JP2000099543A - Information retrieval device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an information retrieval device which integrally deals with document information of different document formats and retrieves a document through the use of relevance shown in the document. SOLUTION: A document format common part 1 previously makes the document format of document information to be common and it is storaged in a document information storage part 2. When a user designates document information and logic structure from a designation device 5 at the time of retrieval, a logic structure extraction part 3 extracts designated logic structure on designated document information. A document information retrieval part 4 retrieves document information stored in the document information storage part 2 with the content of extracted logic structure as a retrieval key. Retrieved document information is stored in a retrieval result storage part 6. Logic structure is extracted by the logic structure extraction part 3 and a document information retrieval part 4 retrieves document information on respective pieces of document information. A retrieval result is stored in a storage device 9 and relevance between document information is displayed on a display device 7 so that it can be grasped.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a document management system for storing and retrieving document information having a plurality of different document formats, and more particularly to an information retrieval apparatus for retrieving related information from a certain document. is there.

[0002]

2. Description of the Related Art Conventionally, in order to search for document information, a document format in a document management system in which documents are stored is unified, and a search is performed using a search method specific to the document format. For example, in the case of a document management system for structured documents, only the structured documents are stored in the document management system, and retrieval using the structure of the document is possible. However, in such a search, a document having no structure simply written in a document image or a word processor has not been searched. Also, the retrieved documents are displayed as independent entities, and the relevance between the documents has to be checked by the user by reading the full text.

Originally, a document is often written in relation to another document, and there are many citation relationships within the document. For example, quoting a document as a reference indicates the relevance between the cited document and the cited document. By following such a relationship, the user can more easily search for a document. However, such a relationship is rarely used in a conventional document management system.

As a technique for actively using the relationship between documents, for example, JP-A-7-230467 and JP-A-8-287087 disclose references and cited references when they are available. Documents can be displayed. For example, in Japanese Patent Application Laid-Open No. Hei 8-272818, citations and reference relationships between documents are created in advance, and documents can be searched using these relationships. Further, for example, Japanese Patent Application Laid-Open No. Hei 9-146968 discloses that a related document that is newer than the original document is obtained by not only searching for a reference document but also searching for another document that uses the searched document as a reference document. It is possible to take out.

[0005] In each of these techniques, the above-mentioned search is enabled for the first time by processing data stored in a document management system so that citation relations can be used in reference documents. For example, only bibliographic information is extracted from the reference list, a bibliographic information database is constructed separately from the text information, and the bibliographic information database is searched for references.

In the above system,
When handling a document image, it is necessary to code the document information using a character recognition technique or the like in advance and then register it. For example, in Japanese Patent Application Laid-Open No. 10-3483, a citation relationship written under a certain condition is automatically extracted from document image information using a document image processing technology developed in recent years, converted into a character code, and associated. A system for retrieving document information has been proposed. In this system, in order to minimize the costly character recognition processing, a citation relationship is found from within the document using a predetermined image pattern, and the character recognition processing is applied only to the corresponding part, and the search is performed. To retrieve bibliographic information.

[0007] However, various methods are used to describe citation relationships between documents or within a document. For example, [1], [2],
In some cases, a series of numbers may be used, such as..., Or in some cases, such as [nori98] or [taro96], depending on the publication date of the document and the author. Therefore, it is difficult to flexibly cope with various writing methods by a method of searching for a predetermined image pattern as described in the above-mentioned Japanese Patent Application Laid-Open No. 10-3483.

[0008] Further, in any system that can use the above-mentioned association between documents, it is impossible to handle electronic documents and document images in a unified manner. As described above, no technique has been proposed so far that indicates related documents for document information in which different document formats are mixed, such as electronic documents and document images.

[0009]

SUMMARY OF THE INVENTION The present invention has been made in view of the above-described circumstances. For example, document information of different document formats, such as an electronic document and a document image, is handled in a unified manner, and is indicated in the document. It is an object of the present invention to provide an information search apparatus that enables a search of a document using a related relationship such as a cited relationship.

[0010]

According to the present invention, the document formats of a plurality of document information are shared and stored. At this time, for example, if the document information is a document image, the document image is divided into regions having different properties, character recognition is performed for the character region, and the logical structure of the document is determined from the region division result and the character recognition result. If the document information created by a word processor has only the document content and format information, the logical structure of the document information is determined from the change in the format information in the document and the document content. Convert to format.

