KR20050005261A - Information search management system and method tereof - Google Patents

Abstract

PURPOSE: A system and a method for searching/managing database information are provided to construct a low-cost/high-efficiency information system by forming a stable management function in addition to a search function for a database. CONSTITUTION: An index database volume(70) stores an index database. A document database volume(80) stores distributed document databases. A dictionary database volume(50) stores distributed dictionary databases. A kernel(10) performs data I/O(Input/Output) between the index/document database volume and a user memory, and manages files, directories, records, and reverse files in the index/document database volume. A storage engine(20) manages catalogs storing metadata for the databases and manages document/indexes by using the kernel. A search engine(30) performs the search for a user's query. An indexer(40) extracts the index for the document inputted from the user by using the dictionary database. A data manager(60) generates the database by receiving a schema file from a manager, loads a bundle of raw documents in bulk, and performs indexing.

Description

Information retrieval management system and its method {INFORMATION SEARCH MANAGEMENT SYSTEM AND METHOD TEREOF}

The present invention relates to an information retrieval management system and method for retrieving and managing database (DB) information. In particular, the present invention relates to an information retrieval management system and a method for constructing a low-cost and high-efficiency information system by adding a stable management function in addition to a database search function.

In the past, the Internet, which was recognized as the exclusive property of a certain class, became a place of information sharing for the general public with the advent of the web, and searched for information by users who wanted to find their necessary information in the explosive flood of information. The demand for the system has increased significantly.

As information retrieval systems are increasingly using database management systems (DBMSs) as data storage locations, systems that support retrieval by combining structured data from information retrieval systems (DBMSs) with reverse file structures It is mainstream. That is, a database management system (DBMS) is in charge of managing data, and an information retrieval system performs an index and corresponds to a user search.

However, simultaneously using a database management system (DBMS) and an information retrieval system for information services has the disadvantage of increasing the service construction cost, slowing down the development speed, and weighting the system as a whole.

In addition, with the exponential growth of Internet users, the amount of information on the Internet is also rapidly increasing, and information updates frequently occur. The information retrieval system requires a large volume of indexes and frequent data updates due to the demands of users who want to find the information they need in the flood of information.

Conventional information retrieval systems have been difficult to process large amounts of data (index) or have a long response time for many users, and consume a lot of system resources. In addition, there is a difficulty in applying in a field requiring frequent data updates due to the bulk loading method of collecting and processing data updates at once. In addition, the conventional database system, which is widely used in various fields, can frequently update data, but it takes a long time to process a large amount of unstructured documents.

Accordingly, the present invention has been made to solve the above problems, and an object of the present invention is to add a stable management function in addition to the search function for the database, an information retrieval management system and method capable of constructing a low cost and high efficiency information system To provide.

Another object of the present invention is to provide an information retrieval management system and method for supporting Unicode, Chinese, Chinese and multilingual languages at the storage engine level.

In addition, another object of the present invention is to provide an information retrieval management system and method for efficiently using storage space by reducing the space for the index database that is about three times the document size by supporting compression of the document and index database. To provide.

In addition, another object of the present invention is to provide an information retrieval management system and method that takes advantage of the existing information retrieval system and database system to enable the storage and fast retrieval of large data, many simultaneous user processing and frequent data updates. have.

1 is an overall block diagram of an information retrieval management system according to the present invention

2 is a conceptual diagram conceptually illustrating a loading process of a database of an information retrieval management system according to the present invention;

3 is a conceptual diagram conceptually showing a form of communication with the process of configuring the information retrieval management system according to the present invention

4A to 4F are conceptual views conceptually showing a search service on the client side of the information retrieval management system according to the present invention;

4A is a conceptual diagram illustrating how a client obtains information of a database;

4B is a conceptual diagram illustrating how a client obtains a section list of a database;

4c is a conceptual diagram illustrating how a client requests a search;

4d is a conceptual diagram illustrating how a client requests similar document search;

4E is a conceptual diagram illustrating how a client requests a list of search results;

4F is a conceptual diagram illustrating how a client requests the original text of a document.

5A and 5B are conceptual views conceptually illustrating an online document management service of the information retrieval management system according to the present invention.

5A is a conceptual diagram illustrating a method of inserting and changing a single document;

5B is a conceptual diagram illustrating a method of deleting one document.

6 is a diagram illustrating indexing of a title section of a document according to two indexing methods;

<Explanation of symbols for the main parts of the drawings>

10: Kernel 11: File and Directory Manager

12: Record Manager 13: Back File Manager

14: transaction manager 15: I / O manager

20: storage engine 21: catalog manager

22: Document Manager 23: Index Manager

30: search engine 40: indexer

50: Dictionary Database 60: Data Manager

70: List Database Volume 80: Document Database Volume

101: database schema file 102: raw document file

103: stopwords list 110: loader

210: Job Scheduler

220/1 to 220 / n, 330/1 to 330 / n: first fire to nth fire

230: set manager 232: set

240: Data Manager 250: List Database

260: first to nth database

310: Client 320: Job Scheduler

340: set manager 350: data manager

Information retrieval management system according to the present invention for achieving the above object,

A list database volume that stores the list database;

A document database volume in which one or several document databases are distributed and stored;

A dictionary database volume in which one or several dictionary databases are distributed and stored;

A kernel that performs data input / output between the list and document database volumes and user memory, and manages files, directories, records, and inverse files in the list and document database volumes;

A storage engine for performing catalog management, document and index management using the kernel to store meta information about a database;

A search engine that searches for a user's query;

An indexer for extracting an index word for the document input by the user using the dictionary database; And

It includes a data manager that receives a schema file created by an administrator, creates a database, bulk loads a bundle of source documents, and performs an index.

The kernel may further include a function of managing pages and buffers for accessing the list and document database volumes, and performing logging and locking functions therefor.

The kernel may include a file and directory manager that maps a physical identifier of a disk in which a record is stored to a logical identifier to facilitate access, and issues logical identifiers for a plurality of volumes; It supports the management of objects that can be stored on one page internally and long data items more than one page, and the ability to sequentially access records in the file from the beginning, and update operations related to inserting, modifying, and deleting records. A record manager; A reverse file manager which compresses a key for efficient storage space and performs insertion, modification, deletion, and retrieval; A transaction manager providing a start and end of a transaction, aborting, a save point function, and a function of recording a transaction information in a log file; And an input / output manager for managing data input / output between the list and document database volumes and the user memory.

The transaction manager uses a checkpoint to reduce recovery time, and if a specified task fails in the middle, all the tasks are returned to the starting point of the transaction to maintain the integrity of the data.

The input / output manager may include: a page manager which allocates, deletes, and maintains pages in the list and document database volumes; A buffer manager for mapping the pages of the disk to the pages of the memory; And a lock manager that different requirements are used to access the same object.

