JP2009064191A - Unit and method for information retrieval, program and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a unit and a method for information retrieval, a program and a recording medium which enable an inexperienced person to retrieve target information in a relatively short time. <P>SOLUTION: A hierarchical history storage means 13 successively stores operation information and retrieval result information obtained from a client 2 into a retrieval history database 14, as retrieval history data. A navigation pattern storage means 15 extracts a navigation pattern for each use of navigation from the retrieval history data, so as to store it in a navigation pattern management database 16. On receiving a navigation request 55 from the client 2, a navigation means 17 reads out the navigation pattern corresponding to the navigation request 55 from the navigation pattern management database 16, and generates navigation information 56 based on the readout navigation pattern, so as to transmit it to the client 2. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an information search support apparatus, an information search support method, a program, and a recording medium that support information search.

  In the Internet world, a portal site provides a plurality of search services so that a large amount of information can be quickly accessed, and many users use it. For example, there is a directory search service that provides a collection of links in which information is classified into a hierarchical structure or a directory structure, or a keyword search service that provides information hit by inputting a keyword.

  In the directory search service, many of the tasks for classifying information into each directory are performed manually, so that there are problems in that maintenance costs increase and information updates cannot catch up. In the keyword search service, there is a problem that a user cannot obtain an optimum result for information requested by the user unless an appropriate search keyword is input. Specifically, if the search condition is severe, for example, if it is an AND condition of many keywords, the information that can be obtained by the search becomes too small. Conversely, if the search condition is loose, there is information that can be obtained by the search. Become enormous.

  In particular, when the amount of information that can be obtained by searching is enormous, the user needs to perform a search again by adding a search keyword in order to narrow down the information. However, if the user is not familiar with the field of information to be searched, the user often does not know the words related to the information to be searched, and it is not easy to specify an appropriate keyword. In the worst case, even if the search is repeated many times, it may not be possible to narrow down the information requested by the user.

  There is a content navigation apparatus as a conventional technique that assists in facilitating information retrieval. This content navigation apparatus stores identification information for identifying content selected by the user from content indicated by a search result searched based on a keyword and the keyword in association with each other. Further, related keywords are grouped based on the correspondence relationship between the keyword and the content indicated by the identification information associated with the keyword. Thereafter, when a keyword is designated for the search, other keywords included in the group including the designated keyword are output as related keywords (see, for example, Patent Document 1).

  At present, information technology (abbreviated as “IT”) has been introduced in various companies, and electronic data has been developed throughout the company activities, and a large amount of data has been accumulated. In particular, the manufacturing industry faces the 2007 problem of baby boomers retiring all at once, that is, the problem of the loss of manufacturing technology and know-how, which are the company's assets. There is an urgent need to respond so that it can be reused for training.

  The manufacturing process in the manufacturing industry begins with the creation of a product concept, goes through concept design, detailed design, and process design, then moves to the production stage, where the final product is manufactured and shipped to the market. In the stage from concept design to detailed design in the upstream stage of this process, the design leader divides the overall functions required for the product into units, that is, sub-functions, and hierarchically structures the functions. Then, the layout and shape of the entire product are optimized, and detailed design of parts for realizing each function is assigned to the designer.

  Designers who are entrusted with detailed design of each function, in order to realize the functional specifications required for the parts, within multiple constraints such as quality, cost, development period and environmental measures, Repeat the trial and error to create the final part design. In the manufacturing process located at the upstream stage (hereinafter referred to as “design process”), the designer accumulates experience of success or failure and accumulates design know-how.

  In companies, the amount of information related to product development accumulated every day is enormous, and when searching for information by keyword as in the above-described search on the Internet, the user must enter an appropriate keyword. I can't get results.

JP 2005-309998 A

  The above-described conventional technology can present a keyword related to the keyword input by the user even if the user is not familiar with the words related to the information to be searched. it can. However, it is not always possible for an inexperienced person who is not familiar with words related to information to be searched to search for target information in a short time just like an expert by simply presenting keywords related to the input keyword. There is no problem.

  An object of the present invention is to provide an information search support device, an information search support method, a program, and a recording medium that support an inexperienced person so that target information can be searched in a relatively short time.