When the document information and a specific logical structure in the document information are designated, the designated logical structure is extracted from the document information designated by the logical structure extracting means, and a document content corresponding to the extracted logical structure is searched for. As document information. Further, a specified logical structure is extracted from the searched document information, and the document information in the document information storage unit is searched using a document content corresponding to the extracted logical structure as a search key. This process is repeated until the search result no longer exists. When the logical structure is extracted, it is possible to refer to the information of each node constituting the document information and extract a structure that is semantically equivalent to the specified logical structure.

In this way, it is possible to retrieve document information having a specified logical structure regardless of the document format. Further, for example, if a structure representing a citation relationship is designated as a logical structure, the retrieved document information can be extracted as information associated with each citation relationship. For example, by displaying the retrieved document information in association with each other, it is possible to provide the user with a format in which the citation relationship of the document information can be easily grasped.

[0013]

FIG. 1 is a block diagram showing an embodiment of an information retrieval apparatus according to the present invention. In the figure, 1 is a document format sharing unit, 2 is a document information storage unit, 3 is a logical structure extraction unit, 4 is a document information search unit, 5 is a designation device, 6 is a search result storage unit, 7 is a display device, 8 Is the central control unit, 9
Is a storage device.

The document format sharing unit 1 shares the document format of the document information. Document information storage unit 2
Accumulates a plurality of pieces of document information having a common document format. The logical structure extracting unit 3 extracts a specific logical structure from the document information. The document information search unit 4 searches the document information stored in the document information storage unit 2 using the content of the logical structure extracted by the logical structure extraction unit 3 as a search key. The designation device 5 includes an input device such as a mouse and a keyboard, and can designate specific document information and a logical structure. The search result storage unit 6 temporarily stores the search results. The display device 7 displays a search result and an operation status. The central controller 8 controls the operation of the entire apparatus. The storage device 9 stores programs and data executed by the central control device 8 and accumulates document information used for extracting a logical structure by the logical structure extraction unit 3 and setting a search key in the document information search unit 4. .

FIG. 2 is a flowchart showing an outline of the operation of the information retrieval apparatus according to the embodiment of the present invention.
In step S11, the document format of the document information is shared by the document format sharing unit 1 in S11 and stored in the document information storage unit 2. Then, when performing a search,
In 12, the user specifies document information and a logical structure for setting a search key from the specifying device 5. The designated document information is stored in a variable A. The document information stored in the variable A is stored in the storage device 9 in S13. S14
In, the logical structure extracting unit 3 extracts a specified logical structure from the document information stored in the variable A.
In step S15, the document information search unit 4 searches the document information stored in the document information storage unit 2 using the content of the logical structure extracted by the logical structure extraction unit 3 as a search key. The searched document information is stored in the search result storage unit 6 in S16.

In S17, it is determined whether or not the searched document information is stored in the search result storage unit 6. If the searched document information is stored in the search result storage unit 6, the process proceeds to S18. Select one of them and take it out,
The value is stored in the variable A and the process returns to S13. Then, after storing the selected document information in the storage device 9, the logical structure extraction unit 3 extracts a specified logical structure from the document information,
The document information is searched by the document information search unit 4 using the extracted contents of the logical structure as a search key. Such a process is repeated until there is no more document information in the search result storage unit 6.

When there is no document information in the search result storage unit 6 in S17, the document information searched so far is used to determine the relationship between the document information, that is, the relationship obtained from which document information is searched for which document information. It is displayed on the display device 7 in S19 so that the user can understand it.

Hereinafter, the outline of the above operation will be described in detail using a specific example. First, the document format sharing unit 1 performs a process of sharing a document format of a plurality of input document information into a specific document format. It is desirable that the specific document format used for the commonization be a document format that can express a logical structure so that a search using the logical structure can be easily performed, and that can be easily described. S as a document format that can handle logical structures
GML and HTML are known. But SGML,
The DTD restricts the tag name representing the logical element and its schema, and is not suitable for the present invention. Because
Since it is assumed that documents that represent various information and have various logical structures are input, it is difficult to predict in advance the schema of all logical structures and the tag names of all logical elements used. It is. Similarly, HTML is not suitable for the present invention because tag names representing logical elements are greatly restricted for easy use. Although a specific document format can be independently defined, the description will be made here using XML as a common document format. XML does not need to define a DTD, and has a specification that allows a user to freely define a tag name. Therefore, the XML is suitable for use as a common format in the present invention.

First, it is easy to convert from a document format in which a logical structure can be described, such as SGML or HTML, to XML. Since SGML is defined as a subset of XML, there is basically no part to change and use. However, since the present invention does not require a document type definition, the document type definition can be deleted and used.

