JP2009251845A - Retrieval result evaluation device and retrieval result evaluation method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To perform the retrieval of a structured document under the consideration of an inter-keyword hierarchical relationship only by entering a plurality of keywords, and to evaluate the retrieval results. <P>SOLUTION: A vocabulary hierarchical relationship storage part 132 stores a vocabulary hierarchical relationship graph for representing an inter-vocabulary hierarchical relationship included in each of a plurality of structured documents stored in a retrieval object data storage part 131 for each structured document. A query creation part 121 creates retrieval formula for retrieving a structured document from a plurality of character strings included in retrieval conditions designated by a user by referring to the vocabulary hierarchical relationship storage part 132. A storage part 13 retrieves the structured document matched with retrieval formula created by the query creation part 121 from a retrieval object data storage part 131. A retrieval result evaluation part 141 evaluates the retrieved structured document based on a plurality of character strings included in the retrieval conditions. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a search result evaluation apparatus and a search result evaluation method for evaluating a searched structured document.

  In general, a document having a logical structure is called a structured document. In a structured document, the logical structure of the document may be indicated by a tag described in the document. A structured document in which a logical structure is expressed using this tag is suitable for processing that is interpreted or processed by a computer for various purposes. As a typical structured document, an XML document described in an XML (Extensible Markup Language) format is known.

  In recent years, XML has been used in a large number of applications, and various data have been described in XML format. As a result, a technique for retrieving an XML document described in the XML format is important.

Here, in the above XML document search condition specification,
1. Condition specification by query language such as XPath and XQuery 2. Specifying conditions by pairs of element names, attribute names and their values Condition specification by keyword (character string) Any method of condition specification in natural language is used.

  Among the above-mentioned condition specifying methods, with respect to 1 and 2, the user (user) has specialized knowledge such as a query language, or specifies the condition after grasping the structure of the XML document to be searched in advance. There is a high possibility that the user can search for a desired XML document.

  On the other hand, 3 and 4 have an advantage that an XML document can be searched even when the user does not have specialized knowledge such as a query language.

  By the way, when searching for the XML document described above, a large number of XML documents may be presented (returned) to the user as search results depending on conditions (search conditions) specified by the user, for example. In such a case, the user needs to search for a desired XML document from among a large number of presented XML documents. However, it is difficult and time consuming to find a desired XML document from among a large number of XML documents.

  For this reason, for example, each of the many XML documents searched (search results) is evaluated, and the search results are returned in a form in which, for example, the XML documents desired by the user are ranked higher, so that the user finally needs Can make it easier to find existing data.

  Here, as a method of evaluating the search result, for example, for each word included in the XML document, the importance in the XML document in which the word appears is stored in advance as an index. There is a method for evaluating the importance of all keywords.

  Further, in the search of the XML document, it is desirable to perform the evaluation in consideration of the structure of the XML document in the above-described evaluation of the search result.

With regard to the technology for evaluating the search results as described above, for example, the structure information and the filter of the document are used by a simple operation of describing which part of the structure of each document is to be created in the format file. A technique (hereinafter referred to as a prior art) that can perform appropriate fitness calculation is disclosed (see, for example, Patent Document 1).
JP 2001-325293 A

  However, in the condition specification by the above query language and the condition specification by the combination of element name, attribute name and its value, the knowledge of the query language such as XPath or XQuery and the structure of the XML document that the user wants to search in advance are grasped After that, it is necessary to specify conditions. Therefore, since a user who does not have such knowledge cannot use it, convenience is lacking.

  On the other hand, the condition specification by the keyword or the condition specification by the natural language cannot represent information for specifying the necessary data (for example, the structure of the XML document). In addition, natural language ambiguity remains for condition specification in natural language. For this reason, a large amount of data is returned as a search result in addition to the data required by the user.

  Further, when evaluating a search result, it is difficult to reflect the user's intention in the search result in a full-text search using existing keywords. Further, in the above-described prior art, for example, the order (ranking) is given based on a specified keyword and a statically defined index or the like. Therefore, in the prior art, for example, it is possible to statically apply the importance according to the structure of the XML document, but it has a hierarchical structure intended by the user based on the search conditions specified by the user at the time of search. It is difficult to dynamically rank XML documents at the top.

  An object of the present invention is to search a structured document in consideration of the hierarchical relationship between the keywords only by inputting a plurality of keywords (character strings), and evaluate the search results according to the keywords. A search result evaluation apparatus and a search result evaluation method are provided.

  According to one aspect of the present invention, a search target data storage unit storing a plurality of structured documents to be searched is searched according to a search condition including a plurality of character strings specified by a user. A search result evaluation apparatus for executing is provided. This search result evaluation apparatus is a vocabulary for storing a lexical hierarchy relation graph indicating a hierarchical relation between vocabularies included in each of a plurality of structured documents stored in the search target data storage means for each structured document. A hierarchical relationship graph storage means; a query creation means for creating a search expression for retrieving a structured document from a plurality of character strings included in the search condition with reference to the vocabulary hierarchical relationship graph storage means; and the creation A search unit that searches the search target data storage unit for a structured document that matches the search expression, and a search result that evaluates the searched structured document based on a plurality of character strings included in the search condition Evaluation means.

  According to the present invention, it is possible to search a structured document in consideration of the hierarchical relationship between the keywords and to evaluate the search results simply by inputting a plurality of keywords.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[First Embodiment]
FIG. 1 is a block diagram mainly showing a functional configuration of a search result evaluation apparatus according to the first embodiment of the present invention. The search result evaluation apparatus 10 executes a search for a structured document according to a search condition including a plurality of character strings specified by a searcher (user) 20, for example.
A search result evaluation apparatus 10 illustrated in FIG. 1 includes a data analysis unit 11, a search execution unit 12, a storage unit 13, and a search result evaluation unit 14.

The data analysis unit 11 creates a vocabulary hierarchy relationship graph in accordance with, for example, a vocabulary hierarchy relationship graph creation request from the administrator 30 of the search result evaluation apparatus 10. The vocabulary hierarchy relationship graph shows a hierarchical relationship between vocabularies included in a structured document to be searched (hereinafter referred to as search target data). Details of this vocabulary hierarchy relationship graph will be described later.
The data analysis unit 11 includes a data analysis control unit 111 and a vocabulary hierarchy relation graph creation unit 112. The data analysis control unit 111 acquires search target data from the storage unit 13 and analyzes the search target data. The lexical hierarchy relationship graph creation unit 112 creates a vocabulary hierarchy relationship graph based on the result analyzed by the data analysis control unit 111. Further, the vocabulary hierarchy relationship graph creation unit 112 registers the created vocabulary hierarchy relationship graph in the storage unit 13.

The search execution unit 12 executes a search process according to a search condition specified by the searcher 20 (search request from the searcher 20). The search execution unit 12 passes the search condition specified by the searcher 20 and the search result (data) that matches the search condition to the search result evaluation unit 14. The search execution unit 12 returns the search result to the searcher 20. The search execution unit 12 includes a query creation unit 121 and a search control unit 122.
The query creation unit 121 creates a list of search conditions specified by the searcher 20 and creates a search expression for searching for search target data from the created list. The search formula is created based on conversion data described later with reference to the vocabulary hierarchy relationship graph storage unit 132. The search condition specified by the searcher 20 is shown in a path format. Hereinafter, the search condition indicated in the path format is referred to as a search condition path expression. The search condition path expression includes a plurality of character strings specified by the searcher 20, for example.
The search control unit 122 performs control for executing search processing on the storage unit 13 based on the search formula created by the query creation unit 121. The search control unit 122 passes the search result data received from the storage unit 13 to the search result evaluation unit 14. The search result data includes search target data searched by the search formula generated by the query generation unit 121, that is, search target data that matches the search conditions specified by the searcher 20.

The storage unit 13 stores various data. The storage unit 13 includes a search target data storage unit 131, a lexical hierarchy relation graph storage unit 132, and a converted data storage unit 133, and stores or retrieves various data as appropriate.
The search target data storage unit 131 stores search target data (structured document) to be searched by the structured document search apparatus 10. The search target data storage unit 131 may be managed (stored) in, for example, an external storage device different from the search result evaluation device 10. The vocabulary hierarchy relationship graph 132 stores (registers) the vocabulary hierarchy relationship graph created by the vocabulary hierarchy relationship graph creation unit 112 of the data analysis unit 11. The conversion data storage unit 133 stores the conversion data described above. This conversion data is data for creating (converting) a search expression from the above-described search condition path expression.