The present invention supports a search device that stores a plurality of pieces of search target information that is a search target and searches for search target information that satisfies a search condition for narrowing down the search target information from among the plurality of stored search target information. An information search support device that performs
Storage means for acquiring operation information representing an operation performed on the search device from the search device, and sequentially storing the acquired operation information in time series;
An extraction means for extracting an operation procedure satisfying a predetermined extraction condition from an operation procedure including a predetermined series of operations configured by operations indicated by operation information stored in time series in the storage means;
Search that supports search of information by generating operation support information for supporting an operation to be performed next in accordance with the operation procedure extracted by the extraction unit, and sequentially transmitting the generated operation support information to the search device. An information search support apparatus including a support means.

  According to the present invention, the operation procedure that satisfies the predetermined extraction condition is operated by a searcher whose degree of skill of the searcher who performs the search is equal to or higher than the predetermined skill level, and the number of times the operation procedure is executed is predetermined. Among the operation procedures that are more than the number of times, the operation procedure is the most frequently executed.

According to the present invention, the extraction means extracts an operation procedure for each of a plurality of predetermined search purposes indicating a search purpose,
The search support means generates operation support information according to an operation procedure that matches a search purpose instructed from the search device among operation procedures for each search purpose extracted by the extraction means.

  In the invention, it is preferable that the operation support information is operation information representing an operation to be performed next in a series of operations indicated by the operation procedure extracted by the extraction unit.

Further, in the present invention, the operation includes a search operation for instructing the search to search.
When the operation is the search operation, the operation information includes search condition information representing the search condition,
The search support means, when an operation to be performed next among the operations indicated by the operation procedure extracted by the extraction means is the search operation, operation information including search condition information of the search operation is displayed as operation support information. It is characterized by producing | generating as.

In the present invention, the search target information includes search target information identification information for identifying each search target information, and detailed information indicating details of the search target information,
The search includes a search operation for instructing the search device to perform a search, and a search result that displays detailed information from a search result indicating search target information searched by the search device that has received the search operation instruction. Including a selection operation to direct the selection of results,
The search result information representing the search result includes search target information identification information of the search target information indicated by the search result,
Further comprising grouping means for grouping a plurality of pieces of search target information stored in the search device according to a predetermined classification condition for classifying into a plurality of groups,
The storage means further acquires search result information representing a search result searched by the search device from the search device, sequentially stores the acquired search result information together with the operation information in time series. The group identification information of the group including the search target information indicated by the search target information identification information included in the acquired search result information among the group identification information for identifying the group grouped by the grouping means, Add to search result information and store it,
The extraction means extracts the operation procedure including the operation result indicated by the operation result information when extracting the operation procedure,
The search support means includes
Among the search results included in the operation procedure extracted by the extraction means, extract the group identification information added to the search result information of the search result selected by the selection operation,
When the operation indicated by the operation information acquired by the storage unit from the search device is a search operation, the search result indicated by the search result information related to the search operation among the search result information acquired by the storage unit from the search device The group identification information of the group including the search target information indicated by the search target information identification information included in the search result information of each search result includes the priority group included in the extracted group identification information, and the extracted group identification information. Search result information that represents search results that are classified into non-priority groups that are not included in and arranged so that the search results classified as priority groups are displayed in preference to the search results classified as non-priority groups. It is generated as support information.

In the present invention, the search target information includes search target information identification information for identifying each search target information, and detailed information indicating details of the search target information,
The search includes a search operation for instructing the search device to perform a search, and a search result that displays detailed information from a search result indicating search target information searched by the search device that has received the search operation instruction. Including a selection operation to direct the selection of results,
The search result information representing the search result includes search target information identification information of the search target information indicated by the search result,
Further comprising grouping means for grouping a plurality of pieces of search target information stored in the search device according to a predetermined classification condition for classifying into a plurality of groups,
The storage means further acquires search result information representing a search result searched by the search device from the search device, sequentially stores the acquired search result information together with the operation information in time series. The group identification information of the group including the search target information indicated by the search target information identification information included in the acquired search result information among the group identification information for identifying the group grouped by the grouping means, Add to search result information and store it,
The extraction means extracts the operation procedure including the operation result indicated by the operation result information when extracting the operation procedure,
The search support means includes
Among the search results included in the operation procedure extracted by the extraction means, group identification information added to the search result information of the search result selected by the selection operation is sequentially extracted for each selection operation,
When the operation indicated by the operation information acquired by the storage unit from the search device is a search operation, the search included in the search result information related to the search operation among the search result information acquired by the storage unit from the search device When the group identification information of the group including the search target information indicated by the target information identification information is included in the sequentially extracted group identification information, the search target information included in the group of the group identification information extracted next to the group identification information The search result information representing the search result including the search target information identification information is generated as operation support information.

