JP2000250930A - Structured document retrieval system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To retrieve all structured documents by designating various structural conditions. SOLUTION: A document retrieval system which deals with the structured documents includes a structure analysis means 106 which analyzes the logical structure of a structured document that is registered, a means 107 for producing structural information holding document data, which produces the structural information holding document data including the information on the hierarchical structure of the document for each of elements that are classified into the logical structures by the means 106, an index information production means 108 which produces the index information for execution of retrieval from the structural information holding document data, a retrieval condition analysis means 109 which analyzes an inputted retrieval condition and converts it into another retrieval condition that is suitable to the retrieval processing and an index information retrieval means 110 which executes retrieval using the index information according to the retrieval condition that is converted by the means 109. Thus, it is made possible to retrieve the structured documents by designating various structural conditions.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document management system for managing a structured document having a logical structure, such as a dissertation or a manual, by a computer. This enables a simple document search.

[0002]

2. Description of the Related Art With the increase in electronic documents, manuals,
Interest in registering and retrieving structured documents having a logical structure such as minutes, specifications, etc. has been increasing. The logical structure of a structured document is defined by DTD (Document Type Definition). Conventionally, a document management system for registering and retrieving structured documents is disclosed in
What is described in -240752 is known. As shown in FIG. 37, this system includes a document structure analysis program 3701 for analyzing a document structure and an analyzed document data storage area 3705 for storing the analyzed document data.
, A structure index creation program 3702 for creating a structure index for performing a search by a document structure, a structure index storage area 3706 for storing the created structure index, and data (structured data) corresponding to the structure index of the document to be registered. Creates a structured full-text data generation program 3703 that generates full-text data), a structured full-text data storage area 3707 that stores this data, and a character string index for performing full-text search from the structured full-text data of each document to be registered String index creation program 37
04 and a character string index storage area 3708 for storing the generated character string index.

In this system, when a structured document is registered, first, a logical structure of a document to be registered is analyzed by a document structure analysis program 3701 to generate analyzed document data, and the analyzed document data is stored in an analyzed document data storage area 3705. register.

Next, a structure index creation program 37
02, the logical structure of each document to be registered is superimposed sequentially in the order of registration, and elements having the same appearance position and type in the document are represented by a single meta element. A certain character string data group is represented by a single meta character string data, so that a meta element group and a meta character string data group (in this system, these are collectively referred to as a metanode) have a tree structure. An index is generated, and an identifier (in this system, this is referred to as a context identifier) for uniquely identifying the meta-nodes constituting the structural index is given to the meta-nodes. Register with.

FIG. 38 is a diagram showing a process of creating the structure index. In FIG. 38, a document 1, a document 2, and a document 3 each represent analyzed document data of a document to be registered. The structure index is formed by sequentially superimposing the structure of the analyzed document data on the existing structure index. First, when document 1 is input, since the structure index is initially in the initial state (empty), a tree structure equivalent to the analyzed data is generated and registered as it is in the structure index. The state shown in 3801 is obtained. The newly generated meta element has context identifiers from E1 to E5, and the newly generated meta character string data has C
Context identifiers from 1 to C3 are assigned.

Next, when the document 2 is input, nothing is performed on the portion where the structure overlaps with the existing structure index (3801), and the partial structure (the network shown in FIG. Only the hung part) is newly registered. The context identifiers E6 and E7 are assigned to the newly generated meta element, and the context identifier C4 is assigned to the newly generated meta character string data.

Next, when the document 3 is input, nothing is performed on the portion where the structure overlaps with the existing structure index (3802), and the partial structure (the network shown in FIG. Only the hung part) is newly registered. Context identifiers E8, E9 and E
10. Context identifiers C5 and C6 are assigned to the newly generated meta-character string data.

[0008] At the stage when three documents are registered in this way, the structure index is in the state shown in 3803.

Next, a structured full-text data generation program 37
According to 03, for each document to be registered, all the character strings included in the parsed document data corresponding to the document,
Generates data (in this system, this is called structured full-text data) composed of the definition of the correspondence relationship with the context identifier indicating the character string in the structure index, and registers it in the structured full-text data storage area 3707 .

Next, a character string index creation program
According to 3704, a character string index for performing a full-text search including a context identifier is created from the structured full-text data corresponding to each registration target document, and registered in the character string index storage area 3708.

In the search in this system, first, a set of context identifiers satisfying a specified structure condition is determined by referring to a structure index.

Next, a document group that satisfies the designated condition is obtained by searching a character string using the context identifier as a key.

[0013]

However, there are a plurality of different document types in a structured document. However, in the related art, registration of documents corresponding to all such general-purpose document types and document There was a problem that a search conscious of the type could not be performed.

Further, according to the method of the prior art, in the case of a search condition such as “document including“ XX ”in chapter and below”,
Since it is necessary to obtain a set of all applicable identifiers by referring to the structure index and perform an OR search based on the identifiers, the search is slow.

The present invention solves the above-mentioned problems of the prior art. It is possible to perform a search by specifying various structural conditions for all structured documents. It is an object of the present invention to provide a structured document search system capable of searching structured documents.

[0016]

Therefore, in the structured document search system of the present invention, a structure analyzing unit for analyzing a logical structure of a registered document when registering the structured document, and data divided for each logical structure. A structure information holding document data creating unit for adding a character string obtained by connecting a tag name, its appearance order, and a delimiter as information of the hierarchical structure to the document, and an index for performing a high-speed search from the structure information holding document data And an index information creating unit for creating information.

With this configuration, when performing a search,
For example, "document whose title contains the character string" structured "", "document whose chapter 1 contains the character string of" registration "", and "document whose chapter title contains the character string of" method "" It is possible to search for a document by designating various structural conditions of the document.

Further, in the structured document search system of the present invention, when registering a structured document, not only the structure character string but also the document type of the registration target document is used as the information on the logical structure of the structured document. Register as a character string.

Thus, for example, if the document type is [paper]
Thus, a search specifying a document type can be performed, for example, "a document in which the title of the first chapter includes a character string" registered "".