The storage engine includes a catalog manager that manages information about a structure of a database to be built; A document manager converting the original document into an internal document structure and performing operations for insertion, deletion or modification of an existing document; And an index manager configured to configure the index information extracted using the indexer into a structure suitable for searching.

Information on the structure of the database is characterized in that it includes a document structure, indexing method, primary key information, compression or not, stopwords.

Another information search management system according to the present invention for achieving the above object,

A list database volume that stores the list database;

A document database volume in which one or several document databases are distributed and stored;

A database schema file describing information about the directory and volume of the list and document database, a database group, a section definition, a section indexing method, a structure of a raw document, and a loading method from a source document to a database;

A raw document file consisting of raw documents to be loaded into the document database;

A stopword list to be loaded into the list database; And

The document database is generated and the document is loaded according to the information received from the database schema file, and the loader is configured to include a loader for receiving commands for generating the document database and loading the document from an administrator.

Through the structure of the original document described in the database schema file, a document having a heterogeneous structure can be loaded into a single database.

Another information retrieval management system according to the present invention for achieving the above object,

It distributes the work according to the status of the first to nth fires received from the client, requests document management from the data manager during online document management, and transmits the result to the client. A job scheduler for notifying the first to nth fire, set managers that the database has been changed;

First to n-th fires that perform a job on a service request received from the job scheduler, transmit a result to the client, and request to save a search result to a set manager;

A set manager that performs a task on a service request received from the first to nth fires and transmits a result to the first to nth fires; And

And a data manager for performing a job on the service request received from the job scheduler and transmitting the result to the job scheduler.

Information retrieval management method according to the present invention for achieving the above object,

An information retrieval management method of an information retrieval management system including a job scheduler, first to n-th fires, set manager, and data manager,

A first step of a client requesting and receiving information of a database from the first to nth fires through the job scheduler;

A second step in which the client requests and receives a section list of the database from the first to nth fires through the job scheduler;

A third step of receiving, by the client, a search result from the first to nth fires through the job scheduler and receiving a search result;

A fourth step in which the client requests a similar document search from the first to nth fires through the job scheduler and receives a search result;

A fifth step in which the client requests and receives a search result list from the first to nth fires through the job scheduler; And

And a sixth step in which the client requests and receives the original text of the first through nth fire documents through the job scheduler.

In the first step, the first to n-th fires retrieve database information by the request of the client received through the job scheduler, and transmit the information to the client.

In the second step, the first to n-th fires may retrieve a section list of a database by the request of the client received through the job scheduler, and transmit the retrieved section list to the client.

In the third step, the client transmits search information including a query, a search target database list, and a section to the first to nth fires through the job scheduler.

The third step may include: performing a search by the search information received from the client in the first to nth fires and transmitting a search result to the set manager; Storing the search results received from the first to nth fires in the set manager, and transmitting a result set number and the number of documents to the first to nth fires; And transmitting the result set number and the number of documents received from the set manager to the client in the first to nth fires.

In the fourth step, the client transmits similar document search information including a search document number, a search target database list, a section, and a search method through the job scheduler.

The fourth step may include: performing a search based on the similar document search information received from the client in the first to nth fires, and then transmitting a search result to the set manager; Storing the search results received from the first to nth fires in the set manager, and transmitting a result set number and the number of documents to the first to nth fires; And transmitting the result set number and the number of documents received from the set manager to the client in the first to nth fires.

In the fifth step, the client transmits, through the job scheduler, search result list information including a result set number received when searching the first to nth fires, a section to be displayed, and the number of search result lists to be received. It is done.

The fifth step may include transmitting the search result list information received from the client in the first to nth fires to the set manager; Transmitting as many documents as the number of search result lists included in the search result list information received from the first to nth fires to the first to nth fires in the set manager; And retrieving section information requested by the client with the documents received from the set manager in the first to nth fires, and transmitting a search result to the client as the document.

In the sixth step, the client transmits the document number to the first to nth fires through the job scheduler.

In the sixth step, the first to nth fires transmit document contents of a document corresponding to the document number received from the client to the client.

Another information retrieval management method according to the present invention for achieving the above object,

An information retrieval management method of an information retrieval management system including a job scheduler, first to n-th fires, set manager, and data manager,

Transmitting, by the client, a document and database information to be inserted or changed to the job scheduler;

Transmitting, by the job scheduler, information transmitted by the client to the data manager;

Transmitting a result after storing or changing a document in the data manager to the job scheduler;

Instructing a set manager to update the result set if the insert or change is successful in the job scheduler;

Transmitting a command execution result to the job scheduler in the set manager;

Sending a final result from the job scheduler to the first to nth fires; And

And a method of inserting and modifying a document, the method comprising: reopening a database in the first to nth fires and transmitting a result to the client.

The information retrieval management method includes transmitting an ID of a document to be deleted from the client to the job scheduler; Transmitting, by the job scheduler, an ID of a document to be deleted received from the client to the data manager; Transmitting a result to the job scheduler after deleting a document in the data manager; Instructing a set manager to update the result set if the deletion is successful in the job scheduler; Transmitting a command execution result to the job scheduler in the set manager; Sending a final result from the job scheduler to the first to nth fires; And re-opening the database in the first to nth fires and deleting the one document having the result of transmitting the result to the client.

Hereinafter, with reference to the accompanying drawings will be described in detail an embodiment of the present invention.

1 is a block diagram showing the overall configuration of an information retrieval management system according to the present invention.

As shown in FIG. 1, the information retrieval management system is largely composed of a kernel 10, a storage engine 20, a search engine 30, an indexer 40, a dictionary database 50, a data manager ( 60), the list database volume (70), and the document database volume (80).

The kernel 10 is an infrastructure of the information retrieval management system, and is responsible for data input and output between the physical disk and the user memory, and is largely a file and directory manager 11, a record manager 12, and a reverse file manager 13. And a transaction manager 14 and an input / output manager 15.

Here, the file and directory manager 11 is a unit used to form records or inverse files that are logically the same set in physical disk space, and is allocated in extent units that are a bundle of pages. The physical identifier of the disk where the record is stored can be mapped to a logical identifier to facilitate access and issue unique logical identifiers for multiple volumes.

The record manager 12 supports management of an object that can be stored in one page internally and a long data item of more than one page, and sequentially accesses a record in a file from the beginning, and inserts and modifies the record. It also supports update operations related to deletes.

The reverse file manager 13 is implemented using a B tree, and the key part may use one of an integer, a real number, and a variable length string, and an element part stores a user-defined structure. To save storage space, the key can be compressed and provide functions such as insert, modify, delete, and search.

The transaction manager 14 provides a series of functions for stably performing a unit of work designated by a user. In other words, it provides the start and end of a transaction, abortion, a save point function, and record the transaction information to a log file, and use checkpoints to reduce recovery time. If a given job fails in the middle, all jobs return to the beginning of the transaction to maintain the integrity of the data.