According to another aspect of the present invention, there is provided a search device that stores a plurality of pieces of search target information that is a search target, and searches for the search target information that satisfies a search condition for narrowing down the search target information from the stored plurality of search target information. An information retrieval support method for supporting,
A storage step of acquiring operation information representing an operation performed on the search device from the search device and sequentially storing the acquired operation information in time series,
An extraction step for extracting an operation procedure satisfying a predetermined extraction condition from an operation procedure including a predetermined series of operations configured by operations indicated by operation information stored in time series in the storage step;
Search that supports search of information by generating operation support information for supporting an operation to be performed next in accordance with the operation procedure extracted in the extraction step, and sequentially transmitting the generated operation support information to the search device. An information search support method including a support step.

According to another aspect of the present invention, there is provided a search device that stores a plurality of pieces of search target information that is a search target, and searches for the search target information that satisfies a search condition for narrowing down the search target information from the stored plurality of search target information. In supporting,
A storage step of acquiring operation information representing an operation performed on the search device from the search device and sequentially storing the acquired operation information in time series,
An extraction step for extracting an operation procedure satisfying a predetermined extraction condition from an operation procedure including a predetermined series of operations configured by operations indicated by operation information stored in time series in the storage step;
Search that supports search of information by generating operation support information for supporting an operation to be performed next in accordance with the operation procedure extracted in the extraction step, and sequentially transmitting the generated operation support information to the search device. A program for causing a computer to execute the support step.

  The present invention is also a computer-readable recording medium in which the program is recorded.

  According to the present invention, a search device that stores a plurality of search target information that is a search target and searches for search target information that satisfies a search condition for narrowing down the search target information from the stored plurality of search target information. In the support, the storage means acquires operation information representing an operation performed on the search device from the search device, and the acquired operation information is sequentially stored in time series.

  Then, the extraction means extracts an operation procedure that satisfies a predetermined extraction condition from an operation procedure including a predetermined series of operations constituted by operations indicated by operation information stored in time series in the storage means, and performs a search. Operation support information for supporting an operation to be performed next is generated by the support means according to the operation procedure extracted by the extraction means, and the generated operation support information is sequentially transmitted to the search device, so that information Search is supported.

  Therefore, since operation support information based on an operation procedure that satisfies a predetermined extraction condition among operation procedures operated in the past is transmitted to the search device, the user follows the operation support information received and displayed by the search device. By performing the operation, even an inexperienced person who is not familiar with words related to information to be searched can search for target information in a relatively short time comparable to that of an experienced person.

  Further, according to the present invention, a plurality of search target information as search targets are stored, and a search for searching for search target information satisfying a search condition for narrowing down the search target information from the stored plurality of search target information. In supporting the apparatus, in the storing step, operation information representing an operation performed on the search apparatus is acquired from the search apparatus, and the acquired operation information is sequentially stored in time series.

  In the extraction step, an operation procedure that satisfies a predetermined extraction condition is extracted from an operation procedure including a predetermined series of operations configured by operations indicated by the operation information stored in time series in the storage step. In the search support step, operation support information for supporting an operation to be performed next is generated according to the operation procedure extracted in the extraction step, and the generated operation support information is sequentially transmitted to the search device, thereby Help search.

  Therefore, if the information search support method according to the present invention is applied, operation support information based on an operation procedure that satisfies a predetermined extraction condition among operation procedures operated in the past is transmitted to the search device. By searching according to the operation support information received and displayed by the search device, even an inexperienced person who is not familiar with the words related to the information to be searched can search for the target information in a relatively short time comparable to that of the experienced person. Can do.