[0020]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS According to the first aspect of the present invention, in a document search system for handling a structured document, a structure analyzing means for analyzing a logical structure of a structured document to be registered, and the structure analyzing means Means for creating structure information holding document data including information of the hierarchical structure for each element divided into logical structures by the above, and an index for performing a search from the created structure information holding document data Index information creation means for creating information, search condition analysis means for analyzing input search conditions and converting the search conditions into search conditions suitable for search processing,
An index information search means for performing a search using the index information is provided, and a structured document can be searched by specifying various structural conditions.

According to a second aspect of the present invention, there is provided a result list displaying means for displaying a list of search results searched by the index information searching means, and an entity for displaying entity data of a document selected from the search result list. Display means for selecting a desired document from a plurality of documents retrieved based on the structural conditions and displaying the actual data.

According to a third aspect of the present invention, the structure information holding document data creating means has a structure in which the logical structure information of each element of the document to be registered is formed by a character string in which a tag name, its appearance order and a delimiter are connected. This is included in the information holding document data. When performing a search, it is possible to perform a search by designating various structural conditions.

According to a fourth aspect of the present invention, the structure information holding document data creating means includes a character string indicating the document type information of the registration target document in the structure information holding document data. In this case, a search with a specified document type can be performed.

According to a fifth aspect of the present invention, a unique identifier is assigned to the document type and tag name of the document to be registered before passing the structurally analyzed data analyzed by the structural analysis means to the structure information holding document data creating means. The document type to which the
A structure table creating means is provided.For example, for documents having different document types but the same logical structure, such as an original language document and its Japanese translation document, the same value is used for the identifier of the document type and the tag name. The assignment makes it possible to extract those documents with one search condition expression.

According to a sixth aspect of the present invention, the index information creating means creates index information including an identifier uniquely assigned to each of the document type and the tag name. In response to a structure-designated search request that specifies a name, efficient and high-speed search can be performed.

According to the present invention, the document type / structure table creating means holds information on a plurality of appearing tag names, and the structure information holding document data creating means stores only the information on the tag name appearing a plurality of times. Since the information on the appearance order is included in the structure information holding document data, and the information on the appearance order can be excluded from the index information, the retrieval speed is increased.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. The present invention is not limited to these embodiments at all, and can be implemented in various modes without departing from the scope of the invention.

(First Embodiment) As shown in FIG. 1, the structured document retrieval system according to the first embodiment displays the progress of document registration, and specifies the progress of document registration. Display means 101 for displaying search conditions and search results;
A recording device 102 for storing documents to be registered, a search engine 104 for actually registering, searching for, and displaying search results of documents to be searched, and an input for inputting various commands and conditions at the time of document registration and document search It comprises means 103 and a search database 105 for storing various data created by the search engine 104.

The configuration of the search engine 104 is roughly divided into registration means, search means, and display means. First, the registration unit includes a structure analysis unit 106 that analyzes a logical structure of a document to be registered.
And a structure information holding document data creation unit 10 for adding structure information to each data divided for each logical structure by the structure analysis unit 106 to create structure information holding document data.
7 and an index information creation unit 108 for creating index information for performing a high-speed search from the structure information holding document data created by the structure information holding document data creation unit 107. Details of these 106 to 108 will be described in the description of the flow of the document registration process.

Next, the search means includes a search condition analysis unit 109 for converting the search condition received from the input unit 103 into a condition suitable for performing a search by the search device, and a search condition analysis unit 1.
An index information search unit 110 that receives the search condition analyzed in 09 and actually performs a search process using the index information is provided. Details of these 109 and 110 will be described in the description of the flow of document search.

Next, the display means includes a result list display section 111 for displaying a search result list on the display means 101, and an entity display for displaying entity data of the document selected by the input means 103 from the search result list on the display means 101. And a unit 112.

The search database 105 includes a structure-analyzed data storage unit 113 for storing the structure-analyzed data created by the structure analysis unit 106, and structure information for storing the data created by the structure information holding document data creation unit 107. Holding document data storage unit 114 and index information storage unit 115 for storing the index information created by index information creation unit 108
And the entity data storage unit that stores the entity data of the registered document
It includes a list data storage unit 117 for storing bibliographic data for a search result list, and stores data used for structured document search and result display.

Next, the document registration process of the structured document search system will be described using a specific example of a structured document.

First, in accordance with a registration request from the input unit 103, a document to be registered is read from the storage unit 102. next,
The structure analysis unit 106 converts the structure of the registration target document into a form that can be understood by the structure information holding document data creation unit 107. This structure analysis unit 106 refers to the DTD of the registration target document,
Analyze its structure. The structure analysis unit 106 converts the structured document as a character sequence into a data structure that can be understood by the structure information holding document data creation unit 107 (hereinafter, referred to as structure analyzed data).

FIG. 2 shows an example of the DTD. This DTD defines the location, order, number of appearances, and the like where the document element should appear. A description of a part of the contents defined in the example of FIG. 2 is shown below. First, the <paper> element has a <bibliography> element and a <body> element as its child elements. This <
The <Bibliography> element is a <Title> element and <Author> as child elements.
Element and a <date> element. The <title>, <author>, and <date> elements have no child elements below them, but have character strings as their contents. The <Body> element has a <Chapter> element as its child element, and this <Chapter> element is followed by a “*” symbol. It means that it may appear.

The structure analysis unit 106 converts a document type structured document defined by the DTD in FIG. 2 into a structured document in the form of FIG. 3, for example. FIG. 4 shows the structured document of FIG. 3 analyzed by the structure analysis unit 106 in a tree structure.

Next, the structure information holding document data creation unit 107
Will be described with reference to FIG. First, the structurally analyzed data of the document to be registered is stored in the structurally analyzed data storage unit.
Read from 113 (step 501). Next, a unique number (hereinafter, referred to as a document number) is assigned to each of the read registration target documents (step 502). Next, a unique number (hereinafter referred to as an element number) within the document is assigned to each component of the document (step 50).
3). Next, the tree structure address of this element is converted into a character string (hereinafter, referred to as the document element name (tag name))
(Referred to as a structural character string) (step 504).