The search result evaluation unit 14 evaluates the search result data passed from the search execution unit 12 based on the search condition (search condition path expression) passed from the search execution unit 12. The search result evaluation unit 14 includes a path type inter-document evaluation unit 141.
The path type inter-document evaluation unit 141 calculates the evaluation value of the search target data for each search target data (structured document) included in the search result data passed from the search execution unit 12. In this case, the path expression inter-document evaluation unit 141 calculates the evaluation value of the structured document based on a location where a plurality of character strings included in the search condition path expression appear in the search target data. The path type inter-document evaluation unit 141 performs ordering (ranking) of search target data included in the search result data based on the calculated evaluation value. The path type inter-document evaluation unit 141 returns the search result data ordered according to the calculated evaluation value to the search execution unit 12.

Next, with reference to the flowchart of FIG. 2, a processing procedure when the lexical hierarchy relation graph is created in the search result evaluation apparatus 10 shown in FIG. 1 will be described.
First, the administrator 30 issues a lexical hierarchy relationship graph creation request for creating a vocabulary hierarchy relationship graph to the data analysis unit 11, for example (step S1). At this time, for example, the administrator 30 may be given a condition such as specifying a vocabulary included in the created vocabulary hierarchy relation graph. In this case, the generated vocabulary hierarchy relation graph may be configured to always hold the vocabulary designated by the administrator 30, for example, or may be configured to preferentially hold the designated vocabulary.

The data analysis control unit 111 of the data analysis unit 11 sends an acquisition request for data to be analyzed (hereinafter referred to as analysis target data) to the storage unit 13 in response to the lexical hierarchy relationship graph creation request from the administrator 30. On the other hand, the data is output (step S2). This analysis target data is search target data stored in the search target data storage unit 131. Hereinafter, search target data processed by the data analysis unit 11 will be described as analysis target data.
In response to this acquisition request, the storage unit 13 searches the analysis target data stored in the search target data storage unit 131 (step S3). The storage unit 13 passes the searched analysis target data to the data analysis control unit 111. Thereby, the data analysis control unit 111 acquires analysis target data.

  The data analysis control unit 111 basically acquires all the search target data stored in the search target data storage unit 131 as analysis target data. However, in step S1, a condition is designated by the administrator 30. In such a case, only the search target data that matches the condition may be acquired as analysis target data. For example, only a difference from the previous analysis, that is, only data that has not been analyzed can be analyzed.

  The data analysis control unit 111 analyzes the acquired analysis target data (step S4). The lexical hierarchy relationship graph creation unit 112 creates a vocabulary hierarchy relationship graph based on the result analyzed by the data analysis control unit 111 (step S5). The vocabulary hierarchy relationship graph creation unit 112 registers the created vocabulary hierarchy relationship graph in the vocabulary hierarchy relationship graph storage unit 132 (step S6).

Next, with reference to the flowchart of FIG. 3, the processing procedure when the search according to the search request from the searcher 20 is executed in the search result evaluation apparatus 10 shown in FIG.
First, the searcher 20 issues a search request to the search execution unit 12 by specifying, for example, a search condition path expression (step S11). This search condition path expression includes a plurality of character strings.

  The query creation unit 121 creates a search formula for executing a search on the search target data storage unit 131 according to the input search condition path formula (step S12). That is, the query creation unit 121 refers to the vocabulary hierarchy relationship graph storage unit 132 and the conversion data storage unit 133, and the query creation unit 121 indicates the plurality of lexical hierarchy relationship graphs including a plurality of character strings included in the search condition path expression as vocabulary. Create a search expression based on the hierarchical relationship between the strings.

The search control unit 122 of the search execution unit 12 designates the search formula created by the query creation unit 121 and outputs a search request to the storage unit 13 (step S13).
In response to the search request from the search control unit 122, the storage unit 13 searches the search target data storage unit 131 for search target data that matches the search request under the control of the search control unit 122 (step S14). . The storage unit 13 returns the search result (data) including the search target data that matches the search request from the search control unit 122 to the search control unit 122. Thereby, the search control unit 122 acquires search result data based on the search formula. The search result data includes search target data (structured document) that matches the search request searched by the storage unit 13.

  The search control unit 122 passes the acquired search result data and the search condition path expression to the search result evaluation unit 14. The search result evaluation unit 14 evaluates each search target data included in the search result data based on the search condition path expression passed from the search control unit 122 (step S15). At this time, the path expression inter-document evaluation section 141 of the search result evaluation section 14 performs the search based on the location where a plurality of character strings included in the search condition path expression appear in the search target data included in the search result data. The evaluation value of the target data is calculated. Details of the evaluation value calculation process will be described later.

The path type inter-document evaluation unit 141 performs ordering for each of the search target data included in the search result data based on the calculated evaluation value. The search result data including the search target data ordered by the path type inter-document evaluation unit 141 is passed to the search execution unit 12.
The search control unit 122 returns the search result data passed from the path type inter-document evaluation unit 141 to the searcher 20 as a search result for the search request (step S16). Thereby, the searcher 20 acquires a search result for the search request (step S17). In this case, the search results are presented (displayed) to the searcher 20 in order from the search target data at the top of the results ordered by the path type inter-document evaluation unit 141.

FIG. 4 is a diagram illustrating an example of search target data stored in the search target data storage unit 131. The search target data shown in FIG. 4 is, for example, weather forecast data 100 described in XML (Extensible Markup Language) format. The weather forecast data 100 shown in FIG. 4 has, for example, “weather forecast” as a route element name. The weather forecast data 100 has, for example, “location”, “forecast”, “weather”, “temperature”, “highest”, “lowest”, and “precipitation probability” as element names.
Furthermore, the weather forecast data 100 has, for example, an attribute name “prefecture” and attribute values “Tokyo”, “Kanagawa”, and “Saitama”. Also, for example, the attribute name is “Region”, and the attribute values are “Tokyo region”, “Northern Izu Islands”, “Southern Izu Islands”, “Ogasawara Islands”, “Eastern”, “Western”, “Northern” and “Southern” Is included. For example, the attribute name is “unit” and the attribute value has “%”.

Here, there are a simple type and a composite type in the elements included in the search target data. The simple type is an element having no attribute and having only text (character string) in the content. In the weather forecast data 100, for example, element names “weather”, “highest”, and “lowest” are simple types.
The composite type includes a composite type having a simple type content and a composite type having no simple type content. A complex type having simple type content is an element having an attribute and having only text in the content. In the weather forecast data 100, for example, the element name “precipitation probability” can be mentioned. On the other hand, a composite type having no simple type content is a composite type having contents other than text. In the weather forecast data 100, for example, the element name “location” and “temperature” are listed.

Next, with reference to FIG. 5 and FIG. 7, the processing of steps S4 and S5 of FIG. 2 described above will be described in detail. First, a processing procedure for creating a structure graph indicating the hierarchical structure of the analysis target data (search target data) will be described with reference to the flowchart of FIG. This structure graph is used to create a lexical hierarchy relationship graph.
The data analysis unit 11 initializes the structure graph (step S21). With this processing, the data analysis unit 11 converts the structure graph into a graph having no vertices and no edges.

Next, the data analysis control unit 111 of the data analysis unit 11 acquires the root element of the analysis target data searched in step S3 of FIG. 2 described above as a target element (step S22). The root element is an element that includes all elements of a structured document (XML data), for example.
The data analysis control unit 111 acquires the element name of the acquired target element (step S23), and determines whether an attribute is specified for the target element (step S24). Further, if it is determined that an attribute is specified for the target element (YES in step S24), the attribute name and attribute value of the attribute specified for the target element are acquired (step S25). Morphological analysis is performed on the acquired attribute value, and nouns included in the attribute value are extracted based on the result (step S26).
If it is determined in step S24 that no attribute is specified for the target element, the processes in steps S25 and S26 described above are not executed.

Next, the data analysis control unit 111 determines whether the target element is the above-described simple type or a complex type having simple type contents (step S27). When it is determined that the target element is a simple type or a complex type having simple type contents (YES in step S27), the contents of the target element are acquired (step S28). Morphological analysis is performed on the acquired content of the target element, and nouns included in the content of the target element are extracted based on the result (step S29).
If it is determined in step S27 that the target element is not a simple type or a complex type having simple type contents, the processes in steps S28 and S29 described above are not executed.