Further, according to the present invention, a plurality of search target information as search targets are stored, and a search for searching for search target information satisfying a search condition for narrowing down the search target information from the stored plurality of search target information. In supporting the device,
A storage step of acquiring operation information representing an operation performed on the search device from the search device and sequentially storing the acquired operation information in time series,
An extraction step for extracting an operation procedure satisfying a predetermined extraction condition from an operation procedure including a predetermined series of operations configured by operations indicated by operation information stored in time series in the storage step;
Search that supports search of information by generating operation support information for supporting an operation to be performed next in accordance with the operation procedure extracted in the extraction step, and sequentially transmitting the generated operation support information to the search device. The support step can be provided as a program for causing a computer to execute the support step.

  The present invention can also be provided as a computer-readable recording medium in which the program is recorded.

  FIG. 1 is a diagram showing a configuration of a search navigation apparatus 1 according to an embodiment of the present invention. The search navigation device 1 that is an information search support device supports information search by a search device that includes an information terminal device (hereinafter referred to as “client”) 2, a search server 3, a data server 4, and a search target database 5. To do. The information search support method according to the present invention is processed by the search navigation apparatus 1.

  The search navigation device 1, the client 2, the search server 3, the data server 4, and the search target database 5 are connected by a communication line such as a LAN (Local Area Network), and can transmit / receive information to / from each other.

  The client 2 is a client device configured by a personal computer, for example. The user performs a search operation on the client 2, refers to a list of search results displayed on the client 2, and displays detailed information on the search results selected from the list on the client 2 for browsing. be able to. The client 2 transmits a navigation request 55 for requesting search support to the search navigation apparatus 1. Then, by receiving and displaying navigation information 56 that is operation support information for supporting the search transmitted from the search navigation device 1 in response to the transmitted navigation request 55, an operation for the search performed by the user. To help.

  The operations performed by the user on the client 2 include, for example, a LOGIN operation for authenticating the user, a LOGOUT operation for instructing the end of the operation on the client 2, and a search operation for instructing a search by inputting a search condition for search The SEARCH operation is a selection operation for selecting one search result from the search result list, acquiring detailed data 54 as detailed information of the selected search result, and instructing display of the acquired detailed data 54. A VIEW operation, a PAGE operation for instructing switching of pages when the number of search results included in the search result list is large and a plurality of pages are displayed, and a HELP operation for displaying an explanation of the operation are included.

  When the SEARCH operation is performed, the client 2 transmits a search request 51 for requesting a search to the search server 3 together with the search condition information indicating the search condition input by the SEARCH operation.

  The search server 3 is a server device configured by, for example, a computer. When receiving the search request 51 from the client 2, the search server 3 searches the search target database 5 for search condition information that satisfies the search condition indicated by the search condition information received together with the search request 51.

  The search target database 5 is a database that stores a plurality of pieces of search target information that are search targets. The plurality of search target information includes a data ID (Identification) that is search target information identification information for identifying each search target information, and detailed data representing the content of the search target information. When the search target information is, for example, document data, the data ID is, for example, a document name, and the detailed data is the content of the document. The search target information stored in the search target database 5 is not limited to document data, but the case where the search target information is document data will be described below as an example.

  The search server 3 transmits a search result list 52 that is a list of searched search results to the client 2. The search result is information indicating search target information that satisfies a search condition such as a keyword, and includes a data ID such as a document name.

  The client 2 displays the search result list 52 received from the search server 3. The user performs a VIEW operation and selects one search result from the search result list 52 displayed on the client 2 in order to view detailed data, for example, the contents of a document. When the VIEW operation is performed, the client 2 transmits a data acquisition request 53 for requesting detailed data corresponding to the search result selected by the VIEW operation to the data server 4. The data acquisition request 53 includes a data ID included in the search result selected by the VIEW operation.

  The data server 4 is a server device configured by, for example, a computer. When the data server 4 receives the data acquisition request 53 from the client 2, the data server 4 reads out the detailed data 54 corresponding to the data ID included in the data acquisition request 53 from the search target database 5. Send to client 2. The client 2 displays the detailed data 54 received from the data server 4. The user can view the detailed data 54 displayed on the client 2, for example, the contents of the document.

  The client 2 transmits operation information representing the operation performed on the client 2 and search result information representing individual search results included in the search result list 52 received from the search server 3 to the search navigation apparatus 1.