The input / output manager 15 includes a page manager for allocating, deleting, and maintaining a page on a physical disk, a buffer manager for mapping a page of a disk to a page of a memory, and an object having different requirements. It consists of a lock manager used for access, and the buffer of the storage system is implemented as shared memory so that multiple processes can access it at the same time.

Next, the storage engine 20 is designed using the kernel 10 as a substructure, and includes a catalog manager 21, a document manager 22, and an index manager 23.

Here, the catalog manager 21 manages information about a structure of a database to be constructed, such as a document structure, an indexing method, basic key information, compression, and stopwords. The document manager 22 converts the original document into an internal document structure to perform operations for insertion, deletion or modification of an existing document, and the index manager 23 is extracted using the indexer 40. Organize index information into a structure suitable for searching.

The storage engine 20 has no limitation on the size of a record (object), and provides a recovery function through transaction processing for stable online insertion and deletion. It also supports various types of object storage and retrieval such as structured documents such as XML and SGML and multimedia data.

Next, the search engine 30 greatly supports a Boolean model and an extended vector model. In this case, the boolean model is an extension of the basic boolean as Boolean operators (and, or, not), proximity operators (near, within), relational operators (=, <=, <,> =,>) and range operators (range). , And truncation operators. The extended vector model is an extension of the pure vector model that produces search results based on the calculation of the similarity value between the user query and the document. The extended vector model is based on vector operations, except for the proximity operator of the Boolean model.

Next, the indexer 40 provides a variety of index formats to allow the user to select a suitable index format for a particular section, it is based on Unicode can easily process documents composed of Gore / Chinese and multilingual. For English, it provides stemming options for indexing the roots of words and for Chinese characters, the option for indexing by converting to Korean. Hangul is processed using real morpheme dictionaries, ending dictionaries, search dictionaries, combined auxiliary term dictionaries, and formal morpheme dictionaries for processing stopwords.

Next, the data manager 60 loads and indexes a large amount of bulk data in a database using a schema created by an administrator. In addition to the plain text file, the data used for loading includes a compressed file compressed in a gzip, and can be compressed and stored for space efficiency. In addition, for Unicode support, documents in other formats are encoded and stored internally in UTF-8. When loading, the primary key is checked to prevent the loading of the same document, and the data is effectively placed in a contiguous space on the disk to quickly perform large documents and indexes and facilitate access later.

Next, the list database volume 70 stores a list database, and the document database volume 80 stores one or several document databases in a distributed manner. Here, the document database stores the actual documents to be built and the indexes of the documents, and the list database contains information about the document structure, indexing method, primary key, compression, and stopwords for the databases. Save them and keep them.

As shown in FIG. 1, the document database is distributed and stored over one or more storage devices. As part of the method for processing large databases, multi-threaded distributed search is used to improve the search performance of large databases.

The information retrieval management system according to the present invention assigns a logical document identifier (RID) to each record to identify the records in multiple databases. The logical document identifier is automatically assigned by the system to be unique throughout the database group and consists of a volume identifier and a record number.

§ Document structure

The information retrieval management system according to the present invention mainly supports retrieval of atypical text documents such as 'book books', 'research reports', 'papers' or 'newspaper articles'. In this system, a database is a collection of text document entities, and a document consists of sections such as 'Title', 'Author', and 'Abstract'. This can be likened to a hierarchy in which a table (or relation) in a relational database consists of records, where each record consists of several fields (or attributes). In the information retrieval management system according to the present invention, sections constituting a document are classified into three categories as follows.

Basic Section

The basic section is the minimum document component that constitutes a document. You can use the default section as the basic unit for specifying the scope of the document search during the database search. For example, the title, author's name, author's institution, abstract, and text can all be basic sections of a paper. In addition, the user can direct the execution of the query by limiting the search scope to two basic sections: 'paper title' and 'green'. Basic sections can also be used as a unit for document indexing. In other words, the information retrieval management system according to the present invention constructs index files for each basic section for quick retrieval of documents. The content of the basic section of the information retrieval management system supports both strings and numbers.

Virtual Section

In the information retrieval management system according to the present invention, in order to support various index types for basic sections, a new section having a single physical meaning can be configured. This type of section is called a virtual section. Virtual sections can also be combined with base sections to form joining sections. For example, you can construct virtual sections of 'Title 2' and 'Title 3' for the basic section of 'Title' and configure various types of indexes such as partial match (INDEX_BY_MA) and exact match (INDEX_AS_IS). .

Union Section

In the information retrieval management system according to the present invention, a plurality of basic sections and virtual sections can be combined to form a new section having a logical meaning. Such a section is called a joining section, which is a fictitious section that is perceived as a single section from the user's point of view without actually indexing. For example, a combined section called 'title' may be formed by combining a basic section or a virtual section of 'subtitle', 'subtitle', 'subtitle'.

§Index form

In general, the index not only speeds up the search of the system but also has a great effect on the search effect. Therefore, specifying the proper indexing method can greatly improve the performance and quality of the system. The information retrieval management system according to the present invention supports various indexing methods for indexing documents, and the database designer applies one of these indexing methods to each basic section and virtual section of the document. There are also options for Chinese characters and STEMMING processing. If the HANJA2HANGLE option is TRUE, if the document is in Chinese, convert the Hanja to Hangul and extract the index word according to the indexing method.The HANJA2HANGLE option is FALSE. ), Extracts index words according to the index method without converting to Hangul. In English, stemming is only done when the stemming option is TRUE.

The information retrieval management system according to the present invention includes six basic indexing methods such as INDEX_AS_IS, INDEX_BY_TOKEN, INDEX_BY_MA, INDEX_BY_CHAR, INDEX_AS_NUMERIC, INDEX_AS_IS_MA, and the like.

INDEX_AS_IS

The INDEX_AS_IS extracts the entire contents or values of a section from a document as one index word. Doing so supports searching for exact matching on that section. Performs a search at the level that checks only Character String equality for the section indexed with INDEX_AS_IS. It is possible to compare string values using the relational operators (<, <=,>,> =,>). The INDEX_AS_IS indexing method is mainly used for basic sections that are uniquely assigned to distinguish records, such as 'control number'.

Example) Title: "Research on Information Retrieval"

Index: "Research on Information Retrieval"

Ex) Title: "硏 究 about 情報 檢索"

Index: "硏 究 about 情報 檢索"

Ex) Control number: "AN00012"

Index word: "AN00012"

INDEX_BY_TOKEN

The INDEX_BY_TOKEN is applicable to a section that needs to support content-based partial matching search such as text search. The index method of INDEX_BY_TOKEN selects index words from among clauses or words in the section, rather than using the entire contents or values of the section as index words. The INDEX_BY_TOKEN is a basic indexing method for extracting words or words except for stopwords from a section as an index word, and uses an index word as it appears in the original text without performing post-processing. Therefore, this method is suitable for sections that mainly contain proper nouns such as' person's name 'or' name 'or' sections that do not require special post-processing, such as' keyword list 'of the paper.

