JP2005251092A - Information retrieval system, information retrieval method and information retrieval program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an information retrieval system for storing and using the past retrieval experience of an operator, and to provide an information retrieval method and an information retrieval program. <P>SOLUTION: In the information retrieving system 1, retrieval for a database 28 is performed by using an inputted query. Furthermore, in this information retrieval system 1, the inputted query is stored as a graph, so as to generate the hint of the change of the query in order to obtain a desired retrieval result. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to an information search system, an information search method, and an information search program that can use past experiences.

  2. Description of the Related Art Conventionally, a technique for structuring a query used for a search in order to obtain an appropriate search result in a database search is known. For example, Patent Document 1 discloses a technique in which a user selects a search term and defines a hierarchical structure between the selected terms. Japanese Patent Application Laid-Open No. 2004-228561 discloses a technique for making a plurality of keywords added for data retrieval into a hierarchical structure.

JP 2000-207422 A JP 2002-342359 A

  However, in the techniques of Patent Documents 1 and 2, since the query does not have a structure that can be flexibly deformed, the dynamic change of the query cannot be expressed by the structure change. For this reason, the search experience of the past operator cannot be appropriately expressed by the structure, and the experience cannot be accumulated and used.

  The present invention has been made to solve this problem, and an object of the present invention is to provide an information search system, an information search method, and an information search program capable of accumulating and using search experiences of past operators.

  In order to solve the above problems, the invention of claim 1 is an information search system for performing a search on a database using an input query, a graphing means for converting the query into a graph, and a change of the graph. Storage means for storing a graph change history as a history.

  The invention according to claim 2 is the information search system according to claim 1, wherein the graphing means converts the query into a graph based on a plurality of characteristics of the query-containing words included in the query. And

  According to a third aspect of the present invention, in the information search system according to the second aspect, the plurality of characteristics include characteristics defined for a single word.

  The invention according to claim 4 is the information search system according to claim 3, wherein the characteristic defined for the single word is an attribute of the word.

  According to a fifth aspect of the present invention, in the information search system according to the second aspect, the plurality of characteristics include characteristics defined for a plurality of word sets.

  According to a sixth aspect of the present invention, in the information search system according to the fifth aspect, the characteristics defined for the plurality of sets of words include a specific abstract relationship, an overall partial relationship, a synonym relationship, a complementary relationship, an equivalent relationship, It is one of an inverse relationship and an order relationship.

  The invention according to claim 7 is the information search system according to any one of claims 2 to 6, wherein the graph is a combination of the query-containing words.

  The invention according to claim 8 is the information search system according to any one of claims 2 to 7, further comprising a dictionary defining characteristics of words, wherein the plurality of characteristics of the query-containing word are based on the dictionary. It is determined.

  According to a ninth aspect of the present invention, in the information retrieval system according to the eighth aspect, the word whose characteristics are defined in the dictionary is a word extracted from data stored in the database.

  The invention according to claim 10 is the information search system according to any one of claims 1 to 9, further comprising detection means for detecting start and end of a series of searches by the operator, wherein the graph change history is: It is stored for each series of searches.

  The invention according to claim 11 is the information search system according to claim 10, further comprising temporary storage means for temporarily storing a graph change history in the series of searches, and in response to the detection of the end, the temporary storage. The graph change history stored in the means is stored in the storage means.

  According to a twelfth aspect of the present invention, in the information search system according to any one of the first to eleventh aspects, RDF is used to represent a schema as a basis for the graphing.

  A thirteenth aspect of the present invention is the information search system according to any one of the first to eleventh aspects, wherein the data stored in the database is represented by RDF.

  The invention according to claim 14 is the information search system according to any one of claims 1 to 11, wherein the graph stored in the storage means is expressed by RDF.

  According to a fifteenth aspect of the present invention, in the information search system according to any one of the first to fourteenth aspects, the evaluation specifying means for specifying the evaluation for the search result using the query that is the conversion source of the graph change history is further provided. And the graph change history and the evaluation are stored in association with each other.

  According to a sixteenth aspect of the present invention, in the information search system according to the fifteenth aspect, the evaluation is an operator's satisfaction with the search result.

The invention of claim 17 is the information search system according to claim 16,
The satisfaction degree is determined based on an operation history of an operator information retrieval system.

  The invention according to claim 18 is the information search system according to claim 16, further comprising satisfaction input means for allowing the operator to input the satisfaction.

  The invention according to claim 19 is the information search system according to any one of claims 1 to 14, wherein a graph change corresponding to a newly inputted query is selected from the graph change history stored in the storage means. It further comprises a specifying means for specifying the history.

  According to a twentieth aspect of the present invention, in the information search system according to the nineteenth aspect, the specifying unit converts the query into a graph obtained by the graphing unit, and changes the graph stored in the storage unit. The graph change history is specified based on the degree of coincidence with the graph included in the history.

  According to a twenty-first aspect of the present invention, in the information search system according to the nineteenth aspect, the graph change history stored in the storage unit is associated with the evaluation, and the specifying unit determines the graph change history based on the evaluation. It is characterized by specifying.

  According to a twenty-second aspect of the present invention, in the information search system according to the twenty-first aspect, the evaluation is an operator satisfaction with respect to a search result, and the specifying means is a graph change associated with a predetermined satisfaction or higher. It is characterized by specifying a history.

  The invention according to claim 23 is the information search system according to any one of claims 19 to 22, wherein the auxiliary information for generating auxiliary information related to the query change is based on the graph change history specified by the specifying means. It further has a generating means.

  According to a twenty-fourth aspect of the present invention, in the information search system according to the twenty-third aspect, the information retrieval system further comprises output means for outputting the auxiliary information so as to be recognized by an operator.

  The invention of claim 25 is an information search method, wherein a) a query input step for inputting a query, b) a graphing step for converting the query into a graph, and c) a graph change of the history of the graph change And a storing step of storing in the storage means as a history.

  The invention according to claim 26 is an information search program, wherein execution of the information search program by a computer includes: a) a query input step for inputting a query to the computer; and b) graphing for converting the query into a graph. And c) storing a graph change history as a graph change history in a storage means.

  According to the first to twenty-sixth aspects of the present invention, since the query change history is stored in the storage means as the graph change history, the past experience related to the query change history can be accumulated and used.