  Next, the vocabulary hierarchy relationship graph creation unit 112 performs scoring on the acquired element name, attribute name, or noun (vocabulary) extracted in step S26 or S28 (step S30). The vocabulary hierarchy relationship graph creation unit 112 executes scoring processing using, for example, evaluation values.

  Here, an example of the evaluation value will be specifically described. The lexical hierarchy relationship graph creating unit 112 executes the scoring process using a plurality of evaluation values among the following first evaluation value to seventh evaluation value, for example. First, an example of an evaluation value in the case of an element name or attribute name will be described.

The first evaluation value is calculated according to the number of occurrences of element names or attribute names in the analysis target data, for example. For example, the evaluation value is high for those having a large number of appearances.
The second evaluation value is calculated according to the appearance position. In this case, the evaluation value is smaller as the appearance position (hierarchy) is deeper. For example, when the depth of the hierarchy is depth, the second evaluation value is calculated by the evaluation formula 1 / depth or 1 / (log depth) +1. The depth of the hierarchy is 1 for the root element.
For example, if the vocabulary specified as a condition by the administrator 30 is an element name, the element name is the specified vocabulary for the depth of elements or attributes below the element whose element name is the specified vocabulary. The depth of the element may be treated as 0, and the above-described second evaluation value may be obtained based on this.

The third evaluation value is calculated according to the child node (child element, attribute). In this case, the evaluation value increases as the number of child nodes increases. For example, if the number of child nodes is count, the third evaluation value is calculated by the evaluation formula count / total number of nodes. Note that a node is a generic term for elements, element contents, attribute names, and attribute values.
The fourth evaluation value is calculated according to whether it is an element name or an attribute name. For example, the coefficient used when calculating the evaluation value is changed depending on the element name or attribute name.
The fifth evaluation value is calculated according to the text content related to the element name or attribute name. For example, when the vocabulary specified as a condition by the administrator 30 is included in the element content or attribute value, the element name or attribute value of the element including the specified vocabulary in the element content is specified. The evaluation value for the attribute name of the attribute including the vocabulary increases.
In the case of an element name or an attribute name (a vocabulary included in the element name), for example, scoring is performed by the sum of a plurality of evaluation values including the first to fifth evaluation values.

Next, examples of evaluation values in the case of element contents or attribute values will be described. The sixth evaluation value is calculated according to the number of appearances of the acquired noun or proper noun as a result of the morphological analysis. For example, the evaluation value is large for a large number of appearances.
The seventh evaluation value is calculated according to the node name related to the content of the element or the attribute value. Here, the node name indicates an element name or an attribute name. For example, when the vocabulary designated as a condition by the administrator 30 is an element name or an attribute name, the element name or the attribute value of the designated vocabulary attribute value becomes larger as the element name or attribute name of the designated element. In the case of the content or attribute value of an element, for example, it is scored by the sum of a plurality of evaluation values including the above-described sixth or seventh evaluation value.

Next, the vocabulary hierarchy relationship graph creation unit 112 has the acquired element name, attribute name, or vocabulary extracted in step S26 or S28 as a label (name), and adds each to the structure graph as a vertex holding the node type. (Step S31). The node type indicates an element name, element content, attribute name or attribute value.
The vocabulary hierarchy relationship graph creation unit 112 adds an edge to the structure graph according to the definition (step S32). An edge represents an element name, an attribute name on the structure graph, or a hierarchical relationship between the vocabularies extracted in step S26 or S28. Here, the above definition includes, for example, “parent element name to child element name”, “element name to attribute name”, “attribute name to attribute value”, “element name to attribute name” depending on the direction of the arrow (side). “Character string included in element content”, “attribute name to child element name from parent element”, “character string to child element name included in attribute value of parent element”, and the like.

  The vocabulary hierarchy relationship graph creation unit 112 determines whether the target element has a child element based on the analysis target data (step S33). When it is determined that the target element has a child element (YES in step S33), the data analysis control unit 111 acquires the child element (step S34), and uses the acquired child element as the target element. Returning to step S23, the process is repeated.

  On the other hand, if it is determined in step S33 that the target element has no child elements, the vocabulary hierarchy relationship graph creating unit 112 extracts the vocabulary based on the result scored in step S30 (step S35). At this time, the vocabulary hierarchy relation graph creating unit 112 has a score (total evaluation value) of vocabularies (here, the acquired element name, attribute name, or vocabulary extracted in step S26 or S28). Extract vocabulary above the value. The predetermined value may be set in advance, or may be set by the administrator 30 for each process. Further, the vocabulary hierarchy relationship graph creating unit 112 may be configured to extract, for example, vocabularies having the number of words designated by the administrator 30 in descending order of evaluation values.

  The lexical hierarchy relation graph 112 deletes the vertex having the vocabulary that is not the vocabulary extracted in step S35 but also the vocabulary designated by the administrator 30 and the edge adjacent to the vertex (step S36). The graph for which the deletion process has been completed in step S36 becomes a structure graph indicating the hierarchical structure of the analysis target data.

  FIG. 6 shows an example of the structure graph 200 of the weather forecast data 100 created by the process shown in the flowchart of FIG. It is assumed that the structure graph 200 is created, for example, by the administrator 30 with the word “weather” specified as a condition.

  As shown in FIG. 6, the structure graph 200 shows “location”, “forecast”, “weather”, “temperature”, and “precipitation probability” as element names. In the structure graph 200, “prefecture” is shown as an attribute name, and “Tokyo”, “Kanagawa”, and “Saitama” are shown as character strings included in the attribute value. In addition, the structure graph 200 shows “clear”, “cloudy”, and “rain” as character strings included in the content of the element whose element name is “weather”. In FIG. 6, in the notation of the graph, the notation of the vertex is changed depending on the node type on XML. In the structure graph 200, for example, an edge is shown from the element name “place” to the element name “forecast”. This side is added according to the definition as described above.

  Of the element names, attribute names, attribute values, or element contents of the weather forecast data 100 shown in FIG. 4, vocabularies not shown in the structure graph 200 are deleted in step S36 of FIG. Shall. The deleted vocabulary is a vocabulary whose evaluation value is smaller than a predetermined value as a result of scoring and is not specified by the administrator 30.

Next, a processing procedure for creating a lexical hierarchy relationship graph using a structure graph will be described with reference to the flowchart of FIG. Note that the structure graph used here is assumed to be the structure graph 200 shown in FIG.
First, the lexical hierarchy relationship graph creation unit 112 refers to the created structure graph 200 and adds a vertex having the same label as the vertex shown in the structure graph 200 to the vocabulary hierarchy relationship graph (step S41). At this time, if there are a plurality of vertices having the same label in the structure graph 200, the vertices are added so that the vertices do not overlap.

  The lexical hierarchy relationship graph creation unit 112 refers to the structure graph 200 and acquires any two vertices within a specific distance (step S42). Here, when the distance of one layer on the structure graph 200 is 1, the specific distance is preferably 2 or more. The specific distance may be specified in advance by the administrator 30, for example, or may be specified by the administrator 30 for each process.

The lexical hierarchy relationship graph 112 adds an edge between the two vertices on the vocabulary hierarchy relationship graph corresponding to the obtained arbitrary two vertices (step S43). Here, the added side has a label. The label of the edge indicates a set of node types on the structure graph 200 of any two obtained vertices.
When updating the vocabulary hierarchy relationship graph storage unit 132, the vocabulary hierarchy relationship graph creation unit 112 initializes the vocabulary hierarchy relationship graph and then executes the processing shown in the flowchart of FIG.

FIG. 8 is a schematic diagram of the lexical hierarchy relation graph created by the process shown in FIG. A vocabulary hierarchy relationship graph 300 shown in FIG. 8 is a vocabulary hierarchy relationship graph created by setting the specific distance to 2 as described above. FIG. 8 shows only a portion centered on a vertex (hereinafter simply referred to as “Tokyo”) having “Tokyo” as a label for convenience. Sides between vertices having labels other than “Tokyo” are indicated by dotted lines, and labels of the sides are omitted. For vertices other than “Tokyo”, for example, a part of the side is omitted.
In the structure graph 200 shown in FIG. 6 above, the vertices existing within a distance of 2 from “Tokyo” are “location”, “prefecture”, “forecast”, “weather”, “temperature”, and “precipitation probability”. It is. Hereinafter, this vertex is referred to as a vertex within a specific distance.