If HANJA2HANGLE

Example) Title: "Research on Information Retrieval"

Index Terms: "Information Search", "About", "Research"

Ex) Title: "硏 究 about 情報 檢索"

Index: "情報 檢索", "in", "about", "硏 究"

If HANJA2HANGLE

Example) Title: "Research on Information Retrieval"

Index Terms: "Information Search", "About", "Research"

Ex) Title: "硏 究 about 情報 檢索"

Index terms: "information search", "about", "research"

INDEX_BY_MA

The INDEX_BY_MA goes one step further to the indexing method of INDEX_BY_TOKEN and uses a Morphological analyzer for indexing Korean text and optionally uses a stemmer for English text. Morphological analysis is performed on each sentence of Hangeul text to classify the smallest morphological units such as nouns, surveys, suffixes, verbs, and adjectives, and then extracts simple nouns that can represent the contents of the section. In the case of English, stems are extracted by processing regular plurals and verb tense changes. Thus, this method performs a better index than the simple index word index of INDEX_BY_TOKEN. Thus INDEX_BY_MA is suitable for basic sections such as 'paper title' or 'abstract'.

If HANJA2HANGLE

Example) Title: "Research on Information Retrieval"

Index terms: "information", "search", "information search", "research"

Ex) Title: "硏 究 about 情報 檢索"

Index terms: "情報 檢索", "硏 究"

Ex) Title: "Information Systems"

Index words: "informat", "system"

If HANJA2HANGLE

Example) Title: "Research on Information Retrieval"

Index terms: "information", "search", "information search", "research"

Ex) Title: "硏 究 about 情報 檢索"

Index terms: "information", "search", "information search", "research"

INDEX_BY_CHAR

The INDEX_BY_CHAR generates index words by extracting one syllable in English and two syllables in Korean. This method is mainly used for the section like 'person's name' in the basic section containing Hangul content. In other words, if a person's name is freely available in Korean, such as 'Hong Gil-dong', 'Hong Gil-dong', and 'Hong Gil-dong', if the user's name is indexed and stored in syllable units, such as 'Hong', 'Gil', 'Dong', Searches for a person's name, regardless of spacing.

If HANJA2HANGLE

Ex) Name: "Hong Gil Dong"

Index words: "hong", "gil", "dong"

Ex) Name: "洪 吉 童"

Index terms: "洪", "吉", "童"

If HANJA2HANGLE

Ex) Name: "Hong Gil Dong"

Index words: "hong", "gil", "dong"

Ex) Name: "洪 吉 童"

Index words: "hong", "gil", "dong"

● INDEX_AS_NUMERIC

The INDEX_AS_NUMERIC extracts an index word for a basic section composed of numbers. It is assumed that the sections designated as INDEX_AS_NUMERIC have only a single atomic value, and a relational operation (<, <=,>,> =, =,-) is possible for these sections. Diagonal styles are often used for sections that have a uniform, numeric form, such as date, year, and date.

Ex) Date: "19961214"

Index word: 19961214

INDEX_AS_IS_MA

The INDEX_AS_IS_MA is an index method that satisfies both the above-described INDEX_AS_IS index method and INDEX_BY_MA index method simultaneously. That is, by extracting the entire contents of a section from the document as an index word, and performing morphological analysis on each word of Hangul text, the minimum morphological units such as nouns, surveys, suffixes, verbs, and adjectives are classified. Extract a simple noun that can be represented as an index word.

If HANJA2HANGLE

Example) Title: "Research on Information Retrieval"

Index terms: "Research", "information", "search", "information search", "research"

Ex) Title: "硏 究 about 情報 檢索"

Index: "硏 究 about 情報 檢索", "情報 檢索", "硏 究"

If HANJA2HANGLE

Example) Title: "Research on Information Retrieval"

Index terms: "Research", "information", "search", "information search", "research"

Ex) Title: "硏 究 about 情報 檢索"

Index terms: "Research", "information", "search", "information search", "research"

DO_NOT_INDEX

The index indicated by the DO_NOT_INDEX is not performed. Currently no search path is provided for sections specified in these ways.

2 is a conceptual diagram conceptually illustrating a loading process of a database of the information retrieval management system according to the present invention.

In FIG. 2, the database schema file 101, the source documents file 102, and the stopwords list 103 are files that a database administrator needs to prepare.

The database schema file 101 describes information about a database directory and document volume, a database group, a section definition, a section indexing method, a structure of a source document, a method of loading a source document into a database, and the like. .

The raw document files 102 are raw documents to be loaded into a database, and documents having heterogeneous structures may be loaded into a single database through the structure of the raw documents described in the database schema file 101. .

The stopwords list 103 is a list of stopwords to be loaded into a database and is referred to only when searching and excluded from a search target. If the user wants to include the stopword in the search, the user should search by prefixing the query with a '+'.

When the database schema file 101, the source document file 102, and the stop word list 103 are prepared, the database is loaded using the loader 110. The loader 110 is a program used to load a database, and imports all information necessary for loading a database from the database schema file 101. The loader 110 is a type of command line interpreter that reads a schema file through an option on a command line, and an administrator may input a command for creating a database and loading a document to the program. . It is convenient for the administrator to create and use a file composed of instructions for input to the loader 110 for convenience. In the information retrieval management system of the present invention, this is called a database schema file (or schema file).

3 is a conceptual diagram conceptually showing a form of communication with the configuration processes of the information retrieval management system according to the present invention.

As shown in FIG. 3, the information retrieval management system includes a job scheduler 210, first to nth fires 220/1 to 220 / n, and a set manager. (230), a data manager (240), and a client (not shown), which are composed of five kinds of processes, and processes transmit data through socket communication or pipe communication. The functions of the above processes are as follows.

The job scheduler 210 accepts the connection request from the client and distributes the job according to the states of the first to nth fires 220/1 to 220 / n. When the online document is managed, the document manager is requested to manage the document and the result is transmitted to the client. In addition, when the database is changed, the data manager 240, the first to nth fires 220/1 to 220 / n, and the set manager 230 are notified that the database has been changed.

The first to nth fires 220/1 to 220 / n have a main function of performing a search and hand over the service request from the job scheduler 210 to the client after the operation. Then, a request is made to store the search results to the set manager 230.

The set manager 230 mainly stores and manages a search result, and after the service request from the first to nth fires 220/1 to 220 / n is operated, the first to The result is passed to the nth fires 220/1 to 220 / n.