  According to the inventions of claims 2 to 9, since the graph is generated based on a plurality of characteristics of the query-containing word, it is possible to use past experiences related to the characteristics of the word.

  According to the invention of claim 8, it is possible to graph the words included in the dictionary.

  According to the inventions of claims 10 to 11, since the graph change history is stored in the storage means every time a series of searches are started to end, the past experience related to the query change history in the series of searches is accumulated and stored. Become available.

  According to the invention of claim 11, since the graph change history is not stored in the storage means until the end of the series of searches, the overhead caused by the storage in the storage means can be eliminated.

  According to the fifteenth to eighteenth aspects of the present invention, since the graph change history and the evaluation are stored in association with each other, the past experience relating to the evaluation can be used.

  According to the nineteenth to twenty-fourth aspects of the present invention, since the graph change history corresponding to the newly input query is specified, the graph change history can be accumulated and used.

  According to the twentieth aspect, it is possible to use a graph change history including a graph having a high degree of coincidence.

  According to the invention of claim 21, the graph change history to which a specific evaluation is given can be used.

  According to the invention of claim 23 or claim 24, auxiliary information based on the specified graph change history can be used.

  According to the invention of claim 24, since the operator can recognize the auxiliary information, the query can be appropriately changed.

  The information search system 1 according to the present embodiment searches a database using an input query. Further, the information search system graphs and stores the input query, and generates a query change hint for obtaining a desired search result. Below, the structure and operation | movement of such an information search system 1 are demonstrated.

<Hardware configuration>
FIG. 1 is a block diagram showing a hardware configuration of a computer 10 for realizing the information search system 1 of the present embodiment.

  The computer 10 includes a CPU 12, a memory 13, a graphics adapter 14, an input device 15, a storage device 16, and a printer 17 connected via a bus 11 and the like. A display 18 serving as an output device is connected to the graphics adapter 14. The input device 15 includes a keyboard 151 and a pointing device 152.

  An operating system (hereinafter also referred to as “OS”) 161 that defines the basic operation of the computer 10 is installed in the storage device 16 configured by a hard disk drive or the like. An information search program 162 that is executed under the management of the OS 161 is also installed in the storage device 16. In the information search system 1, the functions of the functional blocks shown in FIGS. 2, 5, and 9 are realized by the computer 10 executing the information search program 162.

  In the above, the example in which the information search system 1 is realized by the single computer 10 has been described. However, the information search system 1 may be realized by a plurality of computers connected to the network.

<Functional configuration>
FIG. 2 is a block diagram showing a functional configuration of the information search system 1.

  The information search system 1 includes an input unit 21 that detects an operation using the input device 15. The input unit 21 outputs the query input by the operator to the query analysis unit 22. The input unit 21 outputs a search start signal and a search end signal to the data store management unit 23 in response to detection of the start and end of a series of searches, respectively. Further, the input unit 21 specifies the operator's evaluation on the search result and outputs it to the data store management unit 23 at the end of a series of searches.

  The information search system 1 includes a mechanism that improves the suitability, suitability, and usefulness, which are the main indexes relating to the effectiveness of the search results, to improve the suitability and usefulness depending on the operator. Therefore, in the following, description will be given assuming that “satisfaction” as a concept including appropriateness and usefulness is adopted as an evaluation. The degree of satisfaction may be specified by causing the operator to input directly, or may be specified by analyzing the operation history of the operator. Specific examples of how to specify the satisfaction level are shown in the examples.

  The query analysis unit 22 converts the query into a graph and outputs it to the search engine 24, the data store management unit 23, and the hint generation unit 25. However, the graph output to the data store management unit 23 and the hint generation unit 25 is a graph having only a basic structure from which part or all of the query content information is removed. The query analysis unit 22 converts the query into a graph based on the schema 26 and the vocabulary dictionary 27 held in the storage device 16. Details of the schema 26, the vocabulary dictionary 27, and the query analysis unit 22 will be described later.

  The search engine 24 searches the database 28 in which the search target data is stored using the input graph, and outputs the search result to the output unit 30. The database 28 is realized using the storage device 16. The representation of the data stored in the database 28 is preferably done by RDF.

  The information retrieval system 1 includes a data store 29 that stores a history of changes in graphs (basic structure) and satisfaction. The data store 29 is provided with a temporary area 291 and a store area 292. The data store 29 is realized using the storage device 16. The representation of the data (graph) stored in the data store 29 is also preferably done by RDF.

  The temporary area 291 temporarily stores a graph change history G = [G (1), G (2),..., G (N)] of a series of searches being executed. Here, graphs G (1), G (2),..., G (N) mean graphs corresponding to individual queries in the graph change history G. The graph change history G indicates the entire graphs G (1), G (2),..., G (N) including order information. Indexes 1, 2,..., N for identifying graphs correspond to the order in which the graphs are generated.

  In the store area 292, a plurality of graph change histories Gi = [Gi (1), Gi (2),..., Gi (Ni)] (i = 1, 2,... , p) is accumulated. Here, Gi (1), Gi (2),..., Gi (Ni) means a graph corresponding to each query in the graph change history Gi. The graph change history Gi indicates the entire graph Gi (1), Gi (2),..., Gi (Ni) including order information. Indexes 1, 2,..., Ni for identifying graphs correspond to the order in which the graphs are generated. Further, in the store area 292, the satisfaction level Pi for the search result based on the graph change history Gi is stored in association with the graph change history Gi. By accumulating the query change history as the graph change history in the storage unit in this way, it becomes possible to use past experiences related to the query change history in a series of searches. Further, past satisfaction can be used by associating the graph change history with satisfaction.

  Input / output of information to / from the data store 29 is managed by the data store management unit 23. Specifically, when a graph is input from the query analysis unit 22, the data store management unit 23 stores the graph in the temporary area 291. Further, when the search end signal and the satisfaction level are given from the input unit 21, the data store management unit 23 stores the graph change history stored in the temporary area 291 in the store area 292 in association with the satisfaction level, and then temporarily stores it. The graph change history stored in the area 291 is deleted. Thereby, the graph change history is stored in the store area 292 for each series of searches, and the past experience relating to the query change history in the series of searches can be used.

  The hint generation unit 25 searches the store area 292 using a graph corresponding to a newly input query, includes a matching graph in which the basic structure matches the input graph, and gives a satisfaction level equal to or higher than a predetermined level. Specified graph change history (hereinafter also referred to as “high satisfaction graph change history”). Furthermore, the hint generation unit 25 generates a hint serving as auxiliary information for query change based on the high satisfaction graph change history and outputs the hint to the output unit 30. Details of the hint generation unit 25 will be described later.