The search navigation apparatus 1 is configured by a computer, for example. The computer constituting the search navigation apparatus 1 receives information via an input device such as a keyboard and a mouse, an output device including a display device such as a display or a printing device such as a printer, and a communication line such as a LAN (Local Area Network). A communication device that transmits and receives, a time measuring device that measures time, a storage device that includes a semiconductor memory or a hard disk device, stores a program and data, and executes a program stored in the storage device to execute an input device and an output device , Central processing unit for controlling time measuring device and communication device (
Central Processing Unit: hereinafter referred to as “CPU”). The program is a program for controlling the search navigation apparatus 1 and may include an OS (Operating System) and an application program. The computer may be a generally known computer and will not be described in detail.

  Before describing the details of the search navigation apparatus 1, a method for calculating the similarity between data used by the search navigation apparatus 1, a clustering method, an association rule, and a graph mining technique will be described. The following are examples of these, and alternatives are possible.

  As a method for calculating the similarity between data, since many of the information retrieval methods in recent years express data with a vector space model, the search navigation apparatus 1 also expresses the data with a vector space model and calculates the similarity between the data. Calculate by the distance between vectors. When the search target information stored in the search target database 5 is, for example, document data, the importance is calculated for words such as noun phrases extracted by morphological analysis of the character string indicated by the document data, and the importance is high. A word can be expressed as a vector based on a vector space. Morphological analysis is an analysis technique in which a morpheme existing in a character string is decomposed, that is, decomposed into a minimum unit having meaning as a language, and a part of speech is given to the decomposed morpheme.

The degree of similarity between data is expressed by the magnitude of the distance between vectors, for example, the Euclidean distance. If the distance between vectors is large, the similarity between data is low, and if the distance between vectors is small, it is determined that the similarity between data is high. Here, the most widely used method for calculating the importance of words is the TF / IDF method (for example, Gerard Salton (Gerard
Salton), “Automatic Information Organization and Retrieval” (USA), McGraw-Hill, 1968).

  In the TF / IDF method, when the appearance frequency of words in data is represented by tf (term frequency), the total number of data is N, and the data including the corresponding word is n, the logarithm of the ratio, that is, log (N / n) is represented by idf (inverse document frequency). The product of tf and idf (hereinafter referred to as “tf · idf value”) is defined as the importance of the word. The tf · idf value of a word that appears in common in many data is a small value, and the tf · idf value of a word that appears only in specific data is a large value.

  In general, clustering methods can be broadly classified into hierarchical methods and non-hierarchical methods. The hierarchical method sequentially merges the two nearest clusters from the initial state where there are N clusters including only one object, based on a dissimilarity function between the clusters, for example, a distance function. This merging is repeated until all objects are merged into one cluster to generate a hierarchical structure.

The generated hierarchical structure is represented by a binary tree in which each target is each terminal node, and a cluster formed by merging is represented by a non-terminal node. A drawing of this binary tree is called a dendrogram. The hierarchical approach, the difference in the distance function D (C1, C2) of the cluster _{C 1} and the cluster _{C 2,} for example, the distance function D a (C1, C2), the shortest distance method expressed by the following equation (1) There is a longest distance method in which the distance function D (C1, C2) is represented by the following equation (2), and a group average method in which the distance function D (C1, C2) is represented by the following equation (3).

  These hierarchical methods can be applied when a distance D (C1, C2) between arbitrary objects is given, and when the object is described by a vector, the Euclidean distance between the vectors, etc. Can be applied.

  In the non-hierarchical method, an evaluation function is determined at the time of division, and division is repeated until the evaluation function reaches an optimum value. Since the total number of possible divisions is exponential with respect to the total number N of data, in practice, a suboptimal solution will be obtained. A k-means method (hereinafter referred to as “k-means method”), which is a typical non-hierarchical method, is a centroid, that is, an evaluation represented by the following equation (4) with the cluster centroid point ci as the cluster representative point. Divide into k clusters to minimize the function. k is a natural number less than N.

  The search for the optimum solution is performed by alternately repeating the classification into the target cluster and the recalculation of the representative points according to the following procedure.

  In procedure 1, representative points (c1 to ck) of k clusters are generated as initial values. In the procedure 2, the N objects are classified into the clusters of the nearest representative points ci, respectively, and the average of the objects included in each cluster is updated as the representative point ci of each new cluster. In the procedure 3, when the representative point ci does not change, the process is terminated, and when the representative point ci changes, the procedure returns to step 2.

  In the k-means method, since only a local optimum solution is obtained, it is common to randomly change the initial value and select a result that minimizes the evaluation function.

