JP3573471B2 - Full-text database search method - Google Patents

Description

[0001]
[Industrial applications]
The present invention relates to a full-text database search method capable of searching for all words included in a sentence without the need for manual creation of keywords in advance, and particularly to a method of searching document data having a logical structure. A database search method for storing and searching as a database.
[0002]
[Prior art]
In the conventional document search method, a full-text database search method for searching the entire text of a document is used as a method of solving a problem of hit omission due to a poor man-hour required for keyword creation and poor keyword attachment. . However, in this full-text database search method, all documents become a search range, and a useless hit occurs. As a means for solving this problem, a method has been used in which a search request is made by designating the logical structure of a document to increase the search efficiency. As a search method for performing such a search, there is a method in which a document is decomposed and stored by some method to perform a search process. As a known example of such a search method, for example, Japanese Patent Application Laid-Open Nos. H01-189721 and H01-189722 disclose documents arranged in a plurality of preset axes and store the documents. There is a method to specify the search by the axis. However, this method cannot perform a search process specifying a document structure. In Japanese Patent Application Laid-Open No. 03-62169, a document is separated into image, graphic, and text information, and the data structure is automatically extracted and decomposed for each medium, stored as a multimedia file, and searched. .
[0003]
[Problems to be solved by the invention]
According to the above-mentioned prior art, it is possible to execute a search request for an image or a figure in order to perform decomposition of a document for each medium such as an image, a figure, or a text and to store the information as multimedia information in a database. On the other hand, it is necessary to develop a dedicated multimedia database management system.
[0004]
An object of the present invention is to decompose a document having a logical structure into its constituent elements and store it in a general-purpose relational database so that a document structure and a search system recognized by a database user can be realized. By displaying the search request including the document structure information in a description language for relational database search by matching the recognized data structures and displaying the search result by specifying the search request statement including the document structure information Is to provide a full-text database search method that enables
[0005]
[Means for Solving the Problems]
According to the present invention, a document is decomposed in component units of a document, and structural information indicating a logical structure of the document, a component identifier for identifying the component, for example, a tag, and text information following the component identifier A database creation step for creating a database including the following, and a search request statement generation step for converting a search request sentence including the input component element identifier into a language description for database search. Searching the database based on the search request sentence generated by the search request sentence generating step, and a search result organizing step of converting the obtained search result into a component element identifier and a format of text information subsequent thereto, And a display step of displaying the search result.
[0006]
[Action]
In a full-text database search process for a sentence having a logical structure, an input original document is divided in a database creation step, and structural information of the document, a component identifier for identifying the component, and A database including a text following the component identifier is created, and in the search request sentence generation step, the search request sentence specified at the time of the search processing is associated with the logical structure of the document and the database structure. Based on the relations, it is converted into a language description for database search, and in the search result organizing step, the document data obtained as a result of the search processing is sorted and combined based on its ID and sequence number. In the display step, the search result is displayed on a display device, whereby a search request sentence including the document structure information can be specified and the search result can be displayed. .
[0007]
【Example】
Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
[0008]
FIG. 1 is a system configuration diagram of an embodiment of the present invention. In this embodiment, the original document is described in SGML, and SQL is used for an interface with the database management system. SGML (Standard Generalized Markup Language) is a standard defined by ISO (International Organization for Standardization), and is a general-purpose typesetting instruction language that makes it possible to structure a document. SQL (Structured Query Language) is a standard processing language for relational databases. Tags are used to represent the structure of a document using SGML. Tags have a start tag and an end tag. By enclosing the components of the document structure information with start tags and end tags, it is possible to separate the text in the document and to which component in the document structure the text belongs. To clarify. Here, the start tag is obtained by closing the element name with symbols “<” and “>”, and the end tag is obtained by closing the element name with “<//” and “>”. The content of the component following the tag is text. A document including structural information to be searched is composed of a tag and a text.
[0009]
A system configuration for realizing the present invention will be described based on FIG. The present invention comprises an SGML server 1 and a general-purpose relational database management system 2, and the SGML server 1 further comprises a database creating unit 3, an SQL generating unit 4, and an SGML recreating unit 5. The database creation unit 3 inputs an original document described in SGML11, which is one of the languages having document structure information therein, and based on the input SGML, the relational database management system 2 generates a database database. By issuing a DDL (Data Definition Language) 12 which is a language for defining a table, a table of structure information 7, tag information 8, and text information 9 by SGML is created, and original document data using SGML is created. Is stored. The SQL generation unit 4 inputs a search request 16 created in a search language DQL (Document Query Language) from which the document structure can be specified from the workstation 10, creates a search request sentence using the SQL 14, and creates a relational database management system 2. Issues a search request to. The SGML re-creating unit 5 includes a tag using the structure information 7 with respect to the search result 15 generated by the SQL generation unit 4 in the relational database management system 2 and searched by the search request sentence using the SQL 14. The SGML 17 is organized and displayed on the workstation 10.