  The output unit 30 presents search results and hints to the operator using an output device. The output unit 30 is realized by the display 18 that is an output device, and provides a user interface for the operation of the operator.

<Schema>
FIG. 3 is a diagram illustrating a graph G101 that defines a schema serving as a basis for graphing by the query analysis unit 22. Although the representation method of the schema 26 is not limited, the representation by RDF (Resource Description Framework) is adopted in FIG. Since the schema 26 can be changed as appropriate, the graph G101 is merely an example of the schema 26.

  Each node N101 to N106 of the graph G101 has a vocabulary group [Ar (1), Ar (2),..., Ar (Nr)] (r = 1, 2,. ) Is a node for describing a vocabulary (word) selected from (1). For example, the root node (node N101) is a node for describing a vocabulary selected from a vocabulary group [A0 (1), A0 (2),..., A0 (N0)] having an attribute A0.

  In the arc connecting the parent node and the child node, an attribute associated with the child node is described as a property. For example, in an arc connecting the node N102 and the node N103, the attribute A11 associated with the node N103 is described as a property.

  Note that the schema 26 (or RDF used to represent the schema 26) may be updated as appropriate based on the data structure of data stored in the database 28. This update may be performed manually or automatically. Thereby, the search process in the search engine 24 becomes easy.

  In order to facilitate the search process in the search engine 24 and the storage process in the data store management unit 23, the RDF used for expressing the schema 26 and the RDF used for expressing the data stored in the database 28 are used. It is desirable that the RDF used to represent the data stored in the data store 29 has a similar data model, but a data model that is partially different due to a request for improving the processing performance of large-capacity data. It is not impeded to adopt an expression using a method other than RDF in part.

  Such a graph G101 is described by an XML document, for example.

<Vocabulary dictionary>
The vocabulary dictionary 27 includes vocabulary groups [Ar (1), Ar (2),..., Ar (Nr)] (r = 1, 2,...) Extracted from data stored in the database 28. Contains information about the vocabulary group. That is, the vocabulary dictionary 27 includes not only vocabulary data but also vocabulary metadata. Vocabulary information includes characteristics defined for a single vocabulary and characteristics defined for a set of multiple vocabularies.

  Typical examples of properties defined for a single vocabulary are the attributes described above, and the vocabulary group [Ar (1), Ar (2), ..., Ar (Nr)] (r = 1,2, ..) And the attributes of each vocabulary of the vocabulary group are described in the vocabulary dictionary 27. FIG. 4 is a diagram showing a graph G111 describing a vocabulary group [Ar (1), Ar (2),..., Ar (Nr)] having an attribute Ar. The expression method of the vocabulary and attributes in the vocabulary dictionary 27 is not limited, but the expression by RDF is adopted in FIG. Specifically, in the graph G111, the resource is described as an attribute Ar, the property is “vocabulary”, and the literal is described as a vocabulary group [Ar (1), Ar (2),..., Ar (Nr)].

  Typical examples of characteristics defined for a plurality of vocabulary sets are concrete abstract relations, whole partial relations, synonymous relations, complementary relations, equivalent relations, opposite relations, order relations, etc. (hereinafter, these are collectively referred to as “ Also referred to as “multiple word characteristics”. The concrete abstract relationship is a characteristic defined in a vocabulary set in which one of the vocabulary forms the other, such as a vocabulary set of “pencil” and “writing instrument”. The whole partial relationship is a characteristic defined in a vocabulary set, one of which is a vocabulary related to the other part, such as the vocabulary set of “car” and “engine”. Synonymous relationships are characteristics defined in vocabulary sets that can be interchanged with each other, such as the vocabulary set of “automobile” and “automobile”. Complementary relationship is a characteristic defined in a vocabulary set that complements each other's deficiencies, such as a vocabulary set of “male” and “female”. An equivalence relationship is a characteristic defined for a set of vocabularies having the same attribute. Opposite relationship is a characteristic defined in lexical pairs having the same attribute but opposite to each other, such as vocabulary sets of “parent” and “child”. The order relation is a characteristic defined in a set of vocabularies whose order can be defined, such as a set of vocabularies “eldest son” and “second son”. In the above description, a vocabulary group composed of two vocabularies is illustrated, but it is not impeded that the vocabulary group includes three or more vocabularies.

  The vocabulary dictionary 27 is updated as appropriate based on the data stored in the database 28. This update may be performed manually or automatically.

<Query analysis part>
FIG. 5 is a block diagram illustrating a functional configuration of the query analysis unit 22.

  The query analysis unit 22 includes a syntax analysis unit 221 that analyzes a syntax of a natural sentence query and outputs an analysis result to the graph construction unit 222. The syntax analysis unit 221 includes a morpheme analysis unit 221a that divides a natural sentence into words, and a dependency analysis unit 221b that analyzes dependency between content words included in the word.

  The graph construction unit 222 constructs a graph corresponding to the query from the analysis result with reference to the schema 26 and the vocabulary dictionary 27. Specifically, the graph construction unit 222 extracts the vocabulary whose attributes are defined in the vocabulary dictionary 27 from the content words, and sets the node in which the word is described as the whole or part of the structure defined in the schema 26. Join with arcs to match. An example of the graph G121 constructed in this way is shown in FIG. The graph G121 has a structure in which the vocabulary W1 and the vocabulary W2 are connected by an arc AW2, and the vocabulary W1 and the vocabulary Wn are connected by an arc AWn. Further, the graph construction unit 222 extracts a set of words defined by a plurality of word characteristics in the vocabulary dictionary 27 from the content words, and connects nodes corresponding to the set of words with an arc. In this arc, a multi-word characteristic defined in a set of connected words is described. An example of the graph G131 constructed in this way is shown in FIG. In the graph G131, in addition to the structure of the graph G121, the word W2 and the word Wk are connected by an arc, and a multi-word characteristic of order is described in the arc. Thus, by converting a query to a graph having a structure that can be flexibly deformed, the past experience can be easily used. Further, by referring to the vocabulary dictionary 27 in constructing the graph, the vocabulary included in the vocabulary dictionary 27 can be graphed.