The data manager 240 is a main function of online document management, and after the job request from the job scheduler 210 is returned to the job scheduler 210. Pass it over.

The client transmits a user's request to the job scheduler 210.

There are three types of client-side services of the information retrieval management system according to the present invention: retrieval service, online document management service, and error processing service. to be.

4A through 4F are conceptual views conceptually showing a search service on a client side of an information retrieval management system according to the present invention.

First, FIG. 4A is a conceptual diagram illustrating how a client obtains information of a database.

As shown in FIG. 4A, when the search service requests database information from the client 310 (step S1), the job scheduler 320 first to idle the database information requested by the client 310. The data is transmitted to the nth fires 330/1 to 330 / n (step S2). Next, the first to n-th fires 330/1 to 330 / n retrieve the database information requested by the client 310 and transmit it to the client 310 (step S3).

4B is a conceptual diagram illustrating how a client obtains a section list of a database.

As shown in FIG. 4B, the search service requests a section list of the database from the client 310 (step S11). Then, the job scheduler 320 transmits the section list of the database requested by the client 310 to the first to nth fires 330/1 to 330 / n in the idle state (step S12). Next, the first to nth fires 330/1 to 330 / n retrieve the section list of the database requested by the client 310 and transmit it to the client 310 (step S13).

4C is a conceptual diagram illustrating a method of requesting a search by a client.

As shown in FIG. 4C, the search service transmits information such as a query, a search target database list and a section, to the client 310 (step S21). Next, the job scheduler 320 transmits the information transmitted from the client 310 to the first to nth fires 330/1 to 330 / n in the idle state (step S22). Next, the first to nth fires 330/1 to 330 / n search for the information transmitted from the client 310 and transmit the search result to the set manager 340 (step S23). Next, the set manager 340 stores the received search result and transmits the result set number and the number of documents to the first to nth fires 330/1 to 330 / n (step S24). Next, the first to n th fires 330/1 to 330 / n transmit the search information received from the set manager 340 to the client 310 (step S25).

4D is a conceptual diagram illustrating a method of requesting a similar document search by a client.

As shown in FIG. 4D, the search service transmits information such as a search document number, a search target database list, a section, a search method, and the like (step S31). Then, the job scheduler 320 transmits the information transmitted from the client 310 to the first to nth fires 330/1 to 330 / n in the idle state (step S32). Next, the first to nth fires 330/1 to 330 / n search for the information transmitted by the client 310 and transmit the search result to the set manager 340 (step S33). Next, the set manager 340 stores the received search result and transmits the result set number and the number of documents to the first to nth fires 330/1 to 330 / n (step S34). Next, the first to nth fires 330/1 to 330 / n transmit the search information received from the set manager 340 to the client 310 (step S35).

4E is a conceptual diagram illustrating how a client requests a search result list.

As shown in FIG. 4E, the search service transmits the result set number obtained when the client 310 searches, the section to be shown, and the number of results to be obtained (step S41). Next, the job scheduler 320 transmits the information transmitted from the client 310 to the first to nth fires 330/1 to 330 / n in the idle state (step S42). Next, the first to nth fires 330/1 to 330 / n transmit the request information transmitted from the client 310 to the set manager 340 (step S43). Next, the set manager 340 transmits the requested number of search results to the first to nth fires 330/1 to 330 / n (step S44). Next, the first to n th fires 330/1 to 330 / n transmit the search information received from the set manager 340 to the client 310 (step S45).

4F is a conceptual diagram illustrating how a client requests the original text of a document.

As shown in FIG. 4F, the search service transmits the document number to the job scheduler 320 at the client 310 (step S51). Then, the job scheduler 320 transmits the document number received from the client 310 to the first to nth fires 330/1 to 330 / n in the idle state (step S52). Next, the first to nth fires 330/1 to 330 / n transmit the document contents of the corresponding document to the client 310 (step S53).

5A and 5B are conceptual views conceptually illustrating an online document management service of the information retrieval management system according to the present invention.

First, FIG. 5A is a conceptual diagram illustrating a method of inserting and changing a single document.

As shown in FIG. 5A, in the method of inserting and changing the single document, the client 310 transmits inserting and changing information such as a document and a database name to be inserted or changed to the job scheduler 320 (step S61). ).

Next, the job scheduler 320 transmits the insertion and change information received from the client 310 to the data manager 350 (step S62).

Then, the data manager 350 transmits the result after storing or changing the document to the job scheduler 320 (step S63).

The job scheduler 320 then instructs the set manager 340 to refresh the result set if the insertion or change is successful (step S64).

Next, the set manager 340 transmits a command execution result to the job scheduler 320 (step S65).

The job scheduler 320 then sends the final result to the first to nth fires 330/1 to 330 / n in idle state (step S66).

Next, the first to nth fires 330/1 to 330 / n open the database again and transmit the result to the client 310 (step S67).

5B is a conceptual diagram illustrating a method of deleting one document.

As illustrated in FIG. 5B, the method of deleting one document transmits an ID (ID) of the document to be deleted from the client 310 to the job scheduler 320 (step S71).

Next, the job scheduler 320 transmits the ID (ID) of the document to be deleted received from the client 310 to the data manager 350 (step S72).

Then, the data manager 350 transmits the result after deleting the document to the job scheduler 320 (step S73).

The job scheduler 320 then instructs the set manager 340 to refresh the result set if the deletion is successful (step S74).

Next, the set manager 340 transmits a command execution result to the job scheduler 320 (step S75).

The job scheduler 320 then sends the final result to the first to nth fires 330/1 to 330 / n in idle state (step S76).

Next, the first to nth fires 330/1 to 330 / n reopen the database and transmit a result to the client 310 (step S77).

The information retrieval management system according to the present invention basically supports Boolean and extended vector models. The boolean model is an extension of the basic booleans: Boolean operators {AND (&), OR (|), NOT (!)}, Proximity operators {NEAR (/ N), WITHIN (/ W)}, relational operators (=, < =, <,> =,>), Range operators (~), and truncation operators. In addition, the vector model is an extension of the pure vector model, which is based on vector arithmetic, but allows other operators except the proximity operator of the Boolean model.

The following describes a basic method for creating a query in two basic models, Boolean and extension vectors, provided by the information retrieval management system according to the present invention.

Since the database search engine of the information retrieval management system uses a search target identifier called a section as a basic unit of the search, the query is a search composed of a search target identifier, a search scope indicator (section designator, a relational operator, a range operator), and words. It begins with the composition of words. These search terms may be combined by one or more Boolean Operators / Proximity Operators and parentheses (note that there is no search operator specifically for vector searches).

The following shows the basic format of a query statement.