[0010]
FIG. 2A is a model of a logical structure of a document represented by the SGML 11 of FIG. 1, and a general journal 101 is composed of eight components from a cover 102 to a paragraph 109. FIG. 2B is an example of a diagram showing properties of components of a document expressed in SGML.
[0011]
FIG. 3 is an example of a document represented by SGML having a logical structure, and is an example in which the logical structure of the text 211 to 214 is expressed by using the tag 201 to 208.
[0012]
FIGS. 4 to 6 show an example of each table data format when the document expressed in SGML of FIGS. 2 and 3 is stored in the relational database. The data format of the structure information table shown in FIG. 4 is a tabulation of the models shown in FIGS. 2A and 2B. The order defined between the components in the same hierarchy with the ID 303 is represented by an order number 304 according to the order of the definition, and whether or not a tag appears is represented by a connection type 305.
[0013]
The data format of the tag information table of FIG. 5 is created for each tag hierarchy of FIG. 4, and an ID is added for each SGML tag of FIG. Stored with For a tag that may appear multiple times for the same type of tag in the same document, the author number in the author tag table 504 is set. NO.423. Is added.
[0014]
The data format of the text information table shown in FIG. 6 is obtained by converting the text decomposed from the original document described in SGML into its tag ID and text number, which is a series of serial numbers in the document, for each element. Is added and stored.
[0015]
The method of assigning IDs in FIGS. 4 to 6 is an example, and any ID may be used as long as it is unique.
[0016]
FIG. 7 shows the flow of a database search process according to the present invention, and FIG. 8 shows an example of execution of the database search process according to the present invention. The search request input from the workstation 10 in accordance with the DQL is processed by the SQL generator 4. For example, a search request for “cover of a general magazine including“ print ”in a chapter” is input from the workstation 10 as DQL 701. The document structure information shown in FIG. 2 is called from the FROM clause 712 of the DQL 701 (601). The “chapter” of the WHERE clause 713 of the DQL 701 has a chapter 106, a title 107, a section 108, and a paragraph 109 as constituent elements. It can be seen that the title 107, the section 108, and the paragraph 109 are included. 6, "LIKE '% print%" of the WHERE clause 713 is developed for each table of the title text 513, section text 514, and paragraph text 515 in FIG. 6 (602), and the title ID 417 and the section ID 418 are obtained. , A paragraph ID 419, and a chapter ID which is an element of a higher-order structure whose respective IDs match in the title tag 507, the section tag 508, and the paragraph tag 509 in FIG. Until the chapter ID of the component specified by the WHERE clause 713 is obtained, the table shown in FIG. 5 is searched, and a search request sentence using SQL combining all of them is created. From the document structure information of FIG. 2, it is determined that there is no hierarchical relationship between the elements of “cover” and “chapter” specified by the SELECT clause 711 and the WHERE clause 713 (603), and from the document structure information of FIG. The general magazine 101 existing as a common upper element at the lowest level of the cover 102 and the chapter 106 is determined as the element ID to be finally obtained (605), and the chapter ID matches in the table of the chapter tag 506 in FIG. Until the general journal ID is obtained, the table of FIG. 5 is repeatedly expanded so as to obtain the text ID 415 (607), and the final result is stored in the temporary table. The designated search request is converted into a search request sentence described in SQL, a search request sentence 702 written in SQL is completed, and a search is executed. The element “cover” specified in the SELECT clause 711 is found from the sentence structure information in FIG. 2 to reach a structure having a text in two paths, namely, cover 102-title 103 and cover 102-author 104. SQL731 and SQL732 that can obtain the IDs of all elements existing under the general journal ID 101 of the temporary table, the sequence number of the text, and the contents obtained by the above are created and executed by the respective paths, and the search result 741 is obtained. And a search result 742 is obtained (608). Based on the fact that the cover IDs of the search result 741 and the search result 742 match and the sequence number, it is analyzed that they belong to the same cover, and are arranged in the order of the title “DQL that is easy to understand” and the author “Taro Hitachi” ( 609). In processing 610, a tag is added to the text analyzed in processing 609, and a search result sentence 705 according to SGML is organized (610).
[0017]
【The invention's effect】
According to the present invention, the following effects can be obtained.
[0018]
In a process of searching for a document having a logical structure, a document is decomposed into its constituent elements, stored in a database, searched, and the search result is converted into a form of a constituent element identifier and text information. By displaying the search request, it is possible to specify a search request sentence including logical structure information using a specific language and to display a search result including structure information.
[Brief description of the drawings]
FIG. 1 is a diagram showing a system configuration according to an embodiment of the present invention.
FIG. 2 is a diagram illustrating a model of a logical document structure and a diagram illustrating an example of a document type that is logical structure information.
FIG. 3 is a diagram illustrating an example of an SGML document.
FIG. 4 is an example of a structural information table stored in a relational database management system.
FIG. 5 is a diagram showing an example of a structural information symbol table stored in a relational database management system.
FIG. 6 is a diagram showing an example of a text information table stored in a relational database management system.
FIG. 7 is a diagram illustrating an example of a flow of a search process according to the present invention.
FIG. 8 is a diagram illustrating an example of execution of a search process according to the present invention.
[Explanation of symbols]
Reference Signs List 1 SGML server 2 Relational database management system 3 Database creation unit 4 Search request sentence generation unit 5 Search result organization unit 6 Original document 7 Data base of structure information 8 Data base of tag information 9 Data of text information Tabe 10 Search workstation