  Further, in the lexical hierarchy relation graph 300 shown in FIG. 8, “Tokyo” and each vertex within a specific distance are connected by an edge, and each edge is labeled. For example, in “place” and “Tokyo”, an edge is connected from “place” to “Tokyo”, and the label of the edge is “element name, attribute value”. This indicates that the node type of “place” is an element name and the node type of “Tokyo” is an attribute value. Although omitted in the lexical hierarchy relationship graph 300 shown in FIG. 8, vertices other than “Tokyo” are similarly connected to vertices within a specific distance by edges, and each of the edges has a label. It is attached.

Next, the process of step S12 shown in FIG. 3 will be described in detail with reference to the flowchart of FIG.
The query creation unit 121 of the search execution unit 12 inputs the search condition path expression designated by the searcher 20 in the process of step S11 shown in FIG. Hereinafter, the search condition path expression input by the query creation unit 121 is referred to as an input path expression. Here, the input path expression will be described as being “Tokyo / weather”, for example.

  Note that the permutation of the character string (vocabulary) indicated by the input path expression indicates a hierarchical relationship according to the order of the character string of the permutation. As described above, when the input path expression is “Tokyo / weather”, it indicates that there is a hierarchical relationship from “Tokyo” to “weather”. Hereinafter, this hierarchical relationship is referred to as an input path type hierarchical relationship.

The query creation unit 121 refers to the vocabulary hierarchy relationship graph storage unit 132 and acquires a vocabulary hierarchy relationship graph including the input path expression hierarchy relationship as, for example, a parent-child relationship (step S51). For example, when the vocabulary hierarchy relationship graph including all of the input path expression hierarchical relationships as parent-child relationships does not exist in the vocabulary hierarchy relationship graph storage unit 132, the query creation unit 121 determines whether the input path expression hierarchical relationship A lexical hierarchy relation graph including a part as a parent-child relation is acquired.
For example, when the input path expression is A / B / C (the input path expression has a hierarchical relationship from A to B, and from B to C), the lexical hierarchy relationship graph that completely includes the hierarchical relationship is the lexical hierarchy relationship graph. Assume that the storage unit 132 does not exist. In this case, for example, the query creation unit 121 acquires a lexical hierarchy relationship graph including only the hierarchical relationship from A to B as the parent-child relationship, or a vocabulary hierarchical relationship graph having only the hierarchical relationship from B to C as the parent-child relationship.

The query creation unit 121 determines the node type of each character string included in the input path expression based on the acquired lexical hierarchy relation graph (step S52).
Based on the hierarchical relationship of the input path expression and the determined node type, the query creating unit 121 searches the search target data having a hierarchical relationship between character strings included in the input path expression, for example, XPath Create an expression. That is, the query creation unit 121 converts the input path expression into an XPath expression (step S53). Alternatively, the query creation unit 121 converts it into an XPath expression based on the conversion data stored in the conversion data storage unit 133.

  In step S51, when a plurality of vocabulary hierarchy relationship graphs are acquired by the query creation unit 121, the processing of steps S52 and S53 is executed for each of the acquired vocabulary hierarchy relationship graphs. Thus, when a plurality of XPath expressions are created (converted), the search processing is performed on the search target data storage unit 131 with the plurality of XPath expressions being ORed (ORed) with each other.

  Here, FIG. 10 shows an example of the data structure of the conversion data stored in the conversion data storage unit 133. This conversion data is defined in advance as a conversion method to an XPath expression for each relationship in the lexical hierarchy relationship graph. As illustrated in FIG. 10, the conversion data includes relationship information and XPath expression information associated with the information.

  The relationship information indicates a hierarchical relationship between vocabularies (character strings). In the example shown in FIG. 10, for example, there is a parent-child (element name, element name) relationship. This indicates that each node type of the vocabulary is an element name. Similarly, in the parent and child (element name, element content), the node type of the parent vocabulary is the element name, and the node type of a different vocabulary (child) is the element content (character string included in the element). It shows that. For example, the relationship between siblings (element names, element names,...) Indicates that the relationships between vocabularies are the same hierarchy and the same node type.

  Further, the XPath expression information indicates an XPath expression suitable for searching for search target data (structured document) that matches the vocabulary relationship indicated by the information indicating the associated relationship. For example, the XPath expression information associated with the parent-child (element name, element content) relation information indicates the XPath expression “//element1[.//elemnt2]”. Similarly, for example, XPath expression information associated with parent-child (element name, element content) relation information indicates the XPath expression “//element1[./text()=“context”] ”. In the above XPath expression, element1 and element2 indicate element names, and context indicates the content of the element. In addition, attri indicates an attribute name, and value indicates an attribute value.

  Note that the relationship information included in the conversion data shown in FIG. 10 defines a hierarchical relationship between two vocabularies, but it is also possible to define a hierarchical relationship between three or more vocabularies by combining a plurality of these. It is. Thus, an input path expression including three or more vocabularies can be converted into an XPath expression.

  Returning to step S53 in FIG. 9 again, the query creation unit 121 identifies compatible relationship information from the node type determined in step S52 and the hierarchical relationship of the input path expressions. The query creation unit 121 converts it into an XPath expression based on the XPath expression information associated with the specified relationship information.

For example, the case where the input path expression is “Tokyo / weather” as described above and the lexical hierarchy relation graph 300 shown in FIG. 8 is acquired in step S51 will be described. In this case, in the vocabulary hierarchy relationship graph 300, “Tokyo” is an attribute value, and “weather” is an element name. Further, since “Tokyo” is connected to “weather”, “Tokyo” and “weather” have a parent-child relationship. Therefore, the query creation unit 121 creates an XPath expression indicated by the XPath expression information corresponding to the relationship information indicating the parent and child (attribute value, element name) based on, for example, the conversion data illustrated in FIG. That is, the query creation unit 121 converts the input path expression “Tokyo / weather” into an XPath expression “//*[./@*=“Tokyo” and.//weather] ”. Based on this XPath expression, a search process is executed for the search target data storage unit 131. XPath expression "//*[./@*="Tokyo"and.//weather]" has parent-child relationship between "Tokyo" and "weather", and the node type of "Tokyo" is an attribute This is a search formula for searching XML data (structured document) that is a value and whose element type is the node type of “weather”. It should be noted that “Tokyo” and “weather” do not need to have a direct parent-child relationship, and can be searched for, for example, a grandchild or great-grandchild relationship.
Here, the conversion data includes XPath expression information, and an example in which an input path expression is converted into an XPath expression has been shown. However, for example, the conversion data may be converted into an arbitrary query language for an arbitrary data source such as XQuery or SQL.

  FIG. 11 shows a search result based on the XPath expression “//*[./@*=“Tokyo” and.//weather] ”created by the query creation unit 121 for the weather forecast data 100 of FIG. 400 is shown. The search result 400 is XML data including an attribute value “Tokyo” and an element name “weather” that is a parent-child relationship (here, a grandchild relationship).

  By the way, for example, when there are a plurality of input path type hierarchical relationships, a hierarchical relationship that is not set in the vocabulary hierarchical relationship graph in the vocabulary hierarchical relationship graph storage unit 132 (unset hierarchical relationship) among the plurality of hierarchical relationships. Assume that is specified. In this case, the configuration (setting level) for the unset hierarchy relationship may be defined in advance. By setting the level in advance, a desired search result range can be obtained.

  FIG. 12 shows an example of a preset setting level. In the example of FIG. 12, levels 1 to 3 are defined for the node type set for the two vocabularies constituting the unset hierarchy relationship and the hierarchy relation set for the two vocabularies constituting the unset hierarchy relation. The level stage is set as appropriate by the searcher 20, for example.

For example, at level 1, it is defined that the unset hierarchical relationship included in the input path expression is ignored. For example, assume that the input path expression is A / B / C. In this case, in one vocabulary hierarchy relationship graph stored in the vocabulary hierarchy relationship graph storage unit 132, there is a hierarchy relationship from A to B, but there is no hierarchy relationship from B to C. This indicates that the hierarchical relationship is ignored and only the hierarchical relationship from A to B is converted into an XPath expression.
For example, at level 2, the type indicates “element name”, and the operation content indicates “set unset hierarchical relationship as parent-child relationship”. That is, the node types of two vocabularies that constitute an unset hierarchical relationship are set as element names, and the input path expression is converted into an XPath expression with the hierarchical relationship between the two vocabularies as a parent-child relationship.
Further, for example, at level 3, the type indicates “all” and the operation content indicates “set unset hierarchical relationship to parent-child relationship or sibling relationship”. That is, the node type of each of the two vocabularies constituting the unset hierarchical relationship is set to the element name, attribute name, attribute value, or element content. In addition, for every set combination, a parent-child relationship or a sibling relationship is set as the hierarchical relationship between vocabularies. Based on all set combinations and hierarchical relationships, the input path expression is converted to an XPath expression.