<Query> = (<query>)

<Query> <boolean operator> <navigator>

<Navigation word> <proximity operator> <navigation word>

<Navigation words>

<Navigation words> = <section list>: <list of search words>

<Section list>: (<search word list>)

<Section list> <relational operator> <search word>

<Section list> <search word> <range operator> <search word>

<Section list> = <section list>, <section>

<Section>

<Search word> = "<word>"

+ <Word>

<Word> *

Here, several sections can be specified by using ','. The search word list may be formed of a string separated by spaces.

● How to specify the section to search (How to configure search terms)

In the information retrieval management system of the present invention, a section corresponds to a field of a database and is similar to field condition designation in SQL, but reflects the data type according to the indexing method of the information retrieval management system in designating a specific section. . Table 1 shows how search terms are constructed according to index types. The information retrieval management system provides a total of six indexing methods. The section designator (:) is used to designate a string type search word and can be used in a total of five string index types, and a relational operator (=, < ,>, <=,> =) Are used for numeric and string parts, and can be used for two index types as shown in Table 1. Finally, the range operation has the same scope as the relational operator.

Search term composition by index type Designator Data type Index type Search term configuration example Section specifier (:) String INDEX_BY_MAINDEX_BY_CHARINDEX_AS_ISINDEX_BY_TOKENINDEX_AS_IS_MA SECTION: WORDSECTION: WORD_LISTSECTION: "WORD" SECTION: "WORD_LIST" SECTION: + WORD Relational operator String INDEX_AS_IS SECTION = WORDSECTION <WORDSECTION> WORDSECTION <= WORDSECTION> = WORD Numeric INDEX_AS_NUMERIC Range Operator (~) String INDEX_AS_IS SECTION: WORD-WORD Numeric INDEX_AS_NUMERIC

The following is a simple example of how to specify the section to be searched. Assume that the title (TITLE) is indexed by string type and the date (DATE) is indexed by numeric type.

TITLE: Gene

TITLE: Genetic Information

TITLE: (Household Computer Systems)

DATE> 1990

DATE: 1990 ~ 2000

As shown in the example above, you need to know the data type according to the index type and compose the search term to get the correct search result. Of course, if you do not consider the type properly, an error code and message will be displayed.

However, for compatibility with existing information retrieval management system queries, a grammar not specified in Table 1 is provided, which allows the use of section specifiers (:) for numeric types. That is, in the above example, the DATE type can be used as in (1), and its meaning is as in (2).

DATE: 1990 ....................... (1)

DATE = 1990 .................. (2)

In the information retrieval management system of the present invention, one section is basically designated as a search word. However, when two or more sections to be searched are specified, a comma (,) and a section name can be entered in order. (3) designates two sections, title (TITLE) and content (CONTENT), for the search term "information search", which has the same meaning as (4).

TITLE, CONTENT: Information Retrieval ......... (3)

TITLE: Information Search ｜ CONTENT: Information Search ....... (4)

As such, the section specifier ':' can be used for all index sections, regardless of index type.

● Scope to which the specified search target section affects

In principle, every search word must contain a section, otherwise an error occurs. Only the section elimination according to the meaning of the query context is allowed, but only if the section designation of the previous search word is inherited. An example is described below. However, the omission is allowed only when the section specifier (:) is used, and the section must be specified when a relational or range operation is used.

∥ If you allow

In the case of Table 2, the user wants to find 'car audio system' in the title (TITLE), which should be expressed like the right column of Table 2, but allows the form of query in the left column for ease of use. The use of parentheses in the left column is acceptable, and 'car', 'audio', and 'system' are all assumed to be section designations called TITLE.

Tolerance vs. Material Allow query Meaning of raw quality TITLE: car & audio & systemTITLE: (car & audio & system) TITLE: (car) & audio & systemTITLE: ((car & audio) & system) TITLE: car & TITLE: audio & TITLE: system

¶ If not allowed

(5) is also allowed, in which case 'system' is possible because it inherits the previous section 'TITLE' by default, and in (6) 'car' is an error because no previous section exists. Take care.

(TITLE: car & audio) & system ...... (5)

car & TITLE: audio & system ...... (6)

Section omission is not allowed for the relational operator and the range operator. An example of this is shown in Table 3.

Example where section omission is not allowed Use of relational operators Use of range operators DATE> 1990 & 2000 DATE> (1990 & 2000) DATE = 1990 & 2000 DATE = (1990 & 2000) 1990-2000

● Operator Priority

When using a combination of operators in a query statement, the operation priority is applied to determine the operation rank between operators. Table 4 illustrates the priority of the operators provided by the search engine of the information retrieval management system of the present invention.

Operator precedence Priority Operator High ↑ ↓ Low Terminology Specifier String Specifier Bracket Section Specifier Relation Operator Scope Operator Proximity Operator Boolean Operator + "" ();, => <> = <=-/ W / NAND (&) OR (|) NOR (!)

● Stopword Processing

When a database administrator loads data using the information retrieval management system data loader, it is possible to assign stopwords to each section. In this case, the designated stop word is used in the search. When the user inputs a search word, the searcher compares the pair (section, search word) entered by the user with the stop word and excludes the registered stop word from the search. Excluding such stopwords from the search is generally used because it improves the performance of the full-text search.

However, a special designator ('+') is provided for users who want to include a stopword in the search, by prefixing the '+' to the search word so that the word is included in the search even if it is a stopword.

For example, if 'system' is a stopword, (7) and (8) suggest different search results.

TITLE: Car Audio System ............... (7)

TITLE: Car Audio + System ............... (8)

(7) finds documents containing only 'car' and 'audio' in the title (TITLE), and (8) finds documents containing all three search words.

When using a stopword designator, you must use it without a space in front of the search word together with the search word.

TITLE: Search for information + &

However, it can be used with the string specifier described below.

TITLE: + "Search for information"

● Use of string specifiers

The string specifier ("") is a method to separate the whole word into one search word regardless of spaces or special characters when separating the search words. The content specified in the string designator is one whole regardless of the index type of the section. It is considered an index word. Mainly useful in INDEX_AS_IS.

FIG. 6 illustrates that the title section of the document is indexed according to two indexing methods. In the INDEX_AS_IS method, a single index word is generated because no spaces or special characters are considered, and the INDEX_BY_TOKEN method uses delimiters. You can see that three index words are generated.

In this case, if the user indexes the title (TITLE) in the same way as INDEX_AS_IS, a correct result can be obtained only when constructing a query using a string specifier as in (9). This is because a list of words that do not use string specifiers, such as (10), generates multiple index words.

TITLE: "AN20020808 Announcements" ...... (9)

TITLE: AN20020808 Notice ...... (10)

For example, (9) finds the correct document with the same title (TITLE), regardless of spaces or special characters, while (10) generates three search words during query processing. As a result, the user does not get the expected results.

Boolean search

Boolean search is the most commonly used search model in search systems. The operators supported in Boolean search are shown in Table 5. The operators must be capitalized.