In this case, the tree structure address of the title character string “structured document management method” in the structured document shown in FIG. 3 is “\ article_1 \ bibliography_1 \ title_1 \ "Can be expressed. Here, the appearance order is a number indicating the number of the element having the same parent element and the same tag name, and the appearance order is continued with "_" being interposed between the tag names. In addition, “\” is used as a layer break. This "\" is also added to the beginning and end of the structure string. In this example, “_” and “\” are used, but these “_” and “\” may be any as long as they are not used in the tag name and the order of appearance. Also, these "_" and "\" need not be fixed with one kind of character, and different characters such as "\" for the first delimiter of the hierarchy and "$" for the second delimiter may be assigned. Absent.

Finally, a record is created as a set with these document number, element number, structural character string, and element contents (character string) (step 505). And step 5
Steps 03 to 505 are repeated for all components (step 506). This record is combined into structure information holding document data, and the structure information holding document data storage unit.
It is stored in 114 (step 507).

FIG. 6 shows an example of the structure information holding document data created from the example of the structured document shown in FIG. Sign 60
Numerals 1 to 607 are assigned to distinguish each of the records shown in FIG. 6 and will be used in the description of the search processing using an actual example described later. Here, the number “1” is assigned to the structured document in FIG. In this example, it is assumed that the document registered second is assigned “2” as the document number, but this document number is unique to each registered document data. The numbers need not be assigned in ascending order as described above. Also, in FIG. 6, numbers are assigned to the respective structural elements in the order in which they appear in the document, but the element numbers may be unique numbers in the same document.

Next, the flow of processing of the index information creating unit 108 will be described with reference to FIG. First, the structure information holding document data of the registration target document is read from the structure information holding document data storage unit 114 (step 701). Next, a character chain having a predetermined number of characters is extracted from the content character string therein (step 702). Regarding this character chain, the corresponding document number, element number, and a number indicating the number of the first character of the character chain in the element (hereinafter, referred to as
Information of the character position number) is created as index information on the content character string (hereinafter, referred to as content character string index information) (step 703). This process is repeated for all the content character strings of the registration target document (step 704).
Next, the same processing as steps 702 to 704 is performed.
This process is performed on the structural character string of the registration target document (steps 705, 706, and 707). The index information created here is referred to as structural character string index information. Finally, the content character string index information and the structure character string index information are added to the index information storage unit 115 (step 708).

FIG. 8 shows the contents of FIG.
FIG. 7 is a diagram showing a part of an example of structural character string index information and content character string index information created for 601 records in the structural information holding data of FIG. In FIG. 8, reference numeral 801 denotes that “the document chain having the document number“ 1 ”has the character string“ ￥ theory ”in the structural character string of the element having the element number“ 1 ”from the beginning.
It exists from the position of the character. "
Although FIG. 8 shows only a part of the index information, index information is actually created for all the content character strings and the structural character strings of the registration target document. Accordingly, when creating index information for the structural character string 602 in FIG. 6, a character chain “￥ theory” exists from the position of the “1” character from the head, as in 601; The information [1-2-1] is added to the index information for "".

In this example, character chains are extracted two characters at a time and index information is created for each character chain. However, the character chains need not be two characters at a time.

The above-described registration process is repeated each time a document to be registered is input, whereby index information is added.

Next, the flow of the document search process according to the present embodiment will be described using specific examples of search conditions. First, it is assumed that a condition “a document including a character string“ structured ”in a title” is given as a search condition from the input unit 103.

Here, the processing flow of the search condition analysis unit 109 will be described with reference to FIG. First, in step 901, it is determined whether or not there is a condition specifying a structure in the search condition.
If there is a search condition specifying the structure, step 902
If the structure specification is a search that specifies only the end of the logical structure, the structure condition is set to the first character using a wildcard (a character representing any character string of 0 or more characters). It is created as a character string for backward matching (hereinafter, referred to as a structural condition character string) (step 903). The “single end” will be described later using a specific example. In the case of a search designating only the end other than the end, a structure condition character string for intermediate matching including a wildcard in the first and last characters is created (step 904). Finally, a search condition expression is created using the structure condition character string as the search condition created here and the specified content character string (step 90).
5). If it is determined in step 901 that the search is performed without specifying a structure, the structure condition character string is set to “Null value” (step 906), and a search condition expression is created.

In this example, when the search condition is “a document including a character string“ structured ”in the title”, the search is performed by designating the structure. In this condition, there is a structure specification for “title”, and since it is a search specifying only one end, the structure condition character string is “* \ title_1 \”. Although the order of appearance of the structural element "title" is not specified, since this element appears only once in the document,
The appearance order is “1”, and a structure condition character string is created. Here, “*” is a wild card. The search condition expression created as a result is “structure character string =“ * \ title_
1 \ ”& Content string =“ * Structured * ””.

Here, "search specifying only the end of the logical structure" means a search in which one of elements having character data below is specified in a tree structure as shown in FIG. Say Specifically, the elements shaded in FIG. 4 are worth the “end”.

On the other hand, when the search condition specifies a logical structure element or less, such as "document including a character string" register "in Chapter 1", the structure condition character string is "*". \ Chapter_01 \ * ", and a structure condition character string for performing an intermediate match is created. Also, even if the search condition is specified as the end of the logical structure such as "document whose title includes the character string" method "", the applicable element is not limited to one (it is not unique). Perform an intermediate match. In the case of this search condition example, since any chapter title may be used, the structure condition character string is “* \ chapter title *”.

Next, the processing flow of the index information search unit 110 will be described with reference to FIG. First, in step 1001, index information is read from the index information storage unit 115. Next, in step 1002, the designated character string is searched using the content character string index information. As a search result, a set of a document number and an element number in which the character string exists is obtained by the number of cases. Next, in Step 1003, if the result of Step 1002 is 0, "No applicable document"
And ends. If it is not 0, the process proceeds to step 1004. If the structure condition character string is “Null”, the result of step 1002 is set as the final result (step 1004).
7) End. If the structure condition character string is not "Null", a search is performed for the specified structure condition character string using the structure character string index information (step 1005). Here, a set of a document number and an element number is obtained as a search result. Finally, in step 1006, of the results of step 1002 and step 1005, the combination of the document number and element number that exist in both ends as the final result.