  Next, with reference to FIGS. 13 and 14, the search result evaluation process (the process of step S15 shown in FIG. 3 described above) in the search result evaluation apparatus 10 will be specifically described. Here, it is assumed that the input path expression (search condition path expression) is “invention / (Goda and structured document)”. This input path expression represents, for example, the meaning of “invented by Goda and including a structured document”. Further, from this input path expression, for example, an XPath expression such as “// invention [.//text()=“Goda” and .//text()=“structured document ”]” is used as the query creation unit 121. Created by.

  The search retrieved from the search target data storage unit 131 in accordance with the above XPath expression “// invention [.//text()=“Goda” and .//text()=“structured document ”]” As a result, the search target data 101 shown in FIG. 13 and the search target data 102 shown in FIG. 14 are obtained.

Hereinafter, for each of the search target data 101 and the search target data 102 obtained as a search result, the path expression inter-document evaluation unit 141 is based on an input path expression (here, “invention / (Aida and structured document)”). An example of an algorithm for calculating an evaluation value (hereinafter referred to as a first algorithm) will be described.
First, the path expression inter-document evaluation unit 141 performs the search in each of the search target data 101 and the search target data 102 obtained as a search result for all keywords (character strings) having a hierarchical relationship in the input path expression, for example. Calculate the distance. Here, the distance between keywords is, for example, that the distance is 1 from an element to its child element and from the element to its attribute name. Similarly, the distance is set to 1 from the element to the content of the element and from the attribute name to the attribute value.

Next, the path type inter-document evaluation unit 141 obtains the reciprocal of the calculated distance for each search target data obtained as a search result, and calculates the sum of them as an evaluation value of the search target data.
Here, processing for calculating the evaluation values of the search target data 101 shown in FIG. 13 and the search target data 102 shown in FIG. 14 will be specifically described.

  Since the input path expression is “invention / (Goda and structured document)”, the hierarchical relationship (keywords) in the input path expression is “invention → Goda” and “invention → structured document”.

  The distance corresponding to “invention → Goda” in the search target data 101 is calculated. In this case, the distance from the “invention” that is the element of the search target data 101 to the “inventor” that is the child element thereof is 1, and the distance from the “inventor” that is the element to the content of the element is “Goda”. 1. Therefore, the distance corresponding to “invention → Goda” in the search target data 101 is calculated as 2.

Further, a distance corresponding to “invention → structured document” in the search target data 101 is calculated. In this case, the distance from the “invention” that is the element of the search target data 101 to the “name” that is the child element is 1, and the distance from the “name” that is the element to the “structured document” that is the content of the element is 1. Therefore, the distance corresponding to “invention → structured document” in the search target data 101 is calculated as 2.
As described above, since the evaluation value is the sum of the reciprocal of the distance between the keywords, the evaluation value of the search target data 101 is 1/2 + 1/2 = 1.

  On the other hand, the distance corresponding to “invention → Goda” in the search object data 102 is calculated. In this case, the distance from the “invention” that is the element of the search target data 102 to the “inventor” that is the child element is 1, and the distance from the “inventor” that is the element to the “name” that is the child element is 1. Yes, the distance from the element "name" to the element content "Goda" is 1. Therefore, the distance corresponding to “invention → Goda” in the search target data 102 is calculated as 3.

In addition, the distance corresponding to “invention → structured document” in the search target data 102 is calculated. In this case, the distance from “invention” that is an element of the search target data 102 to “content” that is the child element is 1, and the distance from “content” that is the element to “name” that is the child element is 1. The distance from the “name” that is an element to the “structured document” that is the content of the element is 1. Therefore, the distance corresponding to “invention → structured document” in the search target data 102 is calculated as 3.
As described above, since the evaluation value is the sum of the reciprocals of the distances between the keywords, the evaluation value of the search target data 102 is 1/3 + 1/3 = 2/3.

The path type inter-document evaluation unit 141 ranks the search target data 101 and the search target data 102 based on the calculated evaluation values of the search target data 101 and the search target data 102. The path type inter-document evaluation unit 141 ranks, for example, search target data (in this case, the search target data 101) having a large calculated evaluation value in a higher rank. Accordingly, in this case, search results are presented to the searcher 20 in the order of the search target data 101 and the search target data 102.
Note that the first algorithm described above is merely an example, and for example, the method for defining the distance between keywords or calculating the evaluation value may be other than that described here.

  As described above, in the present embodiment, the search condition (search condition path expression) in the path format including a plurality of character strings (vocabulary) designated by the searcher is stored in the vocabulary hierarchy relation graph storage unit 132. XPath for searching a structured document (search target data) that matches the hierarchical relationship between character strings included in the search condition based on the lexical hierarchy relationship graph and the conversion data stored in the conversion data storage unit 133 An expression is created. Based on the created XPath expression, the search target data storage unit 131 is searched. As a result, the searcher 20 can perform a search in consideration of the hierarchical relationship of the structured document, for example, by simply specifying a keyword (character string) without being aware of the data structure of the structured document to be searched. It becomes possible to execute.

  In the present embodiment, the search result searched by the above XPath expression is evaluated based on the search condition specified by the searcher 20, so that the structured document having the hierarchical structure intended by the searcher 20 is obtained. Can be ranked higher and return search results. As a result, the searcher 20 can efficiently acquire data (structured document) required by the searcher 20.

  In the present embodiment, a lexical hierarchy relation graph is created on the structure graph indicating the hierarchical structure of the structured document, for example, based on a specific distance between two arbitrary vertices. Thus, when a search process is executed using this vocabulary hierarchy relation graph, even if the vocabulary specified by the searcher 20 is not directly in a hierarchical relation on the structured document to be searched, It becomes possible to search in consideration.

  Further, in the present embodiment, by setting and specifying a pre-defined level, even if the unset hierarchy relationship is specified as an input path expression, the non-set level is set according to the set level. Hierarchical relationships can be set statically or dynamically.

  In the above-described embodiment, the searcher 20 issues a search request to the search execution unit 12, and the administrator 30 issues a lexical hierarchy relationship graph creation request to the data analysis unit 11. A configuration may be provided in which a request control unit for inputting a request is provided. In this case, the request control unit can interpret the input request and automatically distribute the processing to the data analysis unit 11 or the search execution unit 12.

  Moreover, although the search result evaluation apparatus 10 according to the present embodiment has been described as being a search apparatus as described above, the search result evaluation apparatus 10 may have a function of registering search target data in the search target data storage unit 131. . Further, the search result evaluation device 10 may be configured to be used as a management device including a function of updating or deleting search target data.

[Second Embodiment]
Next, a search result evaluation apparatus according to the second embodiment of the present invention will be described with reference to FIG. FIG. 15 is a block diagram mainly showing a functional configuration of the search result evaluation apparatus according to the present embodiment. The same parts as those in FIG. 1 described above are denoted by the same reference numerals, and detailed description thereof is omitted. Here, parts different from FIG. 1 will be mainly described.

  The search result evaluation device 40 illustrated in FIG. 15 includes a search execution unit 41 and a search result evaluation unit 42. The search execution unit 41 includes a query creation unit 411. For example, the query creation unit 411 executes processing for expanding a search condition path expression designated by the searcher 20 into a plurality of search condition path expressions. At this time, the query creation unit 411 expands the search condition path expression based on a plurality of character strings included in the search condition path expression specified by the searcher 20. For example, the query creation unit 411 expands a plurality of search condition path expressions by rearranging a plurality of character strings included in the search condition path expression designated by the searcher 20. Thereby, the query creation unit 411 creates a plurality of search condition path expressions from the search condition path expressions designated by the searcher 20. In addition, the query creation unit 411 refers to the vocabulary hierarchy relation graph storage unit 132 and the conversion data storage unit 133, and uses the search condition path expression specified by the searcher 20 and the search expression ( For example, XPath expression).

  The search result evaluation unit 42 includes an inter-path expression evaluation unit 421. The inter-path-expression evaluation unit 421 evaluates each of the plurality of search condition path expressions developed by the query creation unit 411 based on, for example, the search condition path expression specified by the searcher 20. In this case, the inter-path expression evaluation unit 421 is included in each of the plurality of search condition path expressions developed by the query creation unit 411 and the arrangement order of the plurality of character strings included in the search condition path expression designated by the searcher. Each evaluation value of the developed plurality of search condition path expressions is calculated based on the order of arrangement of the plurality of character strings. The inter-path expression evaluation unit 421 orders a plurality of expanded search condition path expressions based on the calculated evaluation value.