Boolean operators Relational operator Range operator Proximity operator Truncation operator & (AND) | (OR)! (NOR) = <= <> => To N (NEAR) W (WITHIN) *

Boolean operation search

All Boolean operators used in the information retrieval management system are binary operators. Here is an example of a Boolean operation:

ABS: (Information & Search) ............... (11)

ABS: (Information | Search) ............... (12)

TI: (Information! Search) ............... (13)

ABS :! Search ............... (14)

The search result of the query 11 searches for a word including both "information" and "search" in the ABS section. Query 12 searches for words in one or more of the words “information” or “search” in the ABS section. Query 13 searches for documents in the TI section that do not contain the word "search" among the documents that contain "information". The query 14 is an operation that is not supported by the information retrieval management system. remind '!' Operator has the meaning of "AND NOT".

● Relational operation search

DATE> = 20020101 ......... (15)

The search result of query (15) retrieves all documents in the DATE section that contain dates that are greater than or equal to January 1, 2002.

Proximity calculation search

ABS: (Info./W2 Search) ......................... (16)

ABS: (Info./N2 Search) ......................... (17)

The search result of the query 16 searches for words in the ABS section that contain both "information" and "search" in the order of the adjacent words and the distance between words is 2 or less. The search result of the query (17) searches for the words in the ABS section that are less than or equal to two words in length and that both "information" and "search" include adjacent words in any order.

Cutting operation search

The truncation operator supported by the information retrieval management system is '*' and can be used only for right truncation.

ABS: Information * ......... (18)

The search result of the query 18 searches for all words including "information", such as "information", "information search", "information system", "information processing", and the like in the ABS section.

String Expansion Search

There are two types of string expansion: compound noun expansion and multiple string expansion. These are the methods you use to improve recall and precision.

First, in compound noun extension, compound nouns are decomposed into the form of each constituent noun, and then combination of operators improves reproducibility and accuracy. For example, suppose a user enters the query "TI: information retrieval" and TI is indexed with INDEX_AS_MA. The retrieval system gives 'information search' as the input to the indexer and obtains two decomposition results of the construct nouns 'information' and 'search'. A query like (20) is generated by applying the WITHIN operator between constructs of the same level.

TI: Information Retrieval ......... (19)

→ TI: (Info / W1 Search) ........ (20)

Compound nouns in query (19) are expanded into query (20). The search result of the query 19 searches for adjacent documents in the order of "information" and "search" in the TI section.

In addition, the form of a similarly processed query is as follows.

Words concatenated with word separators other than spaces: red-eared

Second, the multi-string expansion is a '&' which is used as a method of applying an operator when entering multiple search words separated by spaces. However, it allows the user to change its meaning and use it.

TI: Information Retrieval Research ............... (21)

→ TI: (Information & Search & Research) ........ (22)

The query 21 is processed internally by the search system in the form of 22.

TI: Information Retrieval Research ......... (23)

→ TI: ((Information & Search) & Research) ........ (24)

The query (23) is a mixture of compound nouns and several search words, and is processed as (24).

● Extended Vector Search

In the information retrieval management system, extended vector retrieval is a model that supports weight calculation using the WITHIN / NEAR proximity operator in addition to basic vector retrieval. Basically, it is a way to take advantage of vectors and boolean searches by supporting vector operations on those given as vector representations and Boolean operators on those where Boolean operations are used. The operators are shown in Table 6.

Boolean operators Relational operator Range operator Proximity operator Truncation operator & (AND), ｜ (OR),! (NOT) = <= <> => To N (NEAR) W (WITHIN) *

Basic vector search

TI: Information Retrieval System ......................... (25)

The search result of query 25 retrieves the ranked documents by applying the vector weight for the "information retrieval system" in the TI section.

Boolean operation search

TI: Information Retrieval & ABS: Retrieval System ......... (26)

TI: Information retrieval | ABS: Retrieval processing ......... (27)

The search results of query 26 intersect the set of documents ranked by applying vector weights for "information retrieval" in the TI section and the set of documents ranked by applying vector weights for "search system" in the ABS section. Search for (Intersection) documents. The search result of query 27 is a set of documents ranked by applying vector weights for "information retrieval" in the TI section and a set of documents ranked by applying vector weights for "search processing" in the ABS section. 'Search the processed document.

Relational operation search

DATE> = 20020101 ..................... (28)

The color search result of query (28) retrieves all documents in the DATE section that contain dates whose date is greater than or equal to January 1, 2002.

Cutting operation search

ARS: Information * ..... (29)

The search result of the query 29 retrieves all words in the ABS section including "information" such as "information", "information search", "information system", "information processing", and the like.

Compound Noun Expansion Search

TI: Information Retrieval ... (30)

→ TI: Search for information / W ... (31)

In query 30, the compound noun expands into query 31. The search result of the query 31 searches the ranked documents by applying vector weights and distance weights for "information search" in the TI section.

The present invention is not limited to the above-described embodiments, but can be variously modified and changed by those skilled in the art, which should be regarded as included in the spirit and scope of the present invention as defined in the appended claims. something to do.

As described above, according to the information retrieval management system and the method according to the present invention, by configuring a stable management function in addition to the retrieval function for the database, it is possible to build a low-cost and high-efficiency information system.

In addition, by supporting Unicode, it is possible to support Gore / Chinese and multilingual languages at the storage engine level.

In addition, by supporting compression of the document and the index database, the space for the index database, which is about three times the document size, can be reduced to efficiently use the storage space.

In addition, through the development of standardization and management techniques based on the construction of terminology dictionary for specialized fields for indexers, the linguistic data infrastructure, which is the basis of Korean information processing technology, is established through continuous expansion of terminology dictionary. Through a systematic approach, it is possible to prevent colonization of linguistic information processing technology and to lay the foundation of research for domestic researchers.

In addition, multiple processors share the functionality of the entire system, minimizing the load on the system and ensuring reliable online transaction processing and data management. In addition, data addition, change, and deletion in the existing information retrieval system cannot be simultaneously processed while performing a retrieval service. In the system of the present invention, a processor is divided into four processors and only a document management agent exists and data is changed. Since there is a separate processor that detects whether the data is lost and propagates it to the process in charge of the search service, the data is changed online while the search service is naturally performed. Therefore, it is highly useful in internet service fields such as specialized portals and corporate portals, electronic library fields, thesis retrieval systems, electronic document management systems (EDMS) / groupware, and even knowledge management systems (KMS).

