JP2006195667A - Structured document search device, structured document search method and structured document search program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a structured document search device that can present in what context a sentence found in a structured document appears. <P>SOLUTION: The structured document search device comprises a storage part 130 for storing a structured document organized in a hierarchical structure of a plurality of components, a reception part 110 for receiving a query, a search part 120 for searching the structured document for a component matching the query, an extraction part 140 for extracting an additional component having an element name relationship with the component found from the structured document, a means 140 for determining hierarchical relationships between the found component and the extracted additional components, and an output part 150 for outputting the hierarchical relationships. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a document search apparatus, and more particularly to a structured document search apparatus that searches for a component of a structured document. Here, a structured document is information such as XML (Extensible Markup Language) that expresses the components of a document (chapters, sections, paragraphs, abstracts, authors, titles, etc.) explicitly in the document in the form of text. Refers to the described electronic document.

  Currently, with the evolution of information and communication technologies such as the Internet, it has become possible to easily obtain enormous amounts of electronic data. On the other hand, since the amount of information is enormous, necessary information is buried in the enormous amount of data, and as a result of being unable to search as expected, there is a problem that it cannot be fully utilized.

  In order to eliminate such adverse effects, researches have been conducted to make electronic data structured documents, thereby facilitating information sharing and making information retrieval more efficient. For example, in HTML (Hyper Text Markup Language), document components, such as document titles, headings, paragraphs, etc., are described by tags. Further, XML (Extensible Markup Language), which has been attracting attention in recent years, is superior in flexibility and extensibility to HTML because it can create this tag independently.

  For structured documents such as XML, a query language having a syntax similar to SQL (Structured Query Language) and capable of describing a search position, a search condition, an information extraction part, and the like is provided.

  On the other hand, research on techniques for retrieving structured documents in natural language is being conducted. For example, an appropriate search result for a question such as “How much virtual reality has contributed to medical care” is searched from a plurality of articles described and structured in XML. At this time, the thesis is structured in units of title, chapter, section, paragraph, etc., and in the conventional document search, it was possible to search only in units of articles, but in the structured document search system, It has become possible to search for more detailed units than conventional text searches such as chapters, sections and paragraphs.

  When searching for an element that constitutes a part of a structured document appropriate for a question from such a structured document, if only the searched component is directly presented to the user, the component is structured. There is a problem that the search result is very poorly browsed because it is not known in what context it appears in the document.

For this reason, in the conventional structured document search apparatus, instead of displaying the search result suddenly, an overview of the search result is displayed, and when the XML document that the user wants to see details is found, the route of the XML document is displayed. The detailed contents of a document (for example, product information) are displayed by clicking on the component name (for example, product information) with a mouse. Further, when there is a part to be viewed in detail, the detailed contents of the lower layer components can be displayed by clicking on the part with the mouse (see, for example, Patent Document 1).
JP 2004-126770 A (pages 6-7, FIGS. 2, 9, 11, 12)

  The conventional structured document search apparatus described above is effective when the searcher knows the tree structure of the database. This is because it is necessary for the user to search the lower layer from the root of the tree structure by clicking with the mouse.

  However, there is a problem that the context in which the search result appears cannot be presented.

  An object of the present invention is to provide a structured document search device, a structured document search method, and a structured document search program capable of presenting in what context a sentence searched from a structured document appears. It is.

  In the first invention, a storage unit for storing a structured document having a hierarchical structure composed of a plurality of components, a reception unit for receiving a question, and a component corresponding to the question are searched from the structured document. A search unit; an extracted unit that extracts an additional component having a related element name from the structured document; the searched component; and the extracted additional component A structured document search apparatus including means for obtaining a hierarchical relationship and an output unit for outputting the hierarchical relationship.

  According to a second aspect, in the structured document according to the first aspect, the extraction unit further extracts the extracted additional component from the extracted additional component by a hierarchical distance from the searched component. A search device.

  According to a third aspect of the present invention, a storage unit that stores a structured document having a hierarchical structure composed of a plurality of components, a reception unit that receives a question, and a component corresponding to the question are searched from the structured document. A search unit; an extracted unit that extracts an additional component related to the searched component and lexical information from the structured document; the searched component; and the extracted additional component A structured document search apparatus including means for obtaining a hierarchical relationship and an output unit for outputting the hierarchical relationship.

  According to a fourth aspect, in the structured document according to the third aspect, the extraction unit further extracts the extracted additional component from the extracted additional component based on a lexical distance from the searched component. A search device.