Here, the details of the search using the index information performed in steps 1002 and 1005 will be described. As an example, consider a case where a search is performed for a character string “structured”. As for the character string “structured”, “structure” and “structured” can be extracted as a character chain of two characters at a time. Here, the number of characters in the character chain to be extracted is the same as the number of characters in the character chain in the index information. Assuming that index information such as that shown at 1110 in FIG. 11 has been created for these two character chains, the document numbers and element numbers are the same, and the "structure"
Are extracted as search results. In the example of FIG. 11, it is possible to extract 1121, 1122, and 1123 assuming that the document number and the element number are the same. Among them, the character position numbers are consecutive at 1121 and 1123, and a set of these document numbers and element numbers is used as a search result.

The results of the search for the content character string and the structure character string by the above-described method are compared, and the combination of the document number and the element number in both search results is obtained as the final search result.

The bibliographic information (title, author, date, etc.) of the resulting document is stored in the list data storage unit 117 as data for displaying a result list.

Finally, the search result display processing will be described. First, the result list display unit 111 reads the search result list data from the list data storage unit 117 and displays it on the display unit 101. Next, when any one document is selected from the search result list as the entity display request from the input unit 103,
The entity display unit 112 reads the entity data of the corresponding document from the entity data storage unit 116 and displays it on the display unit 101.

As described above, according to the present embodiment, when a search is performed by registering information on the logical structure of a structured document using a character string obtained by connecting the tag name and its appearance order, Searches with various structural conditions specified can be performed.

(Second Embodiment) In a second embodiment of the present invention, a structured document retrieval system capable of performing retrieval by specifying a document type will be described.

The configuration diagram of this structured document search system is not different from that of the first embodiment shown in FIG. However, processing including document type information is added to the processing performed by the structure information holding document data generating unit 107, and as a result, the index information generating unit
The processing of the search condition analysis unit 109 and the index information search unit 110 is different from that of the first embodiment.

Here, the flow of the registration process according to the second embodiment will be described. First, the case of the structured document example shown in FIG. 3 as the first document to be registered will be described. The structured document in FIG. 3 has a “paper” type DTD shown in FIG. The processing of the structure analyzing unit 106 is the same as that of the first embodiment, and thus the description is omitted. Next, the processing of the structure information holding document data creation unit 107 is shown in FIG. In the second embodiment, step 505 in FIG. 5, which is the processing of the structure information holding document data creation unit 107 in the first embodiment, is changed to step 1201, and the record for each document component is a document type character. Created to include columns.

FIG. 13 shows an example of structure information holding document data created from the example of the structured document of FIG. 3 in the second embodiment. Compared to FIG. 6 which is an example of the first embodiment, it can be seen that a document type character string is added. In the example of FIG. 3, a document type declaration such as <! DOCTYPE paper SYSTEM “paper.dtd”> is described on the first line, and this “paper” is used as a document character string. Next, the index information creating unit 108 according to the second embodiment creates index information in the same manner as in the first embodiment, but the first embodiment provides a total number of structured character strings and content character strings. While two types of index information are created, in the second embodiment, index information on document-type character strings is also created in addition to the two types, and this is added to the index information storage unit 115.

Next, a case will be described in which an example of a structured document as shown in FIG. 15 having a “recipe” type DTD shown in FIG. 14 different from “thesis” is registered as the second document to be registered. .

The processing of the structure analyzing unit 106 is the same as that of the first embodiment, and the description is omitted.

FIG. 16 shows the structured document of FIG.
It is a tree structure obtained as a result of analysis by 106.

Next, the structure information holding document data creation unit 107
FIG. 17 shows the result of performing the processing shown in FIG. 12 in the same manner as in the first registration target document. As shown in FIG. 17, a document number different from the first registration target document is assigned to this registration target document. The document type character string is a document type “recipe”.

Next, the index information creating unit 108 creates index information on the structural character string, the content character string, and the document type character string in the same manner as in the case of the first document to be registered. Added to 115. As document type character string index information, the document number of a document having the document type is stored in association with the character string representing the document type.

Every time a new document to be registered is input,
The above registration processing is repeated.

Next, the flow of a document search process according to the second embodiment will be described with reference to specific search conditions. It is assumed that “a document whose document type is [thesis] and the title of Chapter 1 includes a character string“ registered ”” is given as a search condition from the input unit 103. In the second embodiment,
Because the registration is performed with the document type in mind, specifically, the registration of the document type string for each component and the index information for this document type string is performed. Search becomes possible.

FIG. 18 shows the flow of processing of the search condition analysis unit 109 according to the second embodiment. Second
In the second embodiment, it is determined whether or not the search is performed by specifying a document type before performing the process of the search condition analysis unit 109 in the first embodiment (step 1801). The document type is a document type condition string,
If not specified, the document type condition character string is set to Null in step 1803, and the flow advances to the next step 901. Steps 901 and subsequent steps are the same as the processing in the first embodiment, and a description thereof will not be repeated.

An example of the search condition here is “a document whose document type is [article] and the chapter title of Chapter 1 includes the character string“ registration ””, and is created as a result of the above processing. The search condition expression is “document type character string =“ paper ”& structure character string =“ * \ chapter_1 \ chapter_1_1 \ ”& content character string =“ * registered * ””.

Next, the flow of processing of the index information search unit 110 in the second embodiment will be described with reference to FIG. First, a step 100 for searching for a content character string
Reference numerals 1, 1002, and 1003 are the same as those in the first embodiment, and a description thereof will not be repeated. Next, if the result of searching for the content character string is not 0, and if the document type character string of the search condition expression is Null, step 100
The process proceeds to step S7, and the result of step 1002 is ended as the final search result. If the document type character string of the search condition expression is not Null, in step 1902, a search is performed for the document type character string using the index information. Next, in step 1004, when the structural character string of the search condition expression is Null, in step 1903, step 1002
Then, a set of a document number and an element number is extracted as a final result based on the result of step 1902, and the processing is terminated. If the structure character string of the search condition expression is not Null, step 1005
Then, a search is performed for the structured character string using the index information. Finally, in step 1904, a set of a document number and an element number is extracted as a final result based on the results of steps 1002, 1902, and 1005, and the process ends.