Next, commonization of document information having only format information created by a word processor or the like will be described. Document information having only format information includes a lot of information necessary for display, such as the font and font size of each character, and paragraphs and indents. Therefore, a logical structure can be estimated from such a change in font information or the like. However, since it is difficult to estimate the exact logical structure here, "sections", "paragraphs", "text"
Only the rough structure such as is assigned. For example, section headings and the like are often emphasized by using a font larger than the text size, or by using bold font. The text may be described with indentation for section headings. Based on such information, the structure is allocated. Therefore, a tag such as a “section” or a “paragraph” is used as a tag name of each node of the XML, thereby forming a logical structure.

FIG. 3 is an explanatory diagram of a specific example of the common processing for document information having only format information. This example shows an example in which document information (RTF document) having only format information as shown in FIG. 3A is converted into a description in XML, and the description in XML as shown in FIG. can get.

In a document image, it is more difficult to estimate the structure. In the case of a document image, the image is divided into regions based on the difference in the properties of the image information, and a structure is assigned to the text region. The information available at this time is
The position information and size information on the image, and the character information in each text area can be used by using the document recognition device. However, the assignment of the structure at this time is only roughly performed based on only the positional relationship between the text regions. For example, the character areas at the top and bottom of the image area may be headers or footers, but may also be text. Therefore, at this stage, the positional relationship of the character area (text block) is set to XM
Only by describing L, the structure is finally determined by the structure matching process in the logical structure extraction unit 3.

FIG. 4 is an explanatory diagram of a specific example of the common processing for a document image. This example shows a case where a document image as shown in FIG. 4A is input. From such document images, "storage and retrieval of document information", "
The character areas of “Taro” and “Tomi Gakuen University” and the character areas indicated by placing “■” beneath each are separated, and character recognition is performed to obtain an XML description as shown in FIG. ing.

The document information converted into XML is stored in the document information storage unit 2. At this time, only the document converted into the common format may be stored, or the original document and the document converted into the common format may be stored as a pair. If stored as a pair, it is also possible to retrieve and use the search results as the original document format.

When performing a search, first, one document for setting a search key is specified. As the document to be specified, the document information stored in the document information storage unit 2 can be specified using the specifying device 5. Alternatively, the document information may be document information of an external document database or document information newly input by a user. In this case, the document information needs to be converted from the document format to the common format by the document format common unit 1 as described above.

The logical structure extracting unit 3 extracts a logical structure to be extracted from the document information specified by the user or the document information extracted from the search result storage unit 6. FIG. 5 is an explanatory diagram of a specific example of a logical structure to be extracted, and FIG. 6 is an explanatory diagram of a specific example of a logical structure included in document information. Here, as a specific example, an example of extracting a logical structure as shown in FIG. 5 showing a reference list listed at the end or the end of an academic paper is shown. In the structure shown in FIG. 5, first, a plurality of nodes having the tag name "document" are arranged as child nodes of the node having the tag name "reference list". The contents of these nodes are bibliographic information of each document for a node with a tag name of “document”, and a node with a tag name of “reference list” is “reference”, “document” or “References”.
Such as the title.

If a logical structure defines each structure with such a tag name from the beginning, the corresponding logical structure can be extracted by a matching operation between the logical structures. However, if the tag name of the stored document is an XML document defined by a different tag name or document information converted from a document image, the designated logical structure can be obtained only by a simple matching operation. Cannot be extracted.

For example, none of the structures shown in FIGS. 6A and 6B have a structure having a tag name as shown in FIG. However, in the structure shown in FIG.
If the node having the tag name of “section title” has the content of “reference” and the “paragraph” of its sibling node includes the bibliographic information of each reference, the structure is shown in FIG. It can be determined that it represents a reference list similar to the structure. Also, regarding the structure shown in FIG. 6 (B), the node having the tag name
If the child node has bibliographic information of a reference, the structure can be determined to be a reference list similar to the structure shown in FIG.

If it is not possible to make a judgment only with such a logical structure, a matching operation taking into account the characteristic contents of each node may be performed. For example, the bibliographic information of each document starts with a number or a specific character string so as to correspond to the citation in the text. While extracting such features from the contents of each node, each node can be matched with the specified logical structure to extract the specified logical structure.

FIG. 7 is a flowchart showing an example of the process of extracting the designated logical structure in the logical structure extraction unit 3. Here, an example is shown in which the structure of the reference list is extracted according to the specific example described above. First, in S21,
The first node is extracted from the document information and stored in a variable A. At this time, it is determined whether or not a node exists in S22, and the following processing is performed until the node no longer exists.