  The fifth invention includes a reception step for accepting a question, a search step for searching for a component corresponding to the question from a structured document having a hierarchical structure composed of a plurality of components, and the searched component An extraction step of extracting an additional component having an element name relationship from the structured document, a step of obtaining a hierarchical relationship between the retrieved component and the extracted additional component, and the hierarchical relationship A structured document search method comprising: an output step of outputting.

  The sixth invention includes a step of accepting a question, a search step of searching for a component corresponding to the question from a structured document having a hierarchical structure composed of a plurality of components, and the searched component Extracting an additional component related to vocabulary information from the structured document; obtaining a hierarchical relationship between the retrieved component and the extracted additional component; and the hierarchical relationship A structured document search method comprising: an output step of outputting.

  According to a seventh aspect of the present invention, there is provided a reception procedure for receiving a question, a search procedure for searching for a component corresponding to the question from a structured document having a hierarchical structure composed of a plurality of components, and the searched component An extraction procedure for extracting additional components related to vocabulary information from the structured document, means for obtaining a hierarchical relationship between the retrieved component and the extracted additional component, and the hierarchical relationship Is a structured document search program that causes a computer to execute an output procedure for outputting.

  According to an eighth aspect of the present invention, there is provided a procedure for accepting a question, a search procedure for searching for a component corresponding to the question from a structured document having a hierarchical structure composed of a plurality of components, and the searched component An extraction procedure for extracting an additional component related to vocabulary information from the structured document, a procedure for obtaining a hierarchical relationship between the retrieved component and the extracted additional component, and the hierarchical relationship Is a structured document search program that causes a computer to execute an output procedure for outputting.

  According to the present invention, there are provided a structured document search device, a structured document search method, and a structured document search program capable of presenting in what context a sentence searched from within a structured document appears. Can do.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. Examples of structured documents include documents described in XML (Extensible Markup Language), SGML (Standard Generalized Markup Language), and the like. XML is a standard defined by the World Wide Web Consortium (W3C). SGML is a standard defined by ISO (International Organization For Standardization). Each is a structured document convention that allows documents to be structured. Hereinafter, a document described in XML as a structured document will be described as an example.

(First embodiment)
FIG. 1 is a schematic configuration diagram of a structured document search apparatus according to the first embodiment. The structured document search apparatus 100 structures an input unit 110 that receives a user's question input with a keyboard or the like, a structured document storage unit 130 that stores a structured document to be searched, and an appropriate answer to the question. A search unit 120 for searching for a component in the document and an additional component from the structured document storage unit that directly represents in what context the searched component appears in the structured document. An additional component extraction unit 140 for extraction and an output unit 150 that outputs to a display or the like and presents a search result to the user are provided.

  FIG. 2 is a flowchart of the structured document search apparatus according to the first embodiment. FIG. 3 is a diagram illustrating an example of a structured document to be searched. Here is an instruction manual for the camera.

  When the user inputs a question sentence with a keyboard or the like, the input unit 110 accepts this (S110). Here, as an example of the question sentence, “how to focus when shooting a nearby object” is given.

  The search unit 120 extracts keywords from the question sentence (S120). As a keyword, an independent word such as a noun may be extracted. Here, “near”, “photographing”, and “focus” are extracted. A search is executed based on this keyword.

  The structured document storage unit 130 stores structured documents in units of constituent elements. The component element is, for example, an element delimited by tags <PARA> and </ PARA>, and an index is attached to each component element. As shown in FIG. 4, the component index is composed of a vocabulary index, a structure information index, and a body text index.

  The body index describes the body of the component. In the structure information index, structure information such as a parent-child relationship and a sibling relationship of each component of the structured text is described. In the parent-child relationship, the upper layer component of the tree structure is a parent, and the lower layer component is a child. A sibling relationship refers to a component in the same hierarchy. For example, in the case of the components in Chapter 1, Section 2, Item 2 shown in FIG. 4, the components of the sibling relationship are Chapter 1, Section 2, Item 1 or Chapter 1, Section 2, Item 3. The term “component” refers to the component in Chapter 1, Section 2, or Chapter 1. In addition, since Chapter 2 Section 2 Item 2 is the lowest component, it can be seen that there are no child-related components.

  The vocabulary index describes independent words and frequency information of the constituent elements and constituent elements of the child relation. In FIG. 4, since Chapter 1 Section 2 Section 2 is the lowest order component, there are no child-related components, and the vocabulary index contains only independent words for the components in Chapter 1 Section 2 Section 2. Is described. Whether the vocabulary index has child element vocabulary depends on the system. If you have a vocabulary of child elements, the speed of the search itself will be faster. However, since the index is redundant, the disk size increases, so if you want to reduce the disk size, it is better not to hire the vocabulary index to have that vocabulary.