Next, a processing procedure when a search is executed in response to a search request from the searcher 20 of the search result evaluation device 40 shown in FIG. 15 will be described with reference to the flowchart of FIG.
First, the searcher 20 designates a search condition path expression and issues a search request (step S61). This search condition path expression includes a plurality of character strings (vocabulary). Here, the search condition path expression will be described as “weather / Tokyo”.

The query creation unit 411 of the search execution unit 41 inputs the search condition path expression designated by the searcher 20. Hereinafter, the search condition path expression input by the query creation unit 411 is referred to as an input path expression.
Next, the query creation unit 411 expands the input path expression (step SS62). The query creation unit 411 expands the input path expression into a plurality of search condition path expressions by rearranging the character strings included in the input path expression. At this time, the query creation unit 411 expands every permutation (arrangement order) of all character strings included in the search condition path expression. Specifically, if the input path expression is “weather / Tokyo”, the input path expression is expanded into a plurality of search condition path expressions such as “weather / Tokyo” and “Tokyo / weather”. The Hereinafter, each of the plurality of search condition path expressions expanded by the query creation unit 411 is referred to as an expanded path expression.

  Note that the permutation of the character strings indicated by each of the expansion path expressions indicates the hierarchical relationship according to the order of the character strings in the permutation. For example, if the development path expression is “weather / Tokyo”, it indicates that there is a hierarchical relationship from “weather” to “Tokyo”. Hereinafter, this hierarchical relationship is referred to as an expanded path type hierarchical relationship.

  Further, in the above-described step S62, the case where expansion is performed for each permutation of all character strings included in the input path expression has been described. For example, referring to the vocabulary hierarchy relation graph storage unit 132, the expansion path expression Only a development path expression in which a vocabulary hierarchy relation graph including a hierarchy relation exists may be developed. As a result, the number of development path expressions processed in the following steps is reduced, and the amount of processing can be reduced.

  Further, a configuration may be adopted in which a character string included in the input path expression is subjected to morphological analysis, and the input path expression is expanded by a character string extracted as a result. In this case, for example, when the input path expression is “weather / cloudy sometimes sunny”, “morphely is sometimes cloudy” is analyzed, and the developed path expressions are “weather / cloudy / sunny”, “weather / cloudy” and “weather / cloudy”. It becomes possible to develop as “sunny”.

  In addition, the expansion of the input path expression is not only based on the simple rearrangement of the character strings included in the input path expression as described above, but also by a logical operation such as “and”, “or” or “not”. Deployment is also included. That is, when the input path expression is “A / B / C”, for example, it can be expanded to “A / (B and C)”.

  Next, the query creation unit 411 performs scoring for each expansion path expression (step S63). For example, scoring processing is executed using the evaluation value. Further, the query creation unit 411 calculates an evaluation value using the vocabulary hierarchy relationship graph stored in the vocabulary hierarchy relationship graph storage unit 132 as necessary.

Here, an example of evaluation values used in the above scoring process will be specifically described. The query creation unit 411 executes the scoring process using a plurality of evaluation values including the following eighth evaluation value to tenth evaluation value, for example.
The eighth evaluation value is calculated according to the difference in the order of the character strings included in both of the input path expression and the expansion path expression. In this case, compared with the order of the character strings included in the input path expression, the evaluation value is smaller for the expanded path expression including many character strings having different orders. For example, if x = the number of character strings in the order different from the input path expression, and X = the number of character strings included in the input path expression or the expansion path expression, for example, the eighth evaluation value by the evaluation expression X / (log ax) + N Is calculated. However, in the evaluation formula X / (log ax) + X, when there is no character string having a different order from the input path formula, that is, when x = 0, the evaluation value 1 is calculated. In the evaluation formula, “a” is the bottom and can be arbitrarily set. The same applies to the following evaluation values.

  The ninth evaluation value is calculated according to the difference in the type of character string (vocabulary) between the input path expression and the expansion path expression. In this case, the evaluation value is small for an expanded path expression that includes many vocabulary types different from the vocabulary types included in the input path expression. Here, for example, when the vocabulary included in the expanded path expression is not the vocabulary itself included in the input path expression but a vocabulary extracted by morphological analysis, the vocabulary types are treated as different. For example, if y = the number of vocabulary types different from the vocabulary type included in the input path expression, and Y = the number of vocabularies included in the input path expression or the expansion path expression, An evaluation value of 9 is calculated. However, when there is no vocabulary of a different type from the vocabulary included in the input path expression in the evaluation expression Y / (log a y) + Y, that is, when y = 0, the evaluation value 1 is calculated.

  The tenth evaluation value is calculated according to the relationship between adjacent character strings (vocabulary) in the development path expression. In this case, for example, when the number of hierarchical relationships that are not set as parent-child relationships is large in each of the hierarchical relationships of the development path expression, the evaluation value becomes small. For example, if Z = the number of hierarchical relationships that are not set as parent-child relationships, the tenth evaluation value is calculated by, for example, the evaluation formula 1 / (log a Z) +1. However, in the evaluation formula 1 / (log a Z) +1, when there is no hierarchical relationship that is not set, that is, when Z = 0, the evaluation value 1 is calculated. The tenth evaluation value is obtained for each of the vocabulary hierarchy relationship graphs, for example, by obtaining the evaluation value with the above-described evaluation formula for each of the vocabulary hierarchy relationship graphs stored in the vocabulary hierarchy relationship graph storage unit 132. It is obtained by summing the evaluation values.

  Each of the development path expressions is scored by, for example, the sum of a plurality of evaluation values including the above-described eighth to tenth evaluation values. Based on the scored result, the query creation unit 411 selects an expanded path expression having a higher score (total evaluation value) (step S64). Hereinafter, an expanded path expression having a higher score selected by the query creation unit 411 is referred to as an intermediate path expression. There may be a plurality of intermediate path expressions.

The search execution unit 41 (query creation unit 411) passes the input path expression and intermediate path expression to the search result evaluation unit.
The inter-path expression evaluation unit 421 of the search result evaluation unit 42 acquires the input path expression and the intermediate path expression passed from the search execution unit 41. The inter-path expression evaluation unit 421 evaluates the acquired intermediate path expressions (each) based on the acquired input path expression (step S65). In this case, the inter-path expression evaluation unit 421 calculates the evaluation value of the intermediate path expression based on the arrangement order of the plurality of character strings included in the input path expression and the arrangement order of the plurality of character strings included in the intermediate path expression. Then, the intermediate path formulas are ordered based on the calculated evaluation values. The inter-path expression evaluation unit 421 returns the ordered intermediate path expressions to the search execution unit 41.

  Here, the process of calculating the evaluation value of the intermediate path expression by the inter-path expression evaluation unit 421 will be specifically described. Here, the input path expression will be described as “Goda / invention / structured document”. This input path expression represents a meaning such as “an invention made by Goda and including a structured document”. In addition, the query creation unit 411 creates intermediate path expressions “invention / (Goda and structured document)” and “Goda / invention / structured document” from the input path expression “Gaida / invention / structured document”. To do. Hereinafter, the intermediate path expression “invention / (Gita and structured document)” is referred to as a first intermediate path expression, and the intermediate path expression “Goda / invention / structured document” is referred to as a second intermediate path expression.

  For example, depending on the result of scoring in step S63, the input path expression and the second intermediate path expression may be the same as described above. In this case, the input path expression and the second intermediate path expression may be handled as one search condition path expression (for example, an input path expression).

Hereinafter, an example of an algorithm (hereinafter referred to as a second algorithm) in which the inter-path expression evaluation unit 421 calculates evaluation values of the first intermediate path expression and the second intermediate path expression will be described.
First, the inter-path expression evaluation unit 421, for example, the number of combinations of the same keyword having the same hierarchical relationship in the input path expression and the intermediate path expression (here, the first intermediate path expression and the second intermediate path expression). (Hereinafter referred to as the first combination number) is calculated.
Here, the hierarchical relationship means a relationship between keywords separated by “/” in an input path expression or an intermediate path expression. Also, it is defined that there is a hierarchical relationship between keywords separated by a plurality of “/”. For example, when the input path expression (intermediate path expression) is “A / B / C”, the input path expression includes a hierarchical relationship of “A → B”, “A → C”, and “B → C”. It is included.