In S23, it is checked whether or not the content of the node extracted in S21 is "reference" or "document". If the content of the node is not “reference” or “document”, the process returns to S21 to retrieve the next node. By repeating such processing, it is possible to find the section heading in which the reference is listed.

When a node having "reference" or "document" is found, processing is performed on the child node or sibling node of this node. First, in S24, it is determined whether or not this node has a child node. If a child node exists, the child node is extracted in S25. In S26, it is determined whether or not child nodes exist, and as long as child nodes exist, bibliographic information is extracted for those nodes. Bibliographic information begins with a special description to correspond to the citation location in the document. Here, as an example, it is assumed that a number enclosed in parentheses is associated with a document. Therefore,
It is determined whether or not the content of each node is text. If it is text, it is determined in S27 whether or not the content is a description starting with a combination of “[”, a number, and “]”. The node starting with this description is regarded as a part of the bibliographic information of the reference, and the contents of the node are extracted in S28. If there are no child nodes, S26 to S2
Returning to step 1, the processing for finding a node having the next reference is continued.

Similarly, when there is a sibling node,
In S29, a sibling node is extracted. In S30, it is determined whether or not there are sibling nodes. As long as there are sibling nodes, bibliographic information is extracted from those nodes. In S31, it is determined whether or not the content of each node is text. If the content is text, it is determined whether or not the content is a description starting with a combination of “[”, a number, and “]”. The node starting with this description is regarded as a part of the bibliographic information of the reference, and the contents of the node are extracted in S32. If there are no more sibling nodes, S
Returning from S30 to S21, the process of finding a node having the next reference is continued.

In this example, matching is performed only by looking at the contents of each node. However, for example, by comparing tag names of nodes, matching can be performed more accurately and quickly. However, as shown in FIG. 5 and FIG. 6, since the tag names of the respective nodes are not unified, it is necessary to perform matching in consideration of the correspondence between the tag names. For example, processing may be performed according to a rule that a “reference” tag matches a “text block” tag or a “section” tag. Here, the search for the sibling node is performed, for example, as shown in FIG. 6A, such that the section title is located as a child node of the section, and the bibliographic information of the document is positioned at the sibling node. This is to cope with the document structure.

Next, the document information search unit 4 searches for information stored in the document information storage unit 2 using the content of each node of the logical structure extracted by the logical structure extraction unit 3 as a search key. The document information search unit 4 can be realized by a general document search method. Further, the bibliographic information can be decomposed into elements such as an author, a title, and an issuing institution by using a specific symbol and used as a search key of a search device. For example, in the example of the document information shown in FIG. 3B, each document content is already divided by punctuation. Therefore, a search formula can be created by specifying an author name, a title, and the like from each of these document contents. First, in order to specify the author name, a personal dictionary is compared with a character string of each document content, and if the character string is registered in the personal dictionary, the document content is assumed to be the author name. If the sequence of numbers matches the date notation, for example, "x month x day", "YYMMDD", etc., the document content can be determined to be date information. In addition, "Paper Collection", "Preliminary Collection", "in Pro
A document content including a keyword such as "c" can be determined to be a source name. The remaining document content is set as a title.

FIG. 8 is an explanatory diagram of a specific example of generating a search formula in the document information search unit. FIG. 8A shows FIG. 3B.
This is the same XML document. "T.Fuji" in the personal dictionary
i "," S. If the name of "Yamada" is registered, these contents are specified as the author name, and "Docum" is specified.
ent Image Analysis "," Docu
"Recognition" is specified as the title, and "in Procof xxx Sympos"
ium. "As the source name and" 1989. "Can be specified as a date. With these, for example, FIG.
It is possible to generate a search formula as shown in FIG.

In this example, the title is used as the keyword as it is. However, the present invention is not limited to this. Further, a morphological analysis may be performed on the title of the bibliographic information, and the keyword may be extracted and searched.

By searching the document information stored in the document information storage unit 2 using the search formula generated as described above, the user can search for a single document and a large amount of document information. Document information (in this case, a reference) can be easily retrieved. Moreover, even if the original document information has a different document format, it can be searched. At this time, the user does not need to specify a process such as conversion of a document format, and can search without any need to set a special keyword. The results searched in this way can be listed and displayed on the display device 7, for example. By selecting the document information that the user wants to refer to from the list, desired document information can be obtained.