  Next, the search unit 120 refers to the vocabulary index in each component index, and calculates the importance of each component for the extracted keyword (S130). The importance may be calculated using a TF (Term Frequency) / IDF (Inverted Document Frequency) method or the like. In this method, the importance of a component is calculated by combining the following two indicators. TF is the frequency of keywords appearing in a document and indicates the completeness of the keywords. IDF is a logarithm of the total number of documents divided by the number of documents in which the keyword appears. Indicates. In order to calculate the IDF, it is necessary to obtain the total number of documents and the number of documents in which keywords appear. In a normal document search, a unit of a document to be searched is explicitly given in advance.

  However, when searching for a component from a document as in the present embodiment, the component that is the unit of search is not known in advance. For example, the total number of documents in the structured document is calculated in the document search. Calculation is performed by replacing with the number of elements and replacing with the number of elements in which the keyword appears. Next, the importance of each component is obtained from the total importance of the extracted keywords, and the component having the highest importance is used as the search result.

  FIG. 5 is a diagram in which search results are extracted from the question text. In the present embodiment, among the three keywords “near”, “photograph”, and “focus” in the question sentence, the first chapter, the second section, the second term, which is a component including two of “near” and “photograph”. Was obtained as a search result. The search unit 120 sends the search result to the additional component extraction unit 140.

  The additional component extraction unit 140 extracts a component that is closely related to the searched component in the tree structure as an additional component. First, component elements (candidate additional component elements) that are candidates for additional component elements are extracted (S150). Here, ancestor elements, sibling elements, and descendant elements of the searched constituent elements are extracted as additional constituent element candidates. If the retrieved component is Chapter 1, Section 2, Section 2, then Chapter 1 Section 2 or simply Chapter 1 as the ancestor element, Chapter 1 Section 2 Section 1 or Chapter 1 as the sibling element Extract Section 3 and Section 3. Note that there are no descendant elements in this case.

  Next, the extracted additional component candidate is selected by element name (S160). FIG. 6 is a list of element names that should remain as additional component candidate. Here, TITLE, MEMO, and SUBITLE are left. These are headings for chapters, sections, etc., and are useful for the user to understand the context.

  Subsequently, from the selected additional component candidates, those having a hierarchical distance from the searched component to a certain distance are excluded from the candidates (S170). As the hierarchical distance, for example, a parent-child relationship distance is set to 1 and a distance greater than 2 is excluded. Through this process, the additional components shown in FIG. 7 are extracted from the structured document shown in FIG. 3 (S180).

  The output unit 150 acquires the searched component and the extracted additional component. By referring to the structure information index of the obtained component element, the hierarchical relationship between the retrieved component element and the extracted additional component element is obtained (S190). As a result, in the search result as shown in FIG. 5 and the additional component as shown in FIG. 7, the additional component (a) is a sibling with the search result, and the appearance order (a) appears first. It can be seen that the additional component (b) is a sibling with the parent element of the search result, and that (b) appears earlier than the parent of the component.

  Next, the output unit generates output information based on the hierarchical relationship described above. For example, a state in which the appearance order is maintained is output for elements that are siblings with the searched constituent element, and an ancestor element is output as an ancestor element, and an output result such as that shown in FIG. 8 can be obtained (S200). Through the processing described above, it is possible to present to the user in an easy-to-understand manner the context in which the search result appears in the structured document.

(Second Embodiment)
Next, a second embodiment will be described using the flowchart of FIG. The second embodiment is different from the first embodiment in the process of determining additional components. In the first embodiment, selection by element name and exclusion by hierarchical distance are performed (S160 and S170 in FIG. 2). On the other hand, in the second embodiment, acquisition of vocabulary information and selection based on closeness of words (S260 and S270 in FIG. 9) are performed. The other parts are the same as those in the first embodiment, so please refer to the description of the same reference numerals.

  The additional component extraction unit 140 extracts additional component candidates as in the first embodiment (S150). Next, referring to the vocabulary index in the component index, the component searched by the search unit 120 and the vocabulary information of the additional component candidate are acquired (S260).

  Then, the additional component extraction unit 140 calculates a lexical distance between the searched component and the additional component candidate. As this calculation method, for example, a vector space model may be used in units of words. For the vector space model, please refer to the literature (Takenobu Tokunaga: Information Retrieval and Language Processing: The University of Tokyo Publishers: 1999). By this calculation, additional components that are lexically similar to the retrieved components are extracted (S270, FIG. 10). The output unit 150 performs output processing similar to that of the first embodiment, and can obtain an output result as shown in FIG.

  Note that the additional component extraction unit 140 has a function of selecting by element name and exclusion by hierarchical distance in the first embodiment, acquisition of vocabulary information in the second embodiment, and selection by word proximity It is possible to combine the functions. FIG. 12 shows the extraction result when combined, and FIG. 13 shows the output result when combined.

