search for the next search result is started, first, the difference value 0x826E is obtained from the 5th and 6th bytes of the first miniblock of the current database block of the character "US". The most significant bits of the digits other than the unit digit are restored to 0. That is, 0x826E is
It is restored to 0x026E. (016E) 16 = (366) 10. Then, the sum of the character positions 1 and 366 before "US" is calculated and the restored character position 367 (1 + 366 = 36
7) is obtained. The sum of the restored character position 367 and the displacement 0 is larger than the current search word matching position. Therefore,
The restored character position does not match the matching position 2.
Using the sum 367 of the current character position 367 and the displacement 0,
The search word matching position is reset. Also, the counter is 1
Is reset to. A search is started for results that match the new search word match position 367 from the current character position in the current miniblock of the current database block for each character. The above process continues until there is no unprocessed character position information for the characters in the search term. This completes the search process for the word "US America".

Fifth Embodiment: The full-text search engine 106 performs full-text search for each document stored in the database 104. Search for the word "America" in the News Documents category.

In step 502, the document category name is input. For example, a dialog box with an input box is displayed prompting the operator to enter the document category name "News".

In step 504, the full-text search engine 106 searches the database 104 for document category information according to the input document category name.
When the document category information of the news document category is found, the full-text search engine 106 acquires the document category number 1 of the news document category from the document category information.

At step 506, a search term is entered. For example, a dialog box with an input box is displayed, prompting the operator with the search term "America".
Prompt you to enter.

In step 508, a data initialization process prior to full text search is performed. This process is as follows: The initial value of the search word matching position is defined as 1, that is, the position of the first appearance of the first character of the search word in the news document category is 1, and the search The process of obtaining the number of characters of the word "America",
A process of converting each character of the search word into an internal code, for example, converting four characters from the shift JIS code into a corresponding system internal code (see columns 1 and 2 of Table 10),
In the process of giving each character a displacement according to the order of each character in the search word, the first character of the search word "America" is set as the starting point, and the four characters are each given a displacement and the "A" The displacement is 0, the displacement of "me" is -1, the displacement of "li" is -2, the displacement of "f" is -3, the input document category name "news" and the 4-character internal code. Include the steps described in the database SQL query statement and emptying the result set.

In Step 510, Step 506
The database inquiry statement composed of is given to the database 104 and a database search is performed. All record items of the position information of each character of the search word are searched. These records form a record set. Each record item includes multiple fields. A database block that stores character position data is included as a field in each record item. That is, one record item corresponds to one database block. The record set includes position information of each character of the search term stored in the database 104. Table 10
Indicates some record items of character position information of four characters “America” in the news document category in the database 104. At step 512, it is determined whether there is a record for each character of the search term in the record set. If there are unrecorded characters in the recordset, the search ends. If there is no character without a record, the process proceeds to step 514.

In step 514, the found record for each character of the search term is aligned character by character according to the minimum position of the database block. For example,
The minimum positions of the database blocks of the three record items of "ri" are 1005, 30494, and 303, respectively.
842978.

In step 516, the counter (I)
The initial value of is set to 1. This indicates that the restoration / matching process starts from the first character "A" of the search word.

If I = 1 <4 in step 518, the process proceeds to step 520. The search word match position initialized to 1 is compared to the sum of the maximum position 303842976 of the database block for the character "A" and the displacement 0 of the character "A" (see columns 1, 4 and 5 of Table 10). It 1 <3038
Since it is 42976 + 0, there is a possibility that there is a character position of "A" that matches the current search word matching position 1 in the database block, that is, the sum of the character position and the displacement 0 is the current search word matching position. Means equal to position 1. The minimum position of the second mini-block in the database block is X1
Suppose The search word match position 1 is compared with the sum of the maximum position X1 of the first miniblock of the database block (the minimum position of the second miniblock) and the displacement 0 of the character "A". It is determined that a character position of "A" equal to the current search word matching position 1 may exist in the first miniblock. Subsequently, the current search word matching position 1 is compared with the sum of the minimum position 1003 of the first miniblock and the displacement 0. The sum of the minimum position 1003 of the first miniblock and the displacement 0 is larger than the current search word matching position 1. Current,
The current character position of "A" is 1003. In step 526, the search word matching position is set to the sum of the current character position 1003 of "A" and displacement 0, and the counter is set to 1. Returning to step 518, the new search word match position 1003 is used again to perform the matching process on the current character position of the first character "A".
The search word matching position 1003 is the current character position 10 of the character.
It is determined to be equal to the sum of 03 and the displacement. I is incremented by 1.