The search result display processing is the same as in the first embodiment, and a description thereof will not be repeated.

As described above, in the present embodiment, not only the structure character string but also the logical
By registering the document type of the registration target document as a document type character string, it is possible to register structured documents with any document type, and it is possible to search by specifying the document type when performing search Become.

(Third Embodiment) In a third embodiment of the present invention, a structured document retrieval system for setting a document type by a document type identifier will be described.

In this system, as shown in FIG. 20, a document type table in which a document type is associated with a document type identifier, a document type identifier and an element of a structured document are stored in the registration means of the search engine 104. A document type / structure table creating unit 2001 for creating the associated structure table is newly provided, and a document type / structure table storage unit 2002 for storing the created document type table and the structure table in the search database 105. Is newly provided. Other configurations are the same as those of the first embodiment (FIG. 1).

Here, the flow of the registration process in the third embodiment will be described. First, the processing of the structure analyzing unit 106 is the same as that of the first and second embodiments, and thus the description is omitted.

FIG. 21 shows a document type / structure table creation unit 20.
It shows the flow of processing in 01. Here, the case of the example of the structured document in FIG. 3 having the DTD shown in FIG. 2 as a registration target document will be described. First, in step 2101, the DTD (FIG. 2) of the registration target document is read. Next, in step 2102, it is determined whether the document type defined by this DTD is registered in the document type table.
The document type table is a table including a document type and an identifier uniquely assigned to the document type (this is called a document identifier here). If it has been registered in the document type table, the process ends. In this example, it is assumed that the registration target document shown in FIG. 3 is the first document to be registered. At this point, the document type table is empty.
, And adds it to a table for managing the correspondence between these document types and document type identifiers (hereinafter, referred to as a document type table) (step 2104).

Next, steps 2105, 2106, 2
Repeat 107. First, in step 2105, a tag name is read, and in step 2106, a unique identifier (hereinafter, referred to as a tag identifier) is assigned to the tag name.
At 7, a tag name, a tag identifier, and a document type identifier are added to a table (hereinafter, referred to as a structure table) for managing the correspondence between them.

FIG. 22 shows an example of a document type table and a structure table created when the DTD of FIG. 2 is read. In the example of FIG. 22, the document type identifier is “1”.
However, the tag identifiers are assigned in ascending order from "1", but these may not be assigned in ascending order or may not be numbers as long as they can uniquely identify the document type and the tag name. .

Next, as another document to be registered, (tentatively,
Assume that a request for registration of a structured document as shown in FIG. 24 having the DTD shown in FIG. 23 (as the tenth document to be registered) is received. The DTD shown in FIG. 23 has the same logical structure as the DTD of FIG. 2, although the element name (tag name) is different between Japanese and English. When registering such a document having a DTD that defines the same structure as the document type already registered in the document type table and the structure table, the document type / structure table creation unit 2001
Then, the same value as the identifier assigned to the corresponding registered document type and tag name is assigned to the document type identifier and tag identifier for the DTD.

FIG. 25 shows the DTD shown in FIG.
This is the result of the document type table and the structure table at the time when the process of the document type / structure table creation unit 2001 is performed.

The flow of processing in the next structure information holding document data creation unit 107 is almost the same as that of the second embodiment, except that a tag identifier is used instead of a tag name when creating a document type character string. . Also, instead of creating a document type character string field for the document type, a document type identifier field is created.

FIG. 26 shows a third embodiment of the present invention.
24 is structure information holding document data obtained as a result of performing the processing of the structure information holding document data creation unit 107 when the structured document of FIG. 24 is registered. Reference numerals 2601 to 2614 are assigned to distinguish each of the records shown in FIG. 26, and will be used in the description of search processing using an actual example described later.

Next, unlike the second embodiment, the index information creating unit 108 according to the third embodiment does not create index information for a document type. This is because, in the second embodiment, a document type character string is held as document type information, but in the third embodiment, the document type information has a document type identifier as a numerical value.

Next, the flow of a document search process according to the third embodiment will be described. As an example of a specific search condition, it is assumed that “a document whose document type is [thesis] and the chapter title of Chapter 1 includes a character string“ method ”” is given.

FIG. 27 shows the flow of processing of the search condition analysis unit 109 according to the third embodiment. In the third embodiment, as in the second embodiment, it is determined whether or not the search is performed by specifying the document type. However, in the case of the search in which the document type is specified, the document type table is read (step 270).
1) Extract the specified document type identifier instead of the document type character string (step 2702). After that, the second embodiment is similar to the second embodiment in that a process of creating a backward matching structural condition character string or an intermediate matching character string is performed according to the presence or absence of the structural condition designation. However, the third embodiment differs from the second embodiment in that the structure table is read (step 2703), and the structure condition character string is created with reference to the structure table.

An example of the search condition here is “a document whose document type is [article] and the chapter title of Chapter 1 includes a character string“ method ””. The conditional expression is “document type identifier = 1 & structure character string =“ * \
7_1 \ 8_1 \ "& content character string =" * registered * ".

Next, the flow of processing of the index information search unit 110 in the third embodiment will be described with reference to FIG. First, a step 100 for searching for a content character string
Reference numerals 1, 1002, and 1003 are the same as those in the first and second embodiments, and a description thereof will not be repeated. Next, when the result of the search for the content character string is not 0, and when the document type identifier of the search condition expression is Null, the process proceeds to step 1007, and the result of step 1002 is ended as the final search result. I do. The document type identifier of the search condition expression is
If it is not Null and the structure character string of the search condition expression is Null in step 1004, the process proceeds to step 2802, and the document type identifier of the document resulting from step 1002 is set to the value specified in the search condition expression. A match is extracted as the final result, and the process ends. The structure string of the search condition expression is
If not Null, a search is performed for the structural character string using index information in step 1005. Next, in step 2803, a set of the document number and the element number existing in both of the results of step 1002 and step 1005 is extracted as a provisional result.
03, a document whose document type identifier matches the value specified in the search condition expression is extracted as a final result from among the provisional results obtained in step 03,
finish.