  The above-described embodiment is a preferable specific example of the present invention, and thus various technically preferable limitations are attached. However, the embodiments may be appropriately combined and changed within a range not departing from the gist of the present invention. Needless to say, you can. For example, the structured document search device is a computer such as a server, and the hierarchical relationship between the searched component and the additional component can be acquired. Therefore, this hierarchical relationship can be output as a bar display to a terminal equipped with viewer software. And the terminal can display the bar which displayed the position which the searched component and the appearance position of the additional component showed clearly, using the visual function of viewer software. FIG. 14 is a diagram showing the output results according to the first embodiment of FIG. 8 in a bar display. The entire bar represents Section 2 “Set focus”, which makes it easier for the user to understand the relative position of the search results.

1 is a schematic configuration diagram of a structured document search device according to a first embodiment. FIG. 5 is a flowchart of the structured document search device according to the first embodiment. Diagram showing an example of a structured document to be searched The figure which shows an example of a component element index. The figure which extracted the search result from the question sentence. A list of element names to be left as additional component candidate candidates according to the first embodiment. The figure which shows the extraction result of the additional component extraction part which concerns on 1st Embodiment. The figure which shows the output result which concerns on 1st Embodiment. 9 is a flowchart of a structured document search device according to a second embodiment. The figure which shows the extraction result of the additional component extraction part which concerns on 2nd Embodiment. The figure which shows the output result which concerns on 2nd Embodiment. The figure which shows the extraction result at the time of combining 1st and 2nd embodiment. The figure which shows the output result at the time of combining 1st and 2nd embodiment. The figure which displayed the output result which concerns on 1st Embodiment of FIG. 8 on the bar display.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Structured document search apparatus 110 Input part 120 Search part 130 Structured document storage part 140 Additional component extraction part 150 Output part

Claims (8)

A storage unit for storing a structured document in which a hierarchical structure is configured by a plurality of components;
A reception unit that accepts questions,
A search unit that searches the structured document for a component corresponding to the question;
An extraction unit that extracts additional structural elements having element names related to the retrieved structural elements from the structured document;
Means for obtaining a hierarchical relationship between the retrieved component and the extracted additional component;
A structured document search apparatus comprising: an output unit that outputs the hierarchical relationship.   The structured document search apparatus according to claim 1, wherein the extraction unit further extracts the extracted additional component from a hierarchical distance to the searched component from the extracted additional component. A storage unit for storing a structured document in which a hierarchical structure is configured by a plurality of components;
A reception unit that accepts questions,
A search unit that searches the structured document for a component corresponding to the question;
An extraction unit that extracts additional structural elements related to the searched structural elements and vocabulary information from the structured document;
Means for obtaining a hierarchical relationship between the retrieved component and the extracted additional component;
A structured document search apparatus comprising: an output unit that outputs the hierarchical relationship.   4. The structured document search apparatus according to claim 3, wherein the extraction unit further extracts the extracted additional component from the extracted component according to a lexical distance from the searched component. A reception step for accepting questions;
A search step of searching for a component corresponding to the question from a structured document in which a hierarchical structure is configured by a plurality of components;
An extraction step of extracting additional structural elements having element names related to the retrieved structural elements from the structured document;
Obtaining a hierarchical relationship between the retrieved component and the extracted additional component;
A structured document search method comprising: an output step of outputting the hierarchical relationship. A reception step for accepting questions;
A search step of searching for a component corresponding to the question from a structured document in which a hierarchical structure is configured by a plurality of components;
An extraction step of extracting from the structured document an additional component related to the searched component and vocabulary information;
Obtaining a hierarchical relationship between the retrieved component and the extracted additional component;
A structured document search method comprising: an output step of outputting the hierarchical relationship. Acceptance procedure to accept questions,
A search procedure for searching for a component corresponding to the question from a structured document in which a hierarchical structure is configured by a plurality of components;
An extraction procedure for extracting from the structured document an additional component related to the searched component and vocabulary information;
Means for obtaining a hierarchical relationship between the retrieved component and the extracted additional component;
A structured document search program for causing a computer to execute an output procedure for outputting the hierarchical relationship. Acceptance procedure to accept questions,
A search procedure for searching for a component corresponding to the question from a structured document in which a hierarchical structure is configured by a plurality of components;
An extraction procedure for extracting from the structured document an additional component related to the searched component and vocabulary information;
Obtaining a hierarchical relationship between the retrieved component and the extracted additional component;
A structured document search program for causing a computer to execute an output procedure for outputting the hierarchical relationship.

Images