  Since the vocabulary dictionary 27 referred to in constructing the graphs G121 and G131 includes information on the characteristics of the words, the characteristics of the words are used to construct the graphs G121 and G131. In this way, by constructing a graph that combines the words based on the plurality of characteristics of the words contained in the query, it becomes possible to use past experiences related to the characteristics of the words.

  The graph constructed by the graph construction unit 222 is output to the search engine 24 and also given to the basic structure extraction unit 223. The basic structure extraction unit 223 generates a graph in which a predetermined node of a given graph is replaced with a wild card (a symbol representing an arbitrary vocabulary). An example of the graph G141 generated by the basic structure extraction unit 223 is shown in FIG. In the graph G141, the node replaced with the wild card is indicated by “?”. The node to be replaced with the wild card can be appropriately set according to the characteristics of the information search system 1 or the operator's request. For example, as shown in FIG. 8, all nodes other than the root node are replaced with the wild card. A mode in which all nodes other than a node corresponding to an attribute common to all vocabularies are replaced with a wild card can be considered. As the number of nodes replaced with wildcards increases, the hints described later approach those that purely reflect only changes in the query structure in a series of searches. On the other hand, when the number of nodes to be replaced with wildcards decreases, the hints described later reflect the query content information in a series of searches. The basic structure of the graph generated by the basic structure extraction unit 223 is output to the data store management unit 23 and the hint generation unit 25.

<Hint generator>
FIG. 9 is a block diagram illustrating a functional configuration of the hint generation unit 25.

  The hint generation unit 25 includes a graph search engine 251 that searches the store area 292 using the graph G corresponding to the input query and extracts a matching graph whose structure matches the graph G. The graph search engine 251 outputs the extracted matching graph to the graph change history specifying unit 252. In FIG. 9, a state in which the graphs G1 (1) and G2 (2) are extracted as the matching graph of the graph G is schematically represented.

  The graph change history specifying unit 252 refers to the satisfaction level (satisfaction levels P1 and P2 in FIG. 9) associated with the graph change history including the matching graph, and a high satisfaction level graph that is given a satisfaction level of a predetermined level or higher. Identify change history. As a result, the information retrieval system 1 can use the high satisfaction graph change history.

  Further, the graph change history specifying unit 252 extracts the graph next to the match graph in the match graph and the high satisfaction graph change history (hereinafter also referred to as “deformed graph”) and outputs the extracted graph to the hint construction unit 253. . The high satisfaction graph change history is a graph change history related to a series of searches in which a search result that obtained a satisfaction level equal to or higher than a predetermined level is obtained by performing a search using a query similar to the conversion source query of the graph G It is. Therefore, if a graph change similar to the high satisfaction graph change history is performed in a series of searches currently being executed, there is a high possibility that a search result with a satisfaction level of a predetermined level or higher will be output. In other words, if a query change that realizes a graph change from a matching graph to a deformed graph is performed in a series of searches that are currently being executed, a search result that provides a satisfaction level equal to or greater than a predetermined level is output. Probability is high. In FIG. 9, the satisfaction level P2 is “satisfaction level or higher”, and the graph change history G2 is specified as a high satisfaction graph change history that can be used for generating hints, and the graph G2 ( The state in which 3) is output as a deformation graph to the hint construction unit 253 is schematically expressed.

  The hint construction unit 253 constructs a hint based on the change from the matching graph to the deformation graph and outputs the hint to the output unit 30. For example, when the matching graph G151 is the graph shown in FIG. 10 and the deformation graph G161 is the graph shown in FIG. 11, the hint constructing unit 253 gives a hint that prompts the operator to add the attribute AWn + 1 by the query. create.

<Overall operation>
FIG. 12 is a flowchart showing the overall operation of the information search system 1. Hereinafter, an operation in a series of searches of the information search system 1 will be described with reference to FIG.

  When the operation of the information search system 1 starts, first, whether or not a series of searches has been started is detected in the input unit 21 (step S101). If the start is detected, the operation flow moves to step S102. If the start is not detected, the operation flow returns to step S101, and the standby state for starting a series of searches continues.

  Subsequently, the input of the query is detected in the input unit 21 (step S102). If a query input is detected, the operation flow proceeds to step S103. If not detected, the operation flow returns to step S102 and the query input standby state continues.

  Subsequently, the query analysis unit 22 converts the query into a graph (step S103), the search engine 24 searches the database 28 using the graph (step S104), and the output unit 30 outputs the search result (step S104). Step S105). Through steps S102 to S105, the operator obtains a search result for the input query. The subroutine in step S103 will be described later.

  After outputting the search result, the data store management unit 23 adds the basic structure of the graph to the temporary area 291 (step S106).

  In step S107 following step S106, detection of whether or not a series of searches is completed is performed in the input unit 21. When the end is detected, the operation flow moves to step S108, and when it is not detected, the operation flow moves to step S110.

  When a series of searches is completed in step S107, the evaluation for the search result by the series of searches is specified (step S108), and the evaluation and the graph change history temporarily stored in the temporary area 291 are associated with each other in the store area. It is stored in 292 (step S109). Thereafter, the operation flow of the entire operation is completed. By copying from the temporary area 291 to the store area 292 in this way, writing to the store area 292 is not performed until the end of a series of searches, thereby eliminating the overhead caused by writing to the store area 292. be able to.

  On the other hand, if the series of searches is not completed in step S107, the hint generation unit 25 searches the store area 292 using the basic structure of the graph used for the search on the database 28 to generate the above hints (step S107). S110). Subsequently, the output unit 30 outputs the hint (step S111). Through steps S <b> 110 to S <b> 111, the operator obtains a hint for changing the input query to a more appropriate query (a query with higher satisfaction). The subroutine in step S110 will be described later. After step S111 ends, the operation flow returns to step S101.

<Graph conversion operation>
Hereinafter, the graph conversion subroutine (step S103) of FIG. 12 will be described with reference to the flowchart of FIG.

  When a query is input from the input unit 21, first, the syntax analysis unit 221 performs syntax analysis of a natural sentence query and outputs the analysis result to the graph construction unit 222 (step S201).

  Subsequently, the graph construction unit 222 generates a graph from the analysis result with reference to the schema 26 and the vocabulary dictionary 27, and outputs the graph to the search engine 24 and the basic structure extraction unit 223 (step S202).

  Subsequently, the basic structure extraction unit 223 extracts the basic structure of the query and outputs it to the data store management unit 23 and the hint generation unit 25 (step S203).