Currently, I = 2, and the displacement of "M" is -1. The maximum position of the database block for the first record item of characters is 303842977 (columns 1 and 4 of Table 10).
And 5). The minimum position of the first mini block is 1004. The maximum position of the first mini block is Y2. First, the sum of the maximum position of the database block of the record item and the displacement -1 of the character is 303842976.
(303842977-1 = 303842976) is calculated. This sum 303842976 is the current search word match position 1003.
And it is determined that the database block may have a character position that matches the current search term match position. The search word matching position 1003 is compared with the sum (Y2-1) of the maximum position (minimum position of the second mini block) of the first mini-block of the database block and the displacement thereof, whereby the current search word matching position 1003. It is determined that the character position that matches with may exist in the first miniblock of the database block. The search word matching position 1003 is the minimum position 100 of the first mini block.
It is compared with the sum of 4 and displacement-1 1003 (1004-1 = 1003). The comparison results are the same, which means that the character position of the second character "me" of the search word that matches the current search word matching position 1003 has been found. The counter is incremented by 1.

Currently, I = 3. Search word match position 10
The character position of the third character "ri" that matches 03 is searched. The displacement of "ri" is -2. The maximum position of the current database block is 1320 (column 1 of Table 10,
4 and 5). The minimum position of the first miniblock is
1004. The maximum position of the first mini block is X3
And First, the sum of the maximum position of the record item database block and the character displacement -2 is 1318 (1320 -2).
= 1318). The sum 1318 is then compared to the current search term match position 1003, thereby determining that the database block may have a character position that matches the current search term match position. The search word matching position 1003 is compared with the sum (X3 -2) of the maximum position (minimum position of the second mini block) of the first mini-block of the database block and the displacement -2, whereby the current search word match is found. It is determined that a character position matching position 1003 may be present in the first miniblock of the database block. The search word matching position 1003 is the sum 100 of the minimum position 1005 of the first mini block and the displacement −1.
3 (1005-2 = 1003). The comparison results are the same, which means that the character position of the third character "ri" of the search word that matches the current search word matching position 1003 has been found. The counter is incremented by 1.

Currently, I = 4. Search word match position 10
The character position of the third character "F" that matches 03 is searched. The displacement of "F" is -3. The maximum position of the current database block is 303842979 (see columns 1, 4 and 5 of Table 10). The minimum position of the first mini block is 1006. The maximum position of the first mini block is
It is assumed to be X4. First, the sum of the maximum position of the database block of the record item and the displacement-3 of the character is 3038.
It is calculated as 42976 (303842979-3 = 303842976). Sum 303842976 is the current search word match position 1
003, which determines that it may have a character position that matches the current search term match position 1003. The search term matching position 1003 is compared with the sum (X4-3) of the maximum position (minimum position of the second mini block) of the first mini-block of the database block and the displacement -3, whereby the current search word match is found. It is determined that a character position matching position 1003 may be present in the first miniblock of the database block. The search word matching position 1003 is the minimum position 1005 of the first mini block.
And the sum of displacement-3 are 1003 (1006-3 = 1003). The comparison results are the same, which means that the character position of the fourth character “” of the search word that matches the current search word matching position 1003 has been found. The counter is incremented by 1.

Currently, I = 5. At step 518, the counter value is greater than the number of characters in the search term. Therefore, it is determined that the search result is found at the matching position 1003. In steps 544 to 548, the document information sharing conversion unit 107 searches for the corresponding document number using the matching position 1003 described above. Search results are stored in the result set. Subsequently, the search word matching position 1004 is incremented by 1, and a new matching position 1004 is obtained as the current search word matching position.
The counter is reset to 1. The current character position of each character of the word "America" is 1003, 100, respectively.
4, 1005 and 1006. A search is started for a new search result that matches the new search word matching position from the current character position in the current miniblock of the current database block for each character.

When searching for the next search result, first, the difference value 0x04 is obtained from the fifth byte of the first miniblock of the current database block of the character "A". Since this difference value has only one digit, the difference value is directly added to the previous character position 1003 and the restored character position 10
07 (1003 + 4 = 1007) is obtained. By comparison,
The sum of the restored character position 1007 of "A" and displacement 0 is
It is larger than the current search word matching position 1004. The current character position 1007 plus the zero displacement 1007 is used to reset the search term match position. The counter is also reset to 1. The search for the matching character position of "A" is restarted. As a result, the search word matching position 1007 is equal to the current character position of "A". I is changed from 1 to 2.