Next, the inter-path expression evaluation unit 421 calculates the number of combinations of arbitrary two keywords (hereinafter referred to as the second combination number) among a plurality of keywords (character strings) included in the input path expression. . As described above, for example, when the input path expression is “A / B / C”, the combination of any two keywords is “A, B”, “A, C”, and “B, C”. The second combination number is 3.
The inter-path expression evaluation unit 421 calculates the first combination number / the second combination number as the evaluation value of the intermediate path expression from the calculated first combination number and second combination number.

Here, the evaluation values of the first intermediate path expression and the second intermediate path expression will be specifically described. First, the hierarchical relationship of the input path expression “Goda / invention / structured document” is “Goda → invention”, “Goda → structured document”, and “invention → structured document”. The hierarchical relationship of the first intermediate path expression “invention / (Goda and structured document)” is “invention → Goda” and “invention → structured document”. The hierarchical relationship of the second intermediate path expression “Goda / invention / structured document” is “Goda → invention”, “Goda → structured document”, and “invention → structured document”.
In this case, the first combination number in the input path expression and the first intermediate path expression is 1 (“invention → structured document”). The number of first combinations in the input path expression and the second intermediate path expression is 3 (“Goda → Invention”, “Goda → Structured Document”, “Invention → Structured Document”).
The second combination number in the input path expression is 3 (“Goda → Invention”, “Goda → Structured Document”, “Invention → Structured Document”).
Thereby, the evaluation value of the first intermediate path expression is 1/3 from the number of first combinations / the number of second combinations. The evaluation value of the second intermediate path expression is 3/3 (= 1).

  The inter-path expression evaluation unit 421 orders the first intermediate path expression and the second intermediate path expression based on the calculated evaluation values of the first intermediate path expression and the second intermediate path expression. For example, an intermediate path expression having a large calculated evaluation value (here, the second intermediate path expression) is ranked higher.

  Note that the second algorithm described above is merely an example. For example, the hierarchical relationship definition or the evaluation value calculation method may be other than the one described here.

Next, the query creation unit 411 refers to the vocabulary hierarchy relationship graph storage unit 132 and acquires a vocabulary hierarchy relationship graph including the hierarchy relationship for each input path expression and each intermediate path expression as, for example, a parent-child relationship (step S66). .
Note that the query creation unit 411, for example, if the lexical hierarchy relationship graph including all of the hierarchical relationships of the input path expression and the intermediate path expression as a parent-child relationship does not exist in the vocabulary hierarchy relationship graph storage unit 132, the input path expression and A lexical hierarchy relation graph including a part of the intermediate path expression hierarchy relation as a parent-child relation is acquired. For example, when the input path expression (intermediate path expression) is A / B / C (the hierarchical relation of the input path expression is A to B, B to C), the lexical hierarchy relation graph that completely includes the hierarchical relation is the vocabulary. Assume that the hierarchical relationship graph storage unit 132 does not exist. In this case, the query creation unit 411 acquires, for example, a lexical hierarchy relationship graph including only the hierarchical relationship from A to B as a parent-child relationship, or a vocabulary hierarchical relationship graph having only a hierarchical relationship from B to C as a parent-child relationship.

The query creation unit 411 determines the node type of each character string (vocabulary) included in the input path expression and the intermediate path expression based on the acquired vocabulary hierarchy relation graph (step S67). Based on the hierarchical relationship of the input path expression and the node type of the character string included in the determined input path expression, the query creating unit 411 has search target data having a hierarchical relationship between the character strings included in the input path expression. Create a search expression to search for, for example, XPath expression. Similarly, the query creation unit 411 has a hierarchical relationship between character strings included in the intermediate path expression based on the hierarchical relationship of the intermediate path expression and the node type of the character string included in the determined intermediate path expression. Create a search expression (XPath expression) for searching the search target data.
That is, the query creation unit 411 converts the input path expression and the intermediate path expression into XPath expressions, respectively (step S68). At this time, the query creation unit 411 executes conversion processing based on the conversion data stored in the conversion data storage unit 133. Since this conversion process is the same as that of the first embodiment described above, a detailed description thereof will be omitted.

Next, the search control unit 122 of the search execution unit 41 specifies the XPath expression created by the query creation unit 411 and outputs a search request to the storage unit 13.
In response to the search request from the search control unit 122, the storage unit 13 searches the search target data storage unit 131 for search target data (structured document) that matches the search request under the control of the search control unit 122. (Step S69). At this time, the storage unit 13 performs a search based on each XPath expression created by the query creation unit 411. The storage unit 13 returns search result data including the searched search target data to the search control unit 122 for each XPath expression created by the query creation unit 411. Thereby, the search control unit 122 acquires search result data for each XPath expression.

  Hereinafter, search result data including search target data searched based on an XPath expression created (converted) from an input path expression is referred to as input path expression search result data. Further, search result data including search target data searched based on an XPath expression created from an intermediate path expression is referred to as intermediate path expression search result data. It should be noted that the expression “search result data” includes both input path type search result data and intermediate path type search result data.

The search control unit 122 passes the acquired search result data and the input path expression to the search result evaluation unit 42.
The path expression inter-document evaluation section 141 of the search result evaluation section 42 is based on the input path expression passed from the search control section 122, the search result data, and the result of ordering of the intermediate path expressions by the path expression evaluation section 421. The search target data (structured document) included in the search result data is evaluated (step S70).

At this time, the path expression inter-document evaluation unit 141 calculates an evaluation value for each search result data of the input path expression and the intermediate path expression based on the input path expression passed from the search control unit 122.
That is, the path expression inter-document evaluation unit 141 calculates each evaluation value of the search target data included in the search result data of the input path expression based on the input path expression passed from the search control unit 122. Thereby, the path type inter-document evaluation unit 141 orders the search target data included in the search result data of the input path type based on the calculated evaluation value.

Further, the path type inter-document evaluation unit 141 calculates each evaluation value of the search target data included in the search result data of the intermediate path type based on the input path type passed from the search control unit 122. As a result, the path expression inter-document evaluation unit 141 orders the search target data included in the intermediate path expression search result data based on the calculated evaluation value.
Note that the evaluation value of the search target data is calculated by, for example, the first algorithm of the first embodiment described above, and thus detailed description thereof is omitted.

  Further, the pass expression inter-document evaluation unit 141 is included in the search result data based on the ordering result for the intermediate path expression and the ordering result for each of the search target data by the inter-pass expression evaluation unit 421 in step S65. Order the search target data. The search result data including the search target data ordered by the path type inter-document evaluation unit 141 is passed to the search execution unit 41.

  The search control unit 122 returns the search result data passed from the path type inter-document evaluation unit 141 to the searcher 20 as a search result for the search request from the searcher 20 (step S71). Thereby, the searcher 20 acquires the search result for the search request (step S72). In this case, the search results are presented (displayed) to the searcher 20 in order from the higher-order search target data, the results ordered by the path type inter-document evaluation unit 141.

  In step S66, when a plurality of vocabulary hierarchy relationship graphs are acquired by the query creation unit 411, the processing of steps S67 and S68 is executed for each of the acquired vocabulary hierarchy relationship graphs. As a result, for example, when a plurality of XPath expressions are created (converted) from the input path expression (or intermediate path expression), the process of step S69 is executed with the plurality of XPath expressions being in an OR (logical ring) relationship with each other. Is done.

Next, a specific example of ordering of search target data by the path type inter-document evaluation unit 141 will be described with reference to FIGS. 17 and 18.
Here, the intermediate path expressions created from the input path expression by the query creation unit 411 are the intermediate path expression 1 and the intermediate path expression 2.
FIG. 17 shows an example of search result data for each search condition path expression (input path expression, intermediate path expression 1, intermediate path expression 2). In the example illustrated in FIG. 17, search target data 1 to 3 are included in the input path type search result data. Similarly, the search result data of the intermediate path expression 1 includes search target data 4 to 6. Further, the search result data of the intermediate path expression 2 includes search target data 7 and 8.