  Since the above-described clustering methods are all exploratory data analysis methods, the result of clustering is not absolute, that is, universal. The number of clusters to be divided must be determined by the user according to the purpose.

The calculation amount H in the clustering method is a function of H (N ² ), that is, N ^{2 in} the hierarchical method, whereas H (N × k), that is, a function of N × k, in the k-means method. The k-means method is superior in calculation cost. However, the k-means method requires that the number of cluster divisions and the representative point of the initial value of the cluster, that is, the seeds, be set in advance, whereas in the hierarchical method, the hierarchy of the dendrogram is moved up and down. By doing so, there is an advantage that the number of divisions can be adjusted.

  The association rule is one of data mining techniques that are techniques for finding knowledge from a large-scale database. A typical example of an association rule is a market basket analysis (hereinafter referred to as “basket analysis”) that extracts regularity that exists between attribute information held by data from data stored in a relational database. There is technology.

  Basket analysis finds the correlation between items in a shopping cart or basket at the same time based on POS (Point of Sales) information when a customer purchases the product using a shopping cart or basket in a supermarket. It is a technique. For example, the correlation rule “a customer who buys a diaper purchases beer at the same time” is expressed in the form “diaper → beer”. In this case, it is famous that sales increased by displaying beer next to diapers.

  Here, listing all combinations of products and checking one by one against the POS information as to whether or not the correlation rule is correct will check even a meaningless correlation rule. Therefore, in the basket analysis, efficiency is achieved by using the standard of support and confidence.

A priori algorithm (hereinafter referred to as “a priori method”) that extracts a combination of frequently appearing products at high speed using the support property has been proposed (for example, R. Agrwal, R. Agrwal) , R. Srikant), "Fast Algorithms for Mining Association Rules", 20th VLDB (Very Large Data Base) meeting minutes (In
Proceedings of the 20th VLDB Conference), 1994, p487-p499).

  Here, the support degree commonly used in the above-described basket analysis, a priori method, and the graph mining technique described later will be described. For example, taking basket analysis as an example, products that a customer purchases at a time in a supermarket are represented as a data set for each customer. For example, the customer A is represented as a data set {beer, magazine,...}, And the customer B is represented as a data set {milk, bread,.

  Each element of these data sets, for example, “beer” is called an item, and the data set itself is called a transaction. The support level is used when obtaining a frequent item set, that is, a subset of transactions, from the database. Frequently, it is included in many data sets. The support level is, for example, the ratio of the item set L, for example, the item set L including “diapers” and “beer” in all transactions, and can be calculated from the following equation.

Support level (L) = number of transactions including item set L / total number of transactions This expression represents the appearance frequency of item set L, and usually sets a threshold value of minimum support level, which is higher than the set threshold value. This can be used when taking out an item set that is a support level. An item set extracted using a threshold value of minimum support is called a frequent item set.

  Graph mining technology is used instead of the a priori method described above. That is, the a priori method described above is a method for extracting a frequently occurring subset at a high speed from regular log information. However, it cannot be used as a method for detecting frequent patterns from complex and heterogeneous large-scale data represented by a graph structure or a tree structure. Therefore, there is a graph mining technique as a method for extending the object of the basket analysis to the data of the graph structure, and this graph mining technique is used.

A typical example of graph mining technology is AGM (Apriori-based Graph Mining).
There are methods using algorithms (for example, Akihiro Higuchi, Takashi Hagio, Hiroshi Motoda, and two others, “A complete high-speed mining method for frequent graph patterns”, Journal of Artificial Intelligence, Vol. 15, No. 6, 2000)). As described above, the method using the AGM algorithm is a method in which the target of the basket analysis is extended to graph-structured data. The vertex and edge of the graph correspond to each item, and one graph corresponds to one transaction. To do. The support level (G) in the method using the AGM algorithm can be calculated by the following equation.
Support (G) = number of graphs including subgraph G / total number of graphs

The frequent item set becomes a frequent pattern set having a support level exceeding the minimum support level, that is, a subgraph set. References "Akihiro Higuchi, Takashi Hagio, Takashi Okada and Hiroshi Motoda," Applying the Apriori-based Graph
Mining Method to Mutagenesis Data Analysis), Journal of Computer Aided Chemistry, Vol. 2, 2001, p87-p92 ", inhibits mutagenic activity and correlation rules that may amplify mutagenic activity from data showing multiple organic compounds and mutagenic activity causing cancer. It is described that a possible association rule was successfully extracted.