In this example, the search condition is “document whose document type is [paper] and the title of Chapter 1 includes a character string“ method ””. = 1
By creating & structure character string = “* \ 7_1 \ 8_1 \” & content character string = “* Registered *”,
26, 2604 and 2611 in FIG. 26 are obtained. 2611 indicates that the document type is [paper] and that the character string "method" is included in [chpt-title], but the document type / structure table creation unit 2001 has the same document type as [paper]. Since the same value is assigned to the type identifier and the corresponding structure character string, both 2604 and 2611 can be obtained as search results.

The search result display processing is the same as in the first and second embodiments, and a description thereof will be omitted.

As described above, in the present embodiment, in the document type / structure table creation unit 2001, the same value is used for the document type identifier and the tag identifier for different DTDs defining the same structure. When a search is performed, documents having the same structure but different document types can be extracted as search results from one search condition expression.

(Fourth Embodiment) In a structured document search system according to a fourth embodiment of the present invention, a document type identifier and a tag identifier are added to index information of a content character string to improve search efficiency. ing.

The configuration of this system is not different from that of the third embodiment shown in FIG. However, the contents of the index information created by the index information creation unit 108 are different, and as a result, the processing of the index information search unit 110 is different from that of the third embodiment.

Here, the flow of the document registration process in the fourth embodiment will be described. First, the case of the structured document example shown in FIG. 3 as the first document to be registered will be described. The structured document in FIG. 3 has a “paper” type DTD shown in FIG. Structure analysis unit 106, document type / structure table creation unit 200
1, and the processing of the structure information holding document data creation unit 107 is the same as that of the third embodiment, and therefore the description is omitted. As a result, the same structure information holding data as in FIG. 26 is obtained.

Next, FIG. 29 shows a processing flow of the index information creating unit 108 according to the fourth embodiment. The difference from the third embodiment is that when creating index information on a character chain from a content character string, information of a document type identifier and a tag identifier is added in addition to a document number, an element number, and a character position number. (Step 2901) Only the point. The method of creating the structural character string index information is the same as in the third embodiment.

FIG. 30 shows a part of an example of content character string index information created for records 2601 and 2608 in the structural information holding data of FIG. 26 by the index information creating unit 108 in the fourth embodiment. FIG. 30 in FIG.
01 is the character string "Book Check" in the content string of the element with the document type identifier "1" and the document number "10" and the element number "1" and the tag identifier "3". It means that the chain exists from the position of the “2” character from the head ”.

Next, the flow of a document search process according to the fourth embodiment will be described. FIG. 31 shows the processing in the search condition analysis unit 109. Processing related to document type designation (steps 1801, 2701, 2702, 1803)
Are the same as those in the third embodiment, and thus description thereof is omitted.
Next, in step 3101, it is checked whether or not a structure search in which only the tag name is specified. A structure search that specifies only the tag name is
For example, structure information such as "document whose document type is [thesis] and the chapter title includes" registration method "" is given structure information of only a tag name, and there may be a plurality of structure elements. If it is a search that specifies only the tag name, the structure table is read (step 2703), and a tag identifier corresponding to the specified tag name is extracted (step 3102).
A search condition expression including a document type identifier, a tag identifier, and a content character string is created (step 3103). When a search condition as in the above example is given, the search condition expression is “document type identifier = 1 & tag identifier = 8 & content character string =“ * registration method * ””. Note that this search condition expression is used when the document type table and the structure table are created as shown in FIG. If the search does not specify only the tag name, the processing from step 901 in FIG. 27 is performed to create a search condition expression similar to that of the third embodiment (step 3104).

FIG. 32 shows the processing of the index information search unit 110 according to the fourth embodiment. Step 32
In step 1001, it is determined whether or not the search condition expression includes a structural character string designation. Then, a search is performed using only the content character string index information (step 3202).
If the search condition expression includes a structural character string in step 3201, the same search processing as in the third embodiment shown in FIG. 28 is performed (step 3203).

The details of the search performed in step 3202 using only the content character string index information will be described with reference to FIG. The search condition formula is “document type identifier = 1 & tag identifier = 8 & content character string =“ * registration method * ””. As for the character string “registration method”, “registration” and “recording method” can be extracted as a character chain of two characters. Here, the number of characters in the character chain to be extracted is the same as the number of characters in the character chain in the index information. Assuming that index information such as that shown by 3310 in FIG. 33 has been created for these two character chains, those having the same document number and element number are extracted from them. In the example of FIG. 33, this is 3321 and 3322.
Where the character position numbers are consecutive from the “registration” chain to the “recording” chain, and the values of the document type identifier and tag identifier match the search condition expression. Extract as In the example of FIG. 33, 3321 can be extracted, and a set of the document number and the element number is set as a search result.

In the example of FIG. 33, both the document type identifier and the tag identifier are specified in the search condition expression. However, when only the document type identifier is specified, the value of the tag identifier of the index information is used. May be ignored and those having the same document type identifier value may be extracted.

The search result display processing is the same as in the first, second, and third embodiments, and therefore, the description is omitted.

As described above, in the fourth embodiment, when registering a document, the index information creating unit 108 adds the document type identifier and the tag identifier of the document to be registered to the index information related to the content character string. By doing so, when performing a search, in the case of a search request in which a structural condition is specified by a document type or a tag name, only the index information on the content character string is searched, and the index information on the structural character string is obtained. Since it is not necessary to perform a search for, a high-speed search can be performed.

(Fifth Embodiment) In a structured document search system according to a fifth embodiment of the present invention, data of the order of appearance is excluded from the structure character string, and the information of the order of appearance is stored outside the structure character string. keeping.

The structure of this structured document search system is the same as that of the third embodiment shown in FIG. However, the contents of the document type table created by the document type / structure table creation unit 2001 are different, and accordingly, the structure information holding document data creation unit 107, the search condition analysis unit 109, and the index information search unit 110
Is different from that of the third embodiment.