The character positions of the remaining three characters "Merika" that match the current search word matching position 1007 are searched respectively. In steps 544 to 548, if all the characters match the matching position, the document information sharing conversion unit 10
7, the obtained matching position is used to find the corresponding document number. The new search results are stored in the result set. The above process is then repeated from 516. When the character of the search word has no unprocessed character position information, the search for the word "America" ends.

Document Information Sharing Conversion Unit The present invention also provides a method for quickly acquiring corresponding document information from a designated character position. Based on the high-speed access function of the cache memory, this method uses the database 104 in the shared memory.
A backup copy of part of the document information is stored.
When performing a full text search, the bisection algorithm quickly and accurately obtains the corresponding document number from one or more matching positions found in the full text search process. The document information sharing conversion unit 107 is activated so that the backup copy of the document information is updated in good timing each time the document recording process or the document deleting process is performed.

The process of the document information sharing converter in FIG. 5 will be described below.

1. According to the input of the document category number, the document information sharing conversion unit 107 first checks whether or not the document information of the designated document category is stored in the shared memory 604. If the document information of the specified document category does not exist in the shared memory 604, the block of the shared memory is applied to the system. Then, the document information in the database 104 is searched. The data items of the document information (document number, start position and end position of document, deletion flag, etc.) of all documents in the specified document category in the database 104 representing the document number and position range of each document are shared memory. Read into 604 and become resident in shared memory according to the in-order record of the document so that it can be used by multiple users to retrieve a specified document category. Other data items of the document information of the database 104 such as creator, title, recording date and time, and deletion date and time are stored in the database 104 as list format information. When the shared memory 604 stores the document information of the designated document category, the document information can be directly read from the shared memory 604 without accessing the database 104. Therefore, the frequency of input / output of the database 104 can be reduced, the access time to the database can be reduced, and the speed of database inquiry can be increased.

2. When performing a full-text search, position information 6 which is a one-dimensional array for storing one or more matching positions
06 is the document information sharing conversion unit 10 as an input parameter.
Given to 7. By the bisection algorithm, the document information sharing conversion unit 107 compares each matching position with the range of each document in the shared memory 604 (start position and end position of the document),
Documents with matching positions can be determined. The deletion flag of the determined document is checked. If the document has been deleted, the delete flag is 1 and the return value for this document is -1. When the document is not deleted, the document information sharing conversion unit 107 outputs the found corresponding document number. Finally, all the document numbers converted from the matching positions are stored in the document information 608 which is a one-dimensional array for storing one or more document numbers. Document information 6
The storage order of the document numbers in 08 is just position information 6
This corresponds to the order of matching positions in 06.

3. When recording a new document and creating a character index, the document index generation unit 105
An interface is provided for timely addition of new document information to the shared memory 604.

4. When a document is deleted, the document index generation unit 105 provides an interface for setting the deletion flag of the deleted document in the shared memory 604 to 1 (indicating that the document has been deleted) at the appropriate time.

Embodiment 7: In this embodiment, the data in the one-dimensional array for storing the matching position is converted into the document information so that the document number corresponding to the matching position can be obtained.

1. According to the input 602 of the document category number 1, the document information sharing conversion unit 107 first checks whether the document information of the document category number 1 is stored in the shared memory 604. If the document information of the specified document category does not exist in the shared memory 604, the block of the shared memory 604 is applied to the system. Then, the document information of all the documents of the document category number 1 in the document information of the database 104 is searched. The document number, start position, and end position of each document are acquired and read into the shared memory 604, and used by multiple users to search a specified document category.
It becomes resident in the shared memory 604. For details, see FIG.
Please refer to.

2. The position information 606, which is a one-dimensional array for storing a plurality of matching positions, is used as the document information sharing conversion unit 1.
It is given to 07.