  The search navigation apparatus 1 includes a cluster information storage unit 11, a cluster information management table 12, a hierarchy history storage unit 13, a search history database 14, a navigation pattern storage unit 15, a navigation pattern management database 16, and a navigation unit 17. The cluster information storage unit 11, the hierarchy history storage unit 13, the navigation pattern storage unit 15, and the navigation unit 17 are functions realized by a CPU included in a computer executing a program stored in a storage device.

  The cluster information storage means 11 that is a grouping means divides the search object information stored in the search object database 5, for example, document data into clusters that are groups using the clustering method described above, and the cluster information management table 12 is stored. create.

  FIG. 2 is a conceptual diagram of clusters divided by the cluster information storage unit 11. In FIG. 2, the entire document data 51 stored in the search target database 5 is roughly divided into three clusters C1 to C3. “C1 to C3” are cluster IDs that are group identification information for identifying clusters. Cluster 2 is further divided into two clusters, and cluster 3 is further divided into three clusters. If this state is expressed by a dendrogram, it is divided into three in the first layer and divided into six in the lower layer. In the cluster C1, documents with the document name “doc1” are classified.

  FIG. 3 is a conceptual diagram in which the cluster shown in FIG. 2 is divided by layers. One layer is a spatial concept that uses the same criterion as a criterion for calculating a measure of similarity between data, for example, tf · idf values. FIG. 3 shows two layers with layer IDs “L1” and “LC1” for identifying each layer. The layer with the layer ID “L1” is the same as the cluster division shown in FIG. 2, but in the layer with the layer ID “LC1”, the cluster C1 is divided into three clusters C1-1 to C1-3. , The document with the document name “doc1” is classified in the cluster C1-2.

  In the manufacturing industry, a standard format such as a standard format of product specifications is often used for technical data. By dividing the tf / idf value of a word into clusters as a measure of similarity with respect to this fixed format, it is possible to clearly divide a cluster including a product specification and a cluster not including a product specification.

  However, in a cluster including product specifications, the tf / idf values of the words that characterize each product specification are small, so further division may not be the intended result. Therefore, the range of the data set when calculating the tf · idf value is limited to the number of data belonging to the target cluster, and the tf · idf value may be recalculated, that is, the layer may be changed and recalculated. You may be able to get results.

  In FIG. 3, in the layer with the layer ID “L1”, the number of data included in the entire data set when calculating the tf · idf value is N, and in the layer with the layer ID “LC1”, the tf · idf value is calculated. An example in which the number of data included in the entire data set is m is shown. Here, m is a natural number smaller than N.

  FIG. 4 is a diagram showing an example of the cluster information management table 12 created by the cluster information storage unit 11. The cluster information management table 12 is configured by records, and each record includes items of a data ID and a cluster ID for each hierarchy. In the cluster ID item for each layer, the cluster ID and the layer ID into which the document having the document name indicated in the data ID item included in each record is classified are shown for each layer.

  In the cluster information management table 12 shown in FIG. 4, the document name is shown as the data ID, and the cluster ID and the layer ID corresponding to the cluster division shown in FIG. 3 are shown as the cluster ID and the layer ID for each hierarchy. ing.

  For the document with the data ID “doc1”, the cluster ID “C1” and the layer ID “L1” are shown as the first layer cluster ID and the layer ID, and the cluster ID “layer1” is shown as the second layer cluster ID and the layer ID. C1-2 "and layer ID" LC1 "are shown. Similarly, for the document with the data ID “doc2”, the cluster ID “C2” and the layer ID “L1” are shown as the first layer cluster ID and the layer ID, and the second layer cluster ID and layer ID are as follows: A cluster ID “C2-1” and a layer ID “L1” are shown. For the document with the data ID “docN”, the cluster ID “C3” and the layer ID “L1” are shown as the cluster ID and the layer ID of the first hierarchy, and the cluster ID “L1” is shown as the cluster ID and the layer ID of the second hierarchy. C3-1 "and layer ID" L1 "are shown.

  The hierarchical history storage means 13 as storage means acquires operation information