  With such an operation flow, a query is converted into a graph using a plurality of characteristics of words. Since the structure of the graph can be changed dynamically and flexibly according to the change of the query, the trial and error experience in the past search by the operator can be expressed by the structure of the graph. This facilitates the accumulation and use of past experiences. In addition, by appropriately updating the vocabulary dictionary 27 and the schema 26, a graph suitable for the data stored in the database 28 can be generated.

<Hint generation operation>
Hereinafter, the hint generation subroutine (step S110) of FIG. 12 will be described with reference to the flowchart of FIG.

  First, the graph search engine 251 stores graphs Gi (1), Gi (2),..., Gi (Ni) (i = 1, 2,..., P) stored in the store area 292. A matching graph is extracted from the inside (step S301).

  Subsequently, the graph change history specifying unit 252 specifies the high satisfaction graph change history and the modified graph (step S302), and the hint constructing unit 253 generates a hint that prompts the query to be changed and outputs it to the output unit 30 (step). S303).

  By changing the query input by the operator in accordance with the hint output in this way, the possibility that a search result with a satisfaction level equal to or higher than a predetermined level is output is increased.

  Below, the Example which applied the information search system 1 which concerns on embodiment to the paper search system and the manual search system is described. The following embodiments do not limit the application target of the information search system 1, but merely show an application example of the information search system 1.

[Example 1]
Example 1 relates to an information search system 1 (hereinafter also referred to as “paper search system”) applied to search for papers. That is, the paper search system is the information search system 1 in which data stored in the database 28 in FIG. 2 is electronic data related to a paper (hereinafter also referred to as “paper data”). The paper search system searches the database 28 using the input query. Further, the paper search system converts a query into a graph and stores it, and uses the storage to generate a query change hint for obtaining a search result with high satisfaction.

<Schema>
FIG. 15 is a diagram illustrating a graph G201 that defines a schema as a basis for graphing by the query analysis unit 22 in the paper search system.

  In the root node (node N201) of the graph G201, "paper" that is an attribute common to all paper data is described. In the child nodes of the root node (nodes N202 to N204), the author of the paper, title, reference list, and the like are described. The child node (nodes N205 to N206) of the node N202 describes the title and affiliation of the author described in the parent node. Specific vocabulary such as author, title, reference list, title, and affiliation described in the nodes N202 to N206 is described in the vocabulary dictionary 27.

  FIG. 15 shows a simplified graph for explanation. Accordingly, in an actual paper search system, it is not prevented to add a node not included in the graph G201 of FIG. 15 to the graph expressing the schema 26.

<Vocabulary dictionary>
FIG. 16 shows a graph G211 describing a vocabulary group having an attribute (author) corresponding to the node N202 of the graph G201 among the vocabulary groups related to the paper search system. The graph G211 is a graph in which the graph G111 in FIG. 4 is adapted to the specific vocabulary of the paper data.

  As shown in FIG. 16, “author”, which is an attribute of the vocabulary group, is described in the root node (node N211) of the graph. Further, specific author names such as “Tanaka”, “Nakamura”, and “Shimizu” are described in the nodes N212 to N214 that are child nodes of the root node. In the vocabulary dictionary 27, a graph having the same structure as that of FIG. 16 is prepared for each attribute.

<Screen transition>
In the paper search system in which the GUI is implemented, an instruction can be given to the paper search system by performing a predetermined operation with the input device 15 while referring to the GUI screen displayed on the display 18.

  Hereinafter, the transition of the GUI screen will be described with reference to the screen transition diagram of FIG. 17 and the screen examples of FIGS.

○ Menu screen;
A menu screen SC11 illustrated in FIG. 18 is displayed when the paper search system is activated (when the entire operation is started).

  The menu screen SC11 includes a transition link 301 to the search screen SC12. When an operation for selecting the transition link 301 is performed on the menu screen SC <b> 11, the display on the display 18 shifts to a screen corresponding to the transition link 301. The selection of the transition link 301 is also an instruction to start a series of searches for the paper search system. For this reason, when the transition link 301 is selected, a search start signal is output from the input unit 21 to the data store management unit 23.

  Hereinafter, the search screen SC12 and the transition destination screen of the search screen SC12 will be described.

○ Search screen;
A search screen SC12 illustrated in FIG. 19 is a screen for inputting a query. The search screen SC12 is used for the first search in a series of searches.

  The search screen SC12 includes a text area 311 for inputting a query and a search button 312. The operator presses the search button 312 after entering the query in the text area 311 to cause the paper search system to execute a database search (paper search) using the query. That is, when pressing of the search button 312 is detected on the search screen SC12, a natural sentence query input in the text area 311 is given from the input unit 21 to the query analysis unit 22 and converted into a graph. The search result obtained by the database search is displayed on the display 18 (search result list display screen SC13). Further, the basic structure of the graph is added to the temporary area 291 almost simultaneously with the display of the search result on the display 18.

○ Search result list display screen;
The search result list display screen SC13 illustrated in FIG. 20 is a screen for displaying a list of search results.

  The search result list display screen SC13 includes a search result list table 321 and a re-search button 322. In the search result list table 321, for example, a list of titles and categories of articles that have been searched and output is displayed. The titles 321a to 321g of each paper listed are also a transition link to a detailed information display screen corresponding to each paper. Therefore, the operator can call up the detailed information display screen SC15 of the desired paper on the display 18 by selecting the transition link corresponding to the desired paper by a predetermined operation. The re-search button 322 is a GUI instruction member for giving a call instruction for the re-search screen SC14 to the paper search system.

○ Detailed information display screen;
A detailed information display screen SC15 illustrated in FIG. 21 is a screen that displays detailed information of a paper corresponding to the transition link selected on the search result list display screen SC13.

  The detailed information display screen SC15 includes a detailed display table 331, a re-search button 332, and a full text display button 333. In the detail display table 331, for example, a title, a category, an author, a keyword, a strange record, and the number of pages are displayed. The re-search button 332 and the full-text display button 333 are GUI instruction members for giving a call instruction for the re-search screen SC14 and the full-text display screen SC16 to the paper search system, respectively.

○ Full text display screen;
The full text display screen SC16 illustrated in FIG. 22 is a screen that displays the full text of the paper displayed on the detailed information display screen SC15. The full text of the paper displayed on the full text display screen SC16 is a document described in plain text not including format information, text or images including format information, and may be PDF (registered trademark).