Here, the flow of the document registration process in the fifth embodiment will be described. Here, the case of the structured document example shown in FIG. 3 as the registration target document will be described. The structured document in FIG. 3 has a “paper” type DTD shown in FIG. The processing of the structure analysis unit 106 is the same as that of the third embodiment, and thus the description is omitted.

Next, the document type in the fifth embodiment will be described.
The document type table created by the structure table creation unit 2001 will be described. When the DTD of the document to be registered is as shown in FIG. 2, as described in the description of FIG. 2 in the first embodiment, <chapter>, <paragraph>,
It can be seen that multiple <section> elements can appear in a document. These element names (tag names) are added to the fields of the document type table.

FIG. 34 shows a document type table and a structure table when the DTD shown in FIG. 2 is created by the method of the fifth embodiment. As shown in FIG. 34, in addition to the fields of the document type table in the third embodiment,
Has a field to record the tag name of the element that can appear repeatedly.

Next, FIG. 35 is a diagram showing the structure information holding document data created by the structure information holding document data creation unit 107 in the fifth embodiment when the registration target document is that of FIG. Reference numerals 3501 to 3507 are assigned to distinguish each of the records shown in FIG. 35, and will be used in the description of a search process using an example described later. As can be understood from comparison with FIG. 26, in the fifth embodiment, the order of appearance of each element is not included when creating a structural character string. Also,
A repeated appearance field is added, which corresponds to FIG. 34. The “repeated appearance 1” field records the appearance order of the element “chapter” (tag identifier: 7) in the record, and the structural character string of the record Tag identifier corresponding to "chapter": 7
Has no value if is not included. Similarly, "Repeat occurrence 2"
The “repeated appearance 3” field records the appearance order of “paragraph” (tag identifier: 11) and “section” (tag identifier: 9) in the record.

Index information creating section in the fifth embodiment
The processing of 108 is the same as that of the third embodiment, and the description is omitted.

Next, the flow of a document search process according to the fifth embodiment will be described. FIG. 36 shows a flow of processing of the search condition analysis unit 109 according to the fifth embodiment. Only the steps 3601, 3602, 3603, and 3604 are different from the processing in the third embodiment. In step 3601, it is determined whether or not the search is for a specified element or less. A search that specifies a certain element or less is a condition such as "including a character string" OO "in Chapter 1." That is, in this case, “sections” following the “chapter” are included in the search target. This “certain element” is not unique like “Chapter 1”, it is simply “Chapter”
A plurality of items may be used.

According to the result, steps 3602, 36
In step S03, a structure condition character string is created as performed in steps 2704 and 2705 in FIG. 27. At this time, the structure condition string not including the order of appearance is learned according to the structure character string format in the fifth embodiment. Create a string. In the next step 3604, when the specification of the structural condition specifies the appearance order of a certain element, the document type table is referred to and the appearance order of the corresponding repeated appearance is extracted.

As a specific example, the case where the search condition is “document whose document type is [paper] and in the section of Chapter 1 the“ title ”includes“ input ”” will be described. In this case, since the structure condition specification is not a condition specifying “a certain element or less”, “* \ 7 \ 9 \ 10 \” is created from step 3603 as a structure condition character string for backward matching that does not include the order of appearance. Next, the appearance order is specified in the structural condition in "chapter", and the appearance order is 1. Here, from the document type table of FIG. 34, in the document type [thesis], "chapter" is registered in the repeated occurrence 1 field, so that the expression "repeated appearance 1 = 1" is obtained. The search condition formula finally created at the step 905 is “document type identifier = 1 & structure character string =“ * \ 7 \
9 \ 10 \ "& repeated occurrence 1 = 1 & content string =
“* Input *” ”.

Next, the processing in the index information search unit 110 in the fifth embodiment will be described with reference to the search condition formula used in the description of the search condition analysis unit 109. Search condition expression “document type identifier = 1 & structure character string =“ * \ 7 \
9 \ 10 \ "& repeated occurrence 1 = 1 & content string =
For "* input *", a search is performed using the index information for the structural character string and the content character string in the same manner as in the third embodiment, and the value of the document type identifier is 1 in the result. Is 3506 in FIG. In the fifth embodiment, the specification of the repeated appearance field is further checked here. Since “repeated occurrence 1 = 1” is specified in the search condition expression, reference is made to the 3506 repeated occurrence 1 field, and since this value is the same as the specified value “1”, 3506 is used as the final search result. Appropriate.

The search result display processing in the fifth embodiment is the same as in the first, second, third, and fourth embodiments, and therefore will not be described.

In the present embodiment, an example has been described in which a document type identifier and a tag identifier are created as in the third embodiment, and a structural character string is created using the tag identifier. Example of creating a structural character string using tag names as they are as in the second embodiment. 34 is provided with means for holding information on repeated occurrences as in place of the document type table in FIG. 34, and as long as the structural information holding document data holds information on repeated occurrences as shown in FIG. It is feasible.

As described above, in the fifth embodiment, at the time of document registration, only the elements that can appear a plurality of times have information on their appearance order outside the structure character string, and thus the structure character string Fewer characters and faster search speed. Further, by retaining the information of the appearance order outside the structure character string, it is possible to cope with various structure designations.

[0115]

As is clear from the above description, the structured document search system of the present invention converts a registered structured document into
It is possible to search by designating various structural conditions.
Also, by registering the logical structure and the document type of the structured document in consideration of the logical structure and the document type at the time of the search, the search can be realized by specifying the logical structure and the document type. Performs a high-speed structural condition designation search.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a structured document search system according to a first embodiment of the present invention;

FIG. 2 is a diagram showing an example of a DTD (document type definition) according to the first embodiment of the present invention;

FIG. 3 is a diagram showing an example of a structured document according to the first embodiment of the present invention;

FIG. 4 is a diagram schematically showing the structure of the structured document of FIG. 3;

FIG. 5 is a diagram showing a processing procedure of a structure information holding document data creating unit according to the first embodiment of the present invention;

FIG. 6 is a diagram showing an example of structure information holding document data according to the first embodiment of the present invention;

FIG. 7 is a diagram showing a processing procedure of an index information creating unit according to the first embodiment of the present invention;