3. By the bisection algorithm, the document information sharing conversion unit 107 compares the range of each document (the start position and the end position of the document) in the shared memory 604 with each matching position from the first matching position in the array, and the matching position Allows the determination of documents with The deletion flag of the determined document is checked. If the document has been deleted, the delete flag is 1 and the return value corresponding to this document is -1. When the document is not deleted, the document information sharing conversion unit 107 outputs the found corresponding document number. In the seventh embodiment, the first matching position 1001 in the position information 606.
, The shared memory 604
In, the corresponding document whose start position is 998, end position is 1100, the document number is 21 and the deletion flag is 0 indicating that the document has not been deleted is searched. The document information sharing conversion unit 107 stores the document number 2 of the document corresponding to the matching position 1001 in the document information 608 as the first number.
Thus, the process of converting one matching position into the corresponding document number is completed. When processing the matching position 3001 in the position information 606 at the time of searching and comparing, the document information sharing conversion unit 107 converts the matching position 3001 into a document with a start position of 2890, an end position of 3005, and a document number of 41. Judge that there is. Then, it is checked whether the deletion flag is 1. This means that the document has been deleted. When the document information sharing conversion unit 107 returns the value -1, it means that there is no corresponding document at the matching position 3001. The conversion process by the document information sharing conversion unit to another matching position in the position information 606 is the same as the above process.

4. The document number converted from the search word position in the position information 606 is stored in the document information 608 by the document information sharing conversion unit 107. The storage order of the document numbers of the document information 608 corresponds to the storage order of the matching positions in the position information 606.

[Brief description of drawings]

FIG. 1A is a diagram showing an example of the structure of an index creation and full-text search system.

1B is a hardware block diagram of the server of FIG. 1A.

FIG. 2 is a diagram showing a structure of a database block and a mini block for storing character position data.

FIG. 3 is a flowchart of processing of a document index generation unit.

FIG. 4A is a flowchart of processing of a full-text search engine.

FIG. 4B is a flowchart of processing of a full-text search engine.

FIG. 5 is a diagram showing a process of a document information sharing conversion unit.

[Explanation of symbols]

Reference numeral 104 ... Database, 105 ... Document index generation unit, 106 ... Full-text search engine, 107 ... Document information sharing conversion unit, 604 ... Shared memory, 608 ... Document information

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Rihon             People's Republic of China Beijing 100080, High             Dian District, Summer Palais             Slaud, Number 1, Beijing Litho             Sus Hotel, Room 1307 Beijing             Pecan Information System               Within Corporation Limited F-term (reference) 5B075 ND02 NK49

Claims (21)