  The full text display screen SC16 includes a full text 341 of a paper, a print button 342, and a return button 343. The print button 342 is a GUI instruction member for giving a print instruction for the full text 341 of the paper to the paper search system. Therefore, when the print button 342 is pressed on the full text display screen SC16, the full text 341 of the paper is printed from the printer 17, and the display on the display 18 shifts to the print screen SC17. The return button 343 is a GUI instruction member for giving a call instruction for the detailed information display screen SC15 to the paper search system. Accordingly, when the return button 343 is pressed, the display 18 shifts to the detailed information display screen SC15.

○ Print screen;
The print screen SC17 illustrated in FIG. 23 is displayed when a full text print instruction is given on the full text display screen SC16. The print screen SC17 is switched to the detailed information search screen SC15 without waiting for the operator's operation when the print processing is completed.

○ Re-search screen;
The re-search screen SC14 illustrated in FIG. 24 is used for the second and subsequent searches in a series of searches.

  The re-search screen SC14 includes a text area 351 and a search button 352 similar to the search screen SC12. Although the query 351a used in the previous search is displayed by default in the text area 351, the operator can edit the query 351a by a predetermined operation.

  Further, on the re-search screen SC14, a hint HT that prompts the user to change the query is displayed below the text area 351. The hint HT is generated by the hint generator 25 based on the query used in the previous search.

  The re-search screen SC14 includes an end button 353. The end button 353 is a GUI instruction member for instructing the paper search system to shift to the menu screen SC11. By pressing the end button 353, the operator can end a series of searches and switch the display 18 to the menu screen SC11 again. Since pressing the end button 353 also serves as an instruction to end a series of searches, if the pressing of the end button 353 is detected on the re-search screen SC14, the input unit 21 sends a search end signal and satisfaction to the data store management unit 23. Output.

<Satisfaction>
In the screen transition diagram of FIG. 17, when the transition from the search result list display screen SC13 to the re-search screen SC14 is directly performed, it is estimated that the papers required by the operator are not retrieved and output in the search result list display. Therefore, it is considered that the operator's satisfaction with the search result is low. On the other hand, when a transition is made from the search result list display screen SC13 to the detailed information display screen SC15, the full text display screen SC16, or the print screen SC17, it is presumed that the paper required by the operator has been retrieved and output. Further, when the detailed information display screen SC15, the full text display screen SC16, or the print screen SC17 is reached, it is considered that the operator's satisfaction with the search result is improved in this order. For this reason, in the paper search system, when a transition to the detailed information display screen SC15, full-text display screen SC16, or print screen SC17 of the screen display is performed, the parameter P expressing the satisfaction degree of the operator is set to P1 , P2 and P3 (P1 <P2 <P3) are added. This parameter is stored in the store area 292 in association with the graph change history when a transition from the re-search screen SC14 to the menu screen SC11 is performed and a series of searches is completed.

  In this way, by specifying the satisfaction level based on the operation history of the operator's paper search system, it is not necessary for the operator to input the satisfaction level for each series of searches, so the operation load on the operator can be reduced. is there.

<Example of a series of searches>
FIG. 25 and FIG. 26 list changes in queries and graphs in a series of searches, matching graphs and deformation graphs extracted from the store area 292 by the hint generation unit 25, and hints constructed by the hint generation unit 25. . A specific example of a series of searches will be described below with reference to FIGS. 25 and 26.

  First, in the first search in a series of searches, the operator entered a query Q501 “Paper whose author is Tanaka” into the text area 311 of the search screen SC12, and the search was not satisfactory. Therefore, the re-search is performed on the re-search screen SC14. A graph generated based on the query Q501 is a graph G501. In addition, it is assumed that the hint generation unit 25 extracts the matching graph G501a and the modified graph G501b in the first search. In this case, specify the affiliation of “Tanaka” based on the graph change from the matching graph G501a to the deformation graph G501b. ”Is generated and displayed on the re-search screen SC14.

  In the subsequent first re-search, the operator is satisfied with the search by inputting the query Q502 “Papers whose author is Tanaka of the patent university” into the text area 351 of the re-search screen SC14 with reference to the hint HT501. Since the search output was not obtained, it is assumed that the re-search is further performed on the re-search screen SC14. A graph corresponding to the query Q502 is a graph G502. Further, it is assumed that the hint generation unit 25 extracts the matching graph G502a and the modified graph G502b in the first re-search. In this case, based on the graph change from the coincidence graph G502a to the deformation graph G502b, specify the position “Tanaka” additionally. ”Is generated and displayed on the re-search screen SC14.

  In the second subsequent re-search, the operator performs a search by inputting the query Q503 “thesis that is authored by Professor Tanaka of the patent university” into the text area 351 of the re-search screen SC14 with reference to the hint HT502, Since a satisfactory search output was not obtained, it is further assumed that a re-search is performed on the re-search screen SC14. A graph corresponding to the query Q503 is a graph G503. In addition, it is assumed that the hint generation unit 25 extracts the matching graph G503a and the modified graph G503b in the second re-search. In this case, based on the graph change from the matching graph G503a to the modified graph G503b, a hint HT503 “Please specify additional co-authors of“ Tanaka ”” is generated and displayed on the re-search screen SC14.

  In the subsequent third re-search, the operator refers to the hint HT 503 and executes the query Q 504 (converted to graph G 504) “research paper whose author is Nakamura and Professor Tanaka of Patent University” on the re-search screen SC14. It is assumed that search is performed by inputting to the text area 351 and a satisfactory search output is obtained. At this time, the operator performs an operation such as switching the display 18 to the detail display screen SC15, the full text display screen SC16, or the print screen SC17, and then switches the display 18 display to the menu screen SC11 and ends the series of searches. .

  Note that “author”, “affiliation”, and “position” in the graphs G501 to G503 are examples of attributes. The “co-author” of the graph G504 is an example of a multi-word characteristic (equivalent relationship or order relationship).

[Example 2]
The second embodiment relates to an information search system 1 (hereinafter also referred to as “repair manual search system”) applied to search for a repair manual of a motor vehicle (passenger car, truck, motorcycle, etc.). That is, the repair manual search system is an information search system 1 in which data stored in the database 28 in FIG. 2 is electronic data related to a repair manual (hereinafter also referred to as “repair manual data”). The repair manual search system searches the database 28 using the input query. Further, the repair manual search system converts a query into a graph and stores it, and generates a query change hint for obtaining a search result with a high degree of satisfaction using the storage.