Here, it is assumed that the evaluation value of the intermediate path expression 1 calculated by the inter-path expression evaluation unit 421 using, for example, the second algorithm described above is 1/3 and the evaluation value of the intermediate path expression 2 is 1. In this case, the inter-path expression evaluation unit 421 performs ordering in the order of the intermediate path expression 2 and the intermediate path expression 1.
In addition, it is assumed that the path type inter-document evaluation unit 141 calculates an evaluation value for each search result data of the input path type, the intermediate path type 1 and the intermediate path type 2 by the first algorithm described above, for example. As a result, in the search result data of the input path expression, it is assumed that the search target data 1, the search target data 3, and the search target data 2 are ordered in this order based on the calculated evaluation value. In the intermediate path expression 1, it is assumed that the search target data 5, the search target data 4, and the search target data 6 are ordered in this order based on the calculated evaluation value. In the intermediate path expression 2, it is assumed that the search target data 8 and the search target data 7 are ordered in this order based on the calculated evaluation value.
The path expression inter-document evaluation unit 141 orders the entire search target data included in the search result data based on the ordering result for the intermediate path expression and the ordering result for each search result data as described above.

FIG. 18 shows a result of ordering the search target data included in the search result data based on the above intermediate path expression and the result of ordering for each search result data.
As described above, the intermediate path expressions are ordered in the order of the intermediate path expression 2 and the intermediate path expression 1. In this case, the search target data is ordered in the order of the search result data of the input path expression, the search result data of the intermediate path expression 2, and the search result data of the intermediate path expression 1. Further, in ordering the search target data included in each search result data, the order of the search results data described above is the order of order.

Therefore, as shown in FIG. 18, first, the search target data 1, the search target data 3, and the search target data 2 included in the search result data of the input path type are ordered higher. Next, the search target data 8 and the search target data 7 included in the search result data of the intermediate path expression 2 are ordered. Thereafter, the search target data 5, the search target data 4, and the search target data 6 included in the search result data of the intermediate path formula 1 are ordered.
That is, in the case described above, search result data (each search target data included therein) is presented to the searcher 20 in the order shown in FIG.

  As described above, in this embodiment, unlike the first embodiment described above, a plurality of search condition path expressions (intermediate path expressions) are created from the search condition path expression (input path expression) specified by the searcher 20. Is done. In the present embodiment, in addition to the XPath expression created from the input path expression, search processing is executed based on the XPath expression created from this intermediate path expression. Thereby, it is possible to search for a hierarchical relationship other than the hierarchical relationship of the input path expression including a plurality of character strings (vocabulary) designated by the searcher 20. Therefore, the searcher 20 can execute a search considering the hierarchical relationship of the structured document without being aware of the data structure of the structured document, as compared with the first embodiment described above. .

  Further, in this embodiment, by evaluating the search results (data) of the input path expression and the intermediate path expression, the structured documents having the hierarchical structure intended by the searcher 20 are ranked higher and the search results are displayed. Can return. As a result, the searcher 20 can efficiently acquire data (structured document) required by the searcher 20.

  Further, in the present embodiment, the intermediate path expression created from the input path expression is evaluated in advance, and ranking is performed, thereby reducing the evaluation processing of the search target data included in the search result, and as a whole The processing time can be shortened.

  Note that the present invention is not limited to the above-described embodiments as they are, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage. Further, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the above embodiments. For example, some components may be deleted from all the components shown in each embodiment. Furthermore, you may combine the component covering different embodiment suitably.

1 is a block diagram mainly showing a functional configuration of a search result evaluation apparatus according to a first embodiment of the present invention. The flowchart which shows the process sequence at the time of the vocabulary hierarchy relationship graph being created in the search result evaluation apparatus 10 shown in FIG. The flowchart which shows the process sequence when the search according to the search request | requirement from the searcher 20 is performed in the search result evaluation apparatus 10 shown in FIG. The figure which shows an example of the search object data stored in the search object data storage part 131. The flowchart which shows the process sequence which produces the structure graph which shows the hierarchical structure of analysis object data. The figure which shows an example of the structure graph 200 which shows the hierarchical structure of the weather forecast data 100 shown in FIG. The flowchart which shows the process sequence which produces a vocabulary hierarchy relationship graph using a structure graph. The abbreviated figure of the vocabulary hierarchy relationship graph 300 created by the vocabulary hierarchy relationship graph creation unit 112. The flowchart which shows the process sequence which produces a search expression. The figure which shows an example of the data structure of the conversion data stored in the conversion data storage part 133. FIG. The figure which shows an example of the search result by the XPath expression created by the query preparation part 121. FIG. The figure which shows an example of the data structure of the information which shows the setting level defined beforehand. The figure which shows the search object data 101 obtained as a search result. The figure which shows the search object data 102 obtained as a search result. The block diagram which mainly shows a function structure of the search result evaluation apparatus which concerns on the 2nd Embodiment of this invention. The flowchart which shows the process sequence at the time of a search being performed according to the search request | requirement from the searcher 20 of the search result evaluation apparatus 40 shown in FIG. The figure which shows an example of the search result data of each search condition path type | formula. The figure which shows the result by which the search object data contained in search result data was ordered.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 10,40 ... Search result evaluation apparatus, 11 ... Data analysis part, 12, 41 ... Search execution part, 13 ... Memory | storage part, 14,42 ... Search result evaluation part, 111 ... Data analysis control part, 112 ... Vocabulary hierarchy relation graph Creation unit, 121, 411 ... Query creation unit, 122 ... Search control unit, 131 ... Search target data storage unit, 132 ... Vocabulary hierarchy relation graph storage unit, 133 ... Conversion data storage unit, 141 ... Path type inter-document evaluation unit, 421 ... Evaluation unit between pass expressions.

Claims (5)

In a search result evaluation apparatus that performs a search according to a search condition including a plurality of character strings specified by a user, with respect to a search target data storage unit storing a plurality of structured documents to be searched,
A lexical hierarchy relation graph storage means for storing a lexical hierarchy relation graph indicating a hierarchical relation between vocabularies included in each of the plurality of structured documents stored in the search target data storage means, for each structured document;
Query creation means for creating a search expression for searching a structured document from a plurality of character strings included in the search condition with reference to the vocabulary hierarchy relationship graph storage means;
A search unit that searches the search target data storage unit for a structured document that matches the created search formula;
A search result evaluation device comprising: search result evaluation means for evaluating the searched structured document based on a plurality of character strings included in the search condition. The search result evaluation means includes:
A calculation unit that calculates an evaluation value of the structured document based on a location where a plurality of character strings included in the search condition appear in the searched structured document;
The search result evaluation apparatus according to claim 1, wherein the searched structured documents are ordered based on the calculated evaluation value. A conversion data storage means for storing relation information indicating a hierarchical relation between character strings and search expression information indicating a search expression for searching for a structured document having a hierarchical relation between the character strings in advance; Equipped,
The query creation means includes:
Expanding means for expanding a plurality of search conditions from the search conditions by rearranging the character strings included in the search conditions;
Obtaining means for acquiring a lexical hierarchy relation graph including a plurality of character strings included in the search condition as a vocabulary for each of the search condition specified by the user and the plurality of expanded search conditions;
A search for creating a search formula indicated by the search formula information stored in the conversion data storage means in association with the relationship information indicating the hierarchical relationship between the plurality of character strings indicated by the acquired vocabulary hierarchical relationship graph Including formula creation means and
The search means searches the search target data storage means for a structured document that matches the search expression created by the search expression creation means,
The search result evaluation means includes:
A search condition evaluation unit that evaluates each of the plurality of expanded search conditions based on a search condition including a plurality of character strings specified by the user;
The structured document searched by the search unit is evaluated based on a plurality of character strings included in the search condition specified by the user and an evaluation result by the search condition evaluation unit. Search result evaluation device. The search condition evaluation means, based on the arrangement order of a plurality of character strings included in the search condition specified by the user and the arrangement order of a plurality of character strings included in each of the expanded plurality of search conditions, Including a calculation means for calculating an evaluation value of each of the expanded search conditions,
The search result evaluation device according to claim 3, wherein the search result evaluation unit orders the searched structured documents based on the calculated evaluation value. A search target data storage unit storing a plurality of structured documents to be searched and a vocabulary hierarchy relation graph storage unit are included, and the search target data storage unit includes a plurality of character strings designated by the user. A search result evaluation method applied in a search result evaluation device that executes a search according to a search condition,
A lexical hierarchy relation graph indicating a hierarchical relation between vocabularies included in each of the plurality of structured documents stored in the search target data storage means is stored in the vocabulary hierarchy relation graph storage means for each structured document. Steps,
Creating a search expression for searching a structured document from a plurality of character strings included in the search condition with reference to the vocabulary hierarchy relation graph storage means;
Searching the search target data storage unit for a structured document that matches the created search expression;
Evaluating the searched structured document based on a plurality of character strings included in the search condition. A search result evaluation method comprising:

Images