[Claims] 1. A method of creating an index of character information, the process of determining the position of each character in a group of documents to be searched according to the order of all characters in the entire document, and the position data of the character A process of sequentially storing in one or more database blocks corresponding to characters to obtain the maximum position and the minimum position stored in each database block, and a plurality of mini blocks each having a storage space of multiple bytes And obtaining a minimum position stored in each mini-block. 2. The difference value is acquired between each two consecutive positions of all the positions of the character, and the difference value is stored as position data in a database block. Method. 3. The difference value is calculated by a difference algorithm, wherein the difference algorithm converts the difference between the current position and the previous position of the characters of the set of documents into a 127-ary number. Each digit of the 127-digit number is stored as an 8-bit byte, the most significant bit of the byte of each digit is 0, and the 127-digit number is divided to divide each difference value of the characters. 3. The most significant bit of the byte of each digit is set to 1 excluding the unit digit, and the difference value corresponding to the current position of the character is obtained. Method. 4. When indexing the characters of the group of documents, when the remaining bytes of the current miniblock of the characters are not sufficient for the new difference value, then:
Each of the remaining bytes is filled with 0x00, a new miniblock is used, and the current position of the character in the set of documents is the first byte of the new miniblock as the minimum position of the new miniblock. The method of claim 1, wherein the method is stored in :. 5. A process of determining the position of each character in a group of documents to be searched according to the order of all the characters in the entire document, and the position data of the character in one or more database blocks corresponding to the character. A process of sequentially storing and obtaining the maximum position and the minimum position stored in each database block, and dividing each database block into a plurality of mini blocks each having a storage space of multiple bytes, and storing each mini block A storage medium storing a program for executing the process of obtaining the minimum position. 6. A process of determining the position of each character of a group of documents to be searched according to the order of all characters of the entire document, and the position data of the character in one or more database blocks corresponding to the character. A process of sequentially storing and obtaining the maximum position and the minimum position stored in each database block, and dividing each database block into a plurality of mini blocks each having a storage space of multiple bytes, and storing each mini block A computer program executed by a computer to carry out the process of obtaining a minimum position. 7. A process of obtaining a relative positional relationship between each character of the search word to search an index of each character of the search word, and each database block of the character matches the relative positional relationship. Each step of determining whether or not there is a possibility of having a position, and for each database block that may have a position that matches the relative positional relationship, each miniblock of that database block is A step of determining whether or not there is a possibility of having a position that matches, and for a miniblock that may have a position that matches the relative positional relationship, each position of the miniblock is
The method for searching character information based on the index of the character information created according to claim 1, comprising the step of determining whether the character information matches the relative positional relationship. 8. It is determined whether or not there is a possibility that the database block has a position that matches the relative positional relationship, based on the maximum position and the minimum position of each database block. Item 7. The method according to Item 7. 9. The minimum position of the second miniblock of the two consecutive miniblocks can be regarded as the maximum position of the first miniblock, and its maximum position with respect to the last miniblock of the database block. Can be determined as the maximum position of the database block, and based on the maximum position and the minimum position of each miniblock, whether the miniblock may have a position that matches the relative positional relationship. 8. The method of claim 7, wherein is determined. 10. The relative positional relationship between the characters of the search word is set with the first character in the search word as a starting point,
Each character has a displacement of 0, -1, -2, -3 ,. ． ． ,-(N-
8. The method according to claim 7, wherein 1) is sequentially given, where N is represented as the number of characters of the search term. 11. A process of obtaining a relative positional relationship between each character of the search word for searching an index of each character of the search word, and each database block of the character matches the relative positional relationship. Each step of determining whether or not there is a possibility of having a position, and for each database block that may have a position that matches the relative positional relationship, each miniblock of that database block is And a step of determining whether or not there is a possibility of having a position that matches each other, and for a miniblock that may have a position that matches the relative positional relationship, each position of the miniblock is
A storage medium that stores a program for executing a process of determining whether or not the relative positional relationship matches. 12. A process of obtaining a relative positional relationship between each character of the search word to retrieve an index of each character of the search word, and each database block of the character matches the relative positional relationship. Each step of determining whether or not there is a possibility of having a position, and for each database block that may have a position that matches the relative positional relationship, each miniblock of that database block is And a step of determining whether or not there is a possibility of having a position that matches each other, and for a miniblock that may have a position that matches the relative positional relationship, each position of the miniblock is
A computer program executed by a computer to perform a process of determining whether or not the relative positional relationship matches. 13. A method for indexing and searching character information, wherein a certain amount of shared memory is reserved, and a part of each field of the document information of a group of documents to be searched stored in the database is partially retrieved from the database. A method for obtaining related document information directly from the shared memory when a document information sharing conversion unit to be read into the shared memory is installed and a full-text search is performed. 14. The method according to claim 13, wherein the read document information includes at least a field indicating a document number and a field indicating a document range. 15. When performing a full-text search, for one or more search word positions to be searched, the document in which the search word is located is acquired by a bisection algorithm according to the position range of each document. 14. The method of claim 13 characterized. 16. When a document index is created,
14. The method according to claim 13, wherein the document information in the shared memory is updated in a timely manner. 17. When recording or deleting each document, the document information sharing conversion unit provides a corresponding interface for updating the relevant document information in the shared memory at a proper timing. Claim 13
The method described. 18. A process of reserving a shared memory of a certain capacity and copying the document information of a group of documents to be searched stored in a database from the database to the shared memory, and performing the full text search, the sharing A storage medium that stores a program for performing a process of directly acquiring related document information from a memory. 19. A process of reserving a shared memory of a certain capacity and copying the document information of a group of documents to be searched stored in a database from the database to the shared memory, and performing the full text search, the sharing A computer program executed by a computer to perform the process of obtaining related document information directly from a memory. 20. A system for performing indexing and searching of character information, wherein the position of each character of a group of documents to be searched is determined according to the order of all characters of the entire document, and the position data of the character is determined. Are sequentially stored in one or more database blocks corresponding to the character, the maximum position and the minimum position stored in each database block are obtained, and each database block has a plurality of mini blocks each having a storage space of a plurality of bytes. And a relative position relationship between each character of the search word in order to search the index of each character of the search word, and It is determined whether or not each database block of characters may have a position that matches the relative positional relationship. And,
For a database block that may have a position that matches the relative positional relationship, it is determined whether each miniblock of the database block may have a position that matches the relative positional relationship. A full-text search unit that determines whether or not each position of the mini-block may have a position that matches the relative positional relationship, the full-text search unit that determines whether or not each position of the mini-block matches the relative positional relationship, And a storage unit that stores the index information. 21. In a system for indexing and searching character information, a part of each field of document information of a group of documents to be searched, which reserves a certain amount of shared memory and is stored in a database. A system including document information sharing conversion means for directly acquiring related document information from the shared memory when performing full-text search by reading the database from the database into the shared memory.