<Schema>
FIG. 27 is a diagram illustrating a graph G301 that defines a schema serving as a basis for graphing by the query analysis unit 22 in the repair manual search system.

  In the root node (node N301) of the graph, “repair method” which is an attribute common to all repair manual data is described. In a child node (node N302) of the root node, a large classification (type of vehicle such as a passenger car, a truck, and a motorcycle) is described. A child node (node N303) of the node N302 describes a vehicle part (engine, transmission, tire, etc.) described in the parent node. The specific vocabulary of major classifications and parts described in the nodes N302 and N303 is described in the vocabulary dictionary 27.

  FIG. 27 shows a simplified graph for explanation. Therefore, in the actual paper search system, it is not prevented to add a node not included in the graph G301 in FIG. 27 to the graph expressing the schema 26.

<Vocabulary dictionary>
FIG. 28 shows a graph G311 describing a vocabulary group having an attribute (major classification) corresponding to the node N302 of the graph G301 among the vocabulary groups related to the repair manual search system. The graph G311 is a graph in which the graph of FIG. 4 is adapted to the specific vocabulary of the repair manual data.

  As shown in FIG. 28, the root node (node N311) of the graph G311 describes “major classification” that is an attribute of the vocabulary group. Furthermore, specific types of vehicles such as “passenger car”, “truck”, and “bike” are described in nodes N312 to N314 that are child nodes of the node N311. In the vocabulary dictionary 27, a graph having the same structure as that of FIG. 16 is prepared for each attribute.

<Screen transition>
In the repair manual search system in which the GUI is mounted, an instruction can be given to the repair manual search system by performing a predetermined operation with the input device 15 while referring to the GUI screen displayed on the display 18.

  Hereinafter, the transition of the GUI screen will be described with reference to the screen transition diagram of FIG. 29 and the screen example of FIG. The repair manual search system is different from the paper search system in that a satisfaction input screen SC18 is provided in the middle of the transition from the re-search screen SC14 to the menu screen SC11. Therefore, in the following description of the screen transition, only the satisfaction input screen SC18 will be described, and redundant description of the remaining screens will be omitted. The remaining screens are given the same reference numerals as in FIG. Further, the information related to the paper displayed on the remaining screen is information related to the repair manual in the second embodiment.

○ Satisfaction input screen;
The satisfaction input screen SC18 illustrated in FIG. 30 is a screen for inputting the satisfaction with respect to the search result.

  The satisfaction level input screen SC18 includes a character string 361 that prompts the operator to input the level of satisfaction with respect to the search result, “Yes. 362” and “No” button 363. The operator presses the Yes button 362 when necessary information is obtained by the search output, and presses the No button 363 when it is not obtained. When pressing of the Yes button 362 or the No button 363 is detected on the satisfaction input screen SC18, the display on the display 18 shifts to the menu screen SC11. Since the pressing of the Yes button 362 or the No button 363 also serves as an instruction to end a series of searches, when the pressing of the Yes button 362 or the No button 363 is detected on the re-search screen SC14, the input unit 21 sends a search end signal and satisfaction. The degree is output to the data store management unit 23. Here, the degree of satisfaction to be output is P3 when the pressed button is the Yes button 362 and P4 (P4 <P3) when the No button 363 is pressed. That is, in the paper search system, the satisfaction level is specified by analyzing the operation history of the operator, but in the repair manual search system, the operator inputs the satisfaction level directly.

<Example of a series of searches>
FIG. 31 shows a list of queries and graph changes in a series of searches, a match graph and a modified graph extracted from the store area 292 by the hint generation unit 25, and hints constructed by the hint generation unit 25. Hereinafter, a specific example of a series of searches will be described with reference to FIG.

  First, in the first search in a series of searches, the operator inputs a query Q511 “handle repair method” into the text area 311 of the search screen SC12 and performs a search. It is assumed that the search is performed again on the search screen SC14. A graph generated based on the query Q511 is a graph G511. Further, it is assumed that the hint generation unit 25 extracts the matching graph G511a and the modified graph G511b in the first search. In this case, based on the graph change from the coincidence graph G511a to the deformation graph G511b, additionally specify the major classification of “handle”. HT511 is generated and displayed on the re-search screen SC14.

  In the subsequent re-search, the operator inputs a query Q512 “automobile steering wheel repair method” into the text area 351 of the re-search screen SC14 with reference to the hint HT511, and a satisfactory search output is obtained. Since it was not obtained, it is assumed that the re-search is further performed on the re-search screen SC14. The graph corresponding to the query Q512 is the graph G512. In addition, it is assumed that the hint generation unit 25 extracts the matching graph G512a and the modified graph G512b in the first re-search. In this case, specify an alternative expression for “handle” based on the graph change from the matching graph G512a to the deformation graph G512b. Is generated and displayed on the re-search screen SC14.

  Further, in the second re-search, the operator refers to the hint HT512 and queries Q513 (converted to the graph G513) “repair method of the vehicle steering wheel or steering device” in the text area 351 of the re-search screen SC14. Suppose you search by typing and you get a satisfactory search output. At this time, the operator performs an operation such as switching the display 18 to the detailed information display screen SC15, the full text display screen SC16, or the print screen SC17, and then switches the display 18 to the satisfaction input screen SC18 and performs a series of searches. Exit.

  Note that “part” and “major classification” in the graphs G511 and G512 are examples of attributes. Further, “alternative” in the graph G513 is an example of a multi-word characteristic (synonymous relationship).

<Modification>
In the embodiment, the example in which RDF is adopted as the representation of the schema 26 that is the basis of the graphing is shown, but the present invention is not limited to this.

  In the embodiment, the start and end of a series of searches are detected independently in a plurality of steps, but in order to automatically start the next series of searches when the series of searches is completed, The start and end of the search may be detected simultaneously in a single step.

  In the embodiment, only the case where there is a single store area 292 is shown, but it is not hindered to provide a plurality of store areas and change the store area used for each operator or operator group.