FIG. 8 is a diagram showing an example of index information according to the first embodiment of the present invention;

FIG. 9 is a diagram showing a processing procedure of a search condition analysis unit according to the first embodiment of the present invention;

FIG. 10 is a diagram showing a processing procedure of an index information search unit according to the first embodiment of the present invention;

FIG. 11 is a diagram showing details of a search method using index information according to the first embodiment of the present invention;

FIG. 12 is a diagram showing a processing procedure of a structure information holding document data creating unit according to the second embodiment of the present invention;

FIG. 13 is a diagram showing an example of structure information holding document data according to the second embodiment of the present invention;

FIG. 14 shows a DTD according to the second embodiment of the present invention.
A diagram showing a second example of (document type definition),

FIG. 15 is a diagram showing a second example of a structured document according to the second embodiment of the present invention;

FIG. 16 is a diagram schematically showing the structure of the structured document in FIG. 15;

FIG. 17 is a diagram showing a second example of the structure information holding document data according to the second embodiment of the present invention;

FIG. 18 is a diagram showing a processing procedure of a search condition analysis unit according to the second embodiment of the present invention;

FIG. 19 is a diagram showing a processing procedure of an index information search unit according to the second embodiment of the present invention;

FIG. 20 is a configuration diagram of a structured document search system according to a third embodiment of the present invention;

FIG. 21 shows a document type according to the third embodiment of the present invention.
Diagram showing the processing procedure of the structure table creation unit,

FIG. 22 is a diagram showing an example of a document type table and a structure table at an initial point in the third embodiment of the present invention;

FIG. 23 shows a DTD according to the third embodiment of the present invention.
A diagram showing an example of (document type definition),

FIG. 24 is a diagram showing an example of a structured document according to the third embodiment of the present invention;

FIG. 25 is a diagram showing an example of a document type table and a structure table at the time when registration of the DTD of FIG. 23 is completed in the third embodiment of the present invention;

FIG. 26 is a diagram showing an example of structure information holding document data according to the third embodiment of the present invention;

FIG. 27 is a diagram showing a processing procedure of a search condition analysis unit according to the third embodiment of the present invention;

FIG. 28 is a diagram showing a processing procedure of an index information search unit according to the third embodiment of the present invention;

FIG. 29 is a diagram showing a processing procedure of an index information creating unit according to the fourth embodiment of the present invention;

FIG. 30 is a diagram illustrating an example of content character string index information according to the fourth embodiment of the present invention;

FIG. 31 is a diagram showing a processing procedure of a search condition analysis unit according to the fourth embodiment of the present invention;

FIG. 32 is a diagram showing a processing procedure of an index information search unit according to the fourth embodiment of the present invention;

FIG. 33 is a diagram showing details of a search method using only the content character string index information performed in step 3202 of FIG. 32 according to the fourth embodiment of the present invention;

FIG. 34 is a diagram showing an example of a document type table and a structure table according to the fifth embodiment of the present invention;

FIG. 35 is a diagram showing an example of structure information holding document data according to the fifth embodiment of the present invention;

FIG. 36 is a diagram showing a processing procedure of a search condition analysis unit according to the fifth embodiment of the present invention;

FIG. 37 is a diagram showing a configuration of a document registration system according to a conventional technique;

FIG. 38 is a diagram showing a process of generating a structure index in a conventional technique.

[Explanation of symbols]

 101 display means 102 storage means 103 input means 104 search engine 105 search database 106 structure analysis unit 107 structure information holding document data creation unit 108 index information creation unit 109 search condition analysis unit 110 index information search unit 111 result list display unit 112 entity display Unit 113 Structure analyzed data storage unit 114 Structure information holding document data storage unit 115 Index information storage unit 116 Entity data storage unit 117 List data storage unit 2001 Document type / structure table creation unit 2002 Document type / structure table storage unit 3701 Document structure Analysis program 3702 Structural index creation program 3703 Structured full-text data generation program 3704 Character string index creation program 3705 Parsed document data storage area 3706 Structural index storage area 3707 Structured full-text data storage area 3708 Character string index storage area

 ──────────────────────────────────────────────────の Continued on the front page (72) Inventor Ken Tsurubayashi 1006 Kadoma Kadoma, Osaka Prefecture Matsushita Electric Industrial Co., Ltd. F-term (reference) 5B075 ND03 ND35 NK02 NK43 NK54 UU05

Claims (7)

[Claims] In a document retrieval system for handling a structured document, a structure analyzing means for analyzing a logical structure of a structured document to be registered, and a hierarchical structure of each element divided into logical structures by the structure analyzing means. Structure information holding document data creating means for creating structure information holding document data containing information; index information creating means for creating index information for performing a search from the created structure information holding document data; A search condition analyzing unit that analyzes a search condition and converts the search condition into a search condition suitable for a search process; and an index information search unit that performs a search using the index information based on the search condition. Structured document search system. 2. An apparatus according to claim 1, further comprising: a result list displaying means for displaying a search result list searched by said index information searching means; and an entity displaying means for displaying entity data of a document selected from said search result list. Claim 1 characterized by the following:
Structured document search system described in 1. 3. The structure information holding document data generating means converts the logical structure information of each element of the registration target document into a character string obtained by connecting a tag name, its appearance order, and a delimiter to the structure information holding document data. The structured document search system according to claim 1, wherein the structured document search system is included. 4. The structured document according to claim 1, wherein the structure information holding document data creating unit includes a character string indicating document type information of the registration target document in the structure information holding document data. Search system. 5. A document type for assigning a unique identifier to each of a document type and a tag name of a document to be registered before passing the structure-analyzed data analyzed by the structure analyzing means to the structure information holding document data creating means. 2. The structured document search system according to claim 1, further comprising a structure table creating unit. 6. The structured document search system according to claim 5, wherein said index information creating means creates index information including said identifier uniquely assigned to each of a document type and a tag name. 7. The document type / structure table creation means,
The structure information holding document data generating means holds information on a plurality of appearing tag names, and includes information on an appearance order in the structure information holding document data only for elements having a plurality of appearing tag names. Item 6. A structured document search system according to Item 5.