The document information retrieved by the document information retrieval section 4 is temporarily stored in the retrieval result storage section 6. A specific logical structure is extracted again by the document structure extraction unit 3 for each of the searched document information, and the document information stored in the document information storage unit 2 is searched. Such processing is repeated until there is no more document information of the search result stored in the search result storage unit 6. The searched results are sequentially stored in the storage device 9.

In such a search alone, only documents published earlier than the first designated document are stored in the storage device 9. Therefore, the bibliographic information of the next designated document is used as a search key to retrieve the document. The document information stored in the information storage unit 2 is searched. In this case, similarly, the search results are temporarily stored in the search result storage unit 6, and the search is repeatedly performed until there is no more document information in the search result storage unit 6. At the time of retrieval, faster and more accurate retrieval is possible by limiting the range of retrieval within the document information to only the sections in which reference documents are described. The function of the logical structure extraction unit 3 can be used to find the section in which this reference is described. The bibliographic information of a document can be extracted by using information described on the front page of the document, title, author name, and author affiliation. To extract these structures, the function of the logical structure extraction unit 3 can be similarly used. The search result is also stored in the storage device 9.

In this manner, the document information having a reference relationship is stored in the storage device 9 from the document specified by the user. The document information stored in the storage device 9 can be arranged on, for example, the publication date and displayed on the display device 7. FIG. 9 is an explanatory diagram of an example of a display form of a search result. In this example, an example is shown in which a graph is displayed so that the relationship between the document information as search results stored in the storage device 9 can be understood. The document specified first is placed at the center of the display area so that it can be recognized. In FIG. 9, hatching is applied to indicate that the display mode is different. Then, the document information obtained as a result of the search is arranged around the document, and the relevance between the documents is represented by connecting the documents having the cited and cited relationship with a line. By displaying the search results in this manner, the relationship between the search results can be easily grasped at a glance. Of course, the display mode shown in FIG. 9 is an example, and may be displayed in another display mode.

[0042]

As is apparent from the above description, according to the present invention, even if the document format of the original document information is different, the user does not have to worry about the document format of the stored document information. , Search for document information, and further search for all related document information. At the time of the search, it is only necessary to specify the document information and a specific logical structure, and the search using the logical structure can be performed for the document information having no logical structure. Further, the search result can be displayed so that the user can easily grasp it.
As described above, according to the present invention, there are various effects.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of an information search device of the present invention.

FIG. 2 is a flowchart showing an outline of an operation in the embodiment of the information search device of the present invention.

FIG. 3 is an explanatory diagram of a specific example of a common process for document information having only format information.

FIG. 4 is an explanatory diagram of a specific example of a common process for a document image.

FIG. 5 is an explanatory diagram of a specific example of a logical structure to be extracted;

FIG. 6 is an explanatory diagram of a specific example of a logical structure included in document information.

FIG. 7 is a flowchart illustrating an example of a process of extracting a specified logical structure in a logical structure extraction unit.

FIG. 8 is an explanatory diagram of a specific example of generating a search formula in a document information search unit.

FIG. 9 is an explanatory diagram of an example of a display form of a search result.

[Explanation of symbols]

1 ... document format common unit, 2 ... document information storage unit,
3 a logical structure extraction unit, 4 a document information search unit, 5 a designation device, 6 a search result storage unit, 7 a display device, 8 a central control device, 9 a storage device.

Claims (5)

[Claims] 1. A document information storage means for storing a plurality of document information, a document format sharing means for sharing a document format of the plurality of document information and storing the document information in the document information storage means, Means for designating a specific logical structure, a logical structure extracting means for extracting a specified logical structure from designated document information and retrieved document information, and a logical structure extracted by the logical structure extracting means. Search means for searching the document information in the document information storage means using the contents of the document to be searched as a search key, and extracting the logical structure by the logical structure extracting means and searching for the document information searched by the search means An information search apparatus characterized in that a search by means is repeated until a search result no longer exists. 2. The information retrieval apparatus according to claim 1, further comprising display means for displaying the document information sequentially searched by said search means in association with each other. 3. When the document information is a document image, the document format commoning unit divides the document image into regions having different properties, and performs character recognition on a character region.
2. The information retrieval apparatus according to claim 1, wherein a logical structure of the document is determined from a result of the area division and a result of the character recognition, and the document is converted into a common document format. 4. The document format commoner determines a logical structure of the document information from a change in the format information in the document and the document content when the document information has only the document content and the format information. The information retrieval apparatus according to claim 1, wherein the information is converted into a format. 5. The logical structure extracting means according to claim 1, wherein said logical structure extracting means extracts a structure semantically equivalent to a specified logical structure by referring to information of each node constituting the document information. Information retrieval device.