2 is a block diagram showing a hardware configuration of a computer 10 for realizing the information search system 1. FIG. 2 is a block diagram showing a functional configuration of the information search system 1. FIG. It is a figure which shows the graph G101 which defines the schema used as the foundation of graphing of the query analysis part 22. FIG. It is a figure which shows graph G111 describing the vocabulary group [Ar (1), Ar (2), ..., Ar (Nr)] having the attribute Ar. 3 is a block diagram illustrating a functional configuration of a query analysis unit 22. FIG. It is a figure which shows an example of the graph G121 constructed | assembled by the graph construction part 222. FIG. It is a figure which shows an example of the graph G131 constructed | assembled by the graph construction part 222. FIG. It is a figure which shows an example of the graph G141 produced | generated by the basic structure extraction part 223. 4 is a block diagram illustrating a functional configuration of a hint generation unit 25. FIG. It is a figure which shows the coincidence graph G151. It is a figure which shows the deformation | transformation graph G161. 3 is a flowchart showing an overall operation of the information search system 1. It is a flowchart which shows the subroutine of graph conversion. It is a flowchart which shows the subroutine of hint production | generation. In the paper search system, it is a figure which shows the graph G201 which defines the schema used as the foundation of graphing of the query analysis part 22. FIG. It is a figure which shows the graph G211 which described the vocabulary group which has an author attribute. It is a screen transition diagram which shows the transition of the screen displayed on the display 18 of a paper search system. It is a figure which illustrates menu screen SC11. It is a figure which illustrates search screen SC12. It is a figure which illustrates search result list display screen SC13. It is a figure which illustrates detailed information display screen SC15. It is a figure which illustrates full-text display screen SC16. It is a figure which illustrates print screen SC17. It is a figure which illustrates re-search screen SC14. It is a figure which shows the specific example of a series of searches. It is a figure which shows the specific example of a series of searches. FIG. 27 is a diagram illustrating a graph G301 that defines a schema serving as a basis for graphing by the query analysis unit 22 in the repair manual search system. It is a figure which shows the graph G311 describing the vocabulary group which has the attribute (major classification) corresponding to the node N302 of the graph G301. It is a screen transition diagram which shows the transition of the screen displayed on the display 18 of a repair manual search system. It is a figure which illustrates satisfaction input screen SC18. It is a figure which shows the specific example of a series of searches.

Explanation of symbols

11 Bus 301 Transition link 311, 351 Text area 321 Search result list table 331 Detailed display table 341 Full text of paper

Claims (26)

An information search system that searches a database using an input query,
A graphing means for converting the query into a graph;
Storage means for storing a graph change history which is a history of the graph change;
An information retrieval system comprising: The information search system according to claim 1,
The graphing means converts the query into a graph based on a plurality of characteristics of query-containing words included in the query. The information search system according to claim 2,
The information retrieval system, wherein the plurality of characteristics include characteristics defined for a single word. In the information search system according to claim 3,
The information search system, wherein the characteristic defined for the single word is an attribute of the word. The information search system according to claim 2,
The information retrieval system, wherein the plurality of characteristics include characteristics defined for a plurality of sets of words. The information search system according to claim 5,
An information search system characterized in that the characteristic defined for the plurality of word pairs is any one of a specific abstract relation, an entire partial relation, a synonym relation, a complementary relation, an equivalent relation, an opposite relation, and an order relation. . The information search system according to any one of claims 2 to 6,
The information search system, wherein the graph is a combination of the query-containing words. The information search system according to any one of claims 2 to 7,
A dictionary that defines the characteristics of the word,
A plurality of characteristics of the query-containing word are determined based on the dictionary. The information search system according to claim 8,
The information search system, wherein the word whose characteristics are defined in the dictionary is a word extracted from data stored in the database. The information search system according to any one of claims 1 to 9,
A detecting means for detecting the start and end of a series of searches by the operator;
The information search system, wherein the graph change history is stored for each series of searches. The information search system according to claim 10,
A temporary storage means for temporarily storing the graph change history in the series of searches;
In response to the detection of the end, the graph change history stored in the temporary storage means is stored in the storage means. The information search system according to any one of claims 1 to 11,
An information retrieval system, wherein RDF is used to represent a schema as a basis for the graphing. The information search system according to any one of claims 1 to 11,
An information retrieval system, wherein the data stored in the database is represented by RDF. The information search system according to any one of claims 1 to 11,
An information search system, wherein the graph stored in the storage means is expressed by RDF. The information search system according to any one of claims 1 to 14,
Further comprising an evaluation specifying means for specifying an evaluation for a search result using the query of the conversion source of the graph change history,
An information retrieval system that stores the graph change history and the evaluation in association with each other. The information search system according to claim 15, wherein
An information search system, wherein the evaluation is an operator's satisfaction with a search result. The information search system according to claim 16, wherein
An information search system, wherein the satisfaction level is determined based on an operation history of an information search system of an operator. The information search system according to claim 16, wherein
The information search system further comprising satisfaction input means for allowing the operator to input the satisfaction. The information search system according to any one of claims 1 to 14,
An information search system further comprising a specifying unit for specifying a graph change history corresponding to a newly input query from the graph change history stored in the storage unit. The information search system according to claim 19, wherein
The specifying unit specifies a graph change history based on a degree of coincidence between a graph obtained by converting the query by the graphing unit and a graph included in the graph change history stored in the storage unit. An information retrieval system characterized by performing. The information search system according to claim 19, wherein
The graph change history stored in the storage means is associated with the evaluation,
The information search system, wherein the specifying means specifies a graph change history based on the evaluation. The information search system according to claim 21, wherein
The evaluation is the operator's satisfaction with the search result,
The information search system, wherein the specifying unit specifies a graph change history associated with a satisfaction level equal to or higher than a predetermined level. The information search system according to any one of claims 19 to 22,
An information search system further comprising auxiliary information generating means for generating auxiliary information related to the change of the query based on the graph change history specified by the specifying means. The information search system according to claim 23,
The information search system further comprising output means for outputting the auxiliary information so as to be recognized by an operator. An information retrieval method,
a) Query input process for entering a query;
b) a graphing step for converting the query into a graph;
c) a storing step of storing the change history of the graph in the storage means as a graph change history;
An information retrieval method comprising: An information retrieval program,
Execution by the computer of the information retrieval program is performed on the computer.
a) Query input process for entering a query;
b) a graphing step for converting the query into a graph;
c) a storing step of storing the change history of the graph in the storage means as a graph change history;
An information retrieval program characterized by causing

Images