Claims (23)

A list database volume that stores the list database; A document database volume in which one or several document databases are distributed and stored; A dictionary database volume in which one or several dictionary databases are distributed and stored; A kernel that performs data input / output between the list and document database volumes and user memory, and manages files, directories, records, and inverse files in the list and document database volumes; A storage engine for performing catalog management, document and index management using the kernel to store meta information about a database; A search engine that searches for a user's query; An indexer for extracting an index word for the document input by the user using the dictionary database; And An information retrieval management system comprising a data manager that receives a schema file created by an administrator, creates a database, bulk loads a bundle of source documents, and performs an index. The method of claim 1, wherein the kernel, And managing a page and a buffer for accessing the list and document database volumes, and performing a logging and locking function therefor. The method of claim 1, wherein the kernel, A file and directory manager for facilitating access by mapping a physical identifier of a disk where a record is stored to a logical identifier, and for issuing logical identifiers for multiple volumes; It supports the management of objects that can be stored on one page internally and long data items more than one page, and the ability to sequentially access records in the file from the beginning, and update operations related to inserting, modifying, and deleting records. A record manager; A reverse file manager which compresses a key for efficient storage space and performs insertion, modification, deletion, and retrieval; A transaction manager providing a start and end of a transaction, aborting, a save point function, and a function of recording a transaction information in a log file; And And an input / output manager configured to manage data input / output between the list and document database volumes and user memory. The method of claim 3, wherein The transaction manager uses a checkpoint to reduce recovery time, and if a specified job fails in the middle, all the jobs are returned to the starting point of the transaction to maintain the integrity of the data. The method of claim 3, wherein the input / output manager, A page manager for allocating, deleting, and maintaining pages in the list and document database volumes; A buffer manager for mapping the pages of the disk to the pages of the memory; And An information retrieval management system comprising a lock manager for different requirements to access the same object. The method of claim 1, wherein the storage engine, A catalog manager that manages information about a structure of a database to be built; A document manager converting the original document into an internal document structure and performing operations for insertion, deletion or modification of an existing document; And And an index manager configured to configure the index information extracted using the indexer into a structure suitable for searching. The method of claim 1, And information about the structure of the database includes a document structure, an indexing method, primary key information, compression, and stopwords. A list database volume that stores the list database; A document database volume in which one or several document databases are distributed and stored; A database schema file describing information about the directory and volume of the list and document database, a database group, a section definition, a section indexing method, a structure of a raw document, and a loading method from a source document to a database; A raw document file consisting of raw documents to be loaded into the document database; A stopword list to be loaded into the list database; And And a loader configured to perform the document database generation and the document loading by the information received from the database schema file, and to receive a command from the administrator for the document database generation and the document loading. The method of claim 1, And a document having a heterogeneous structure can be loaded into a single database through a structure of a source document described in the database schema file. It distributes the work according to the status of the first to nth fires received from the client, requests document management from the data manager during online document management, and transmits the result to the client. A job scheduler for notifying the first to nth fire, set managers that the database has been changed; First to n-th fires that perform a job on a service request received from the job scheduler, transmit a result to the client, and request to save a search result to a set manager; A set manager that performs a task on a service request received from the first to nth fires and transmits a result to the first to nth fires; And And a data manager for performing a job on the service request received from the job scheduler and transmitting the result to the job scheduler. An information retrieval management method of an information retrieval management system comprising the job scheduler of claim 10, first to nth fires, set manager, and data manager, A first step of a client requesting and receiving information of a database from the first to nth fires through the job scheduler; A second step in which the client requests and receives a section list of the first to nth fire databases through the job scheduler; A third step of receiving, by the client, a search result from the first to nth fires through the job scheduler and receiving a search result; A fourth step in which the client requests a similar document search from the first to nth fires through the job scheduler and receives a search result; A fifth step in which the client requests and receives a search result list from the first to nth fires through the job scheduler; And And a sixth step in which the client receives and receives the original text of the document from the first to nth fires through the job scheduler. The method of claim 11, wherein the first to n-th fire in the first step, And retrieving database information and transmitting the database information to the client at the request of the client received through the job scheduler. The method of claim 11, wherein the first to n-th fire in the second step, And retrieving a section list of a database at the request of the client received through the job scheduler and transmitting the retrieved section list to the client. The method of claim 11, wherein in the third step, the client, And retrieving search information including a query, a search target database list, and a section through the job scheduler through the job scheduler. The method of claim 14, wherein the third step, Performing a search by the search information received from the client in the first to nth fires, and then transmitting a search result to the set manager; Storing the search results received from the first to nth fires in the set manager, and transmitting a result set number and the number of documents to the first to nth fires; And And transmitting the result set number and the number of documents received from the set manager to the client in the first to nth fires. The method of claim 11, wherein in the fourth step, the client, And transmitting similar document search information including a search document number, a search target database list, a section, and a search method through the job scheduler through the job scheduler. The method of claim 16, wherein the fourth step, Transmitting a search result to the set manager after performing a search based on the similar document search information received from the client in the first to nth fires; Storing the search results received from the first to nth fires in the set manager, and transmitting a result set number and the number of documents to the first to nth fires; And And transmitting the result set number and the number of documents received from the set manager to the client in the first to nth fires. The method of claim 11, wherein in the fifth step, the client, And transmitting the search result list information including the result set number, a section to be displayed, and the number of search result lists to be received when searching the first to nth fires through the job scheduler. The method of claim 18, wherein the fifth step, Transmitting the search result list information received from the client in the first to nth fires to the set manager; Transmitting as many documents as the number of search result lists included in the search result list information received from the first to nth fires to the first to nth fires in the set manager; And Searching for section information requested by the client with the document received from the set manager in the first to nth fires, and transmitting a search result to the client as the document; How to manage. The method of claim 11, wherein in the sixth step, the client, And transmitting the document number to the first to nth fires through the job scheduler. The method of claim 20, wherein the sixth step is And the first to nth fires transmit the document contents of the document corresponding to the document number received from the client to the client. An information retrieval management method of an information retrieval management system comprising the job scheduler of claim 10, first to nth fires, set manager, and data manager, Transmitting, by the client, a document and database information to be inserted or changed to the job scheduler; Transmitting, by the job scheduler, information transmitted by the client to the data manager; Transmitting a result after storing or changing a document in the data manager to the job scheduler; Instructing a set manager to update the result set if the insert or change is successful in the job scheduler; Transmitting a command execution result to the job scheduler in the set manager; Sending a final result from the job scheduler to the first to nth fires; And And a method of inserting and modifying a document, the method comprising reopening a database and transmitting a result to the client in the first to nth fires. The method of claim 22, wherein the information retrieval management method, Transmitting an ID of a document to be deleted from the client to the job scheduler; Transmitting, by the job scheduler, an ID of a document to be deleted received from the client to the data manager; Transmitting a result to the job scheduler after deleting a document in the data manager; Instructing a set manager to update the result set if the deletion is successful in the job scheduler; Transmitting a command execution result to the job scheduler in the set manager; Sending a final result from the job scheduler to the first to nth fires; And And re-opening the database in the first to nth fires and deleting a single document comprising transmitting a result to the client.