Claims (1)

In a method of searching the full text of a document having a logical structure,
Creating a relational database by inputting an original document described in a markup language having therein the logical structure information of the document components, the hierarchical relationship between document components in the language and the order of appearance Defining a table for storing logical structure information of a document such as a document, a table for storing component identifiers for identifying the components, and a table for storing text information following the component identifiers, A database creation step of storing original document data in each table using
A search request inputting step of inputting a first search request sentence which is created in a search language capable of designating a component of a document having a logical structure and has an extraction target designation phrase, a search range designation phrase, and a search condition designation phrase; ,
The logical structure information of the document is called by the component specified by the search range specification phrase of the first search request sentence and referred to, and the component specified by the search condition specification phrase of the first search request sentence Generating an SQL statement for searching, from the relational database, a component identifier that satisfies the search condition described in the search condition specification phrase, for all the components having texts that are subordinate components of
Referring to the logical structure information of the called document, a common upper or same rank of the component specified by the extraction target specification phrase and the configuration component specified by the search condition specification phrase of the first search request sentence Determine the lowest component among the components of
A component identifier of the determined component, which has a component identifier of a component that satisfies the search condition described in the search condition specifying phrase at a lower level, is obtained. Generating a new SQL statement by expanding the SQL statement to be stored in a temporary table of a relational database;
A component identifier extracting step of searching the relational beta base using the new SQL statement and storing the component identifier searched in a temporary table of the relational database;
The logical component information of the document is referred to by the component corresponding to the component identifier extracted in the component identifier extracting step, and the component specified in the extraction target specification phrase of the first search request sentence is subordinate to the component. A second component that determines a component having a text and searches the relational database for a component identifier of the document that is the component identifier of the determined component and that is the same as the extracted component identifier. A search request sentence generation step;
A search step of searching the relational database based on the second search request statement ;
A search result organizing step of converting a search result obtained by the search step into a form of a component of a document expressed by the markup language ;
Displaying a search result organized by the search result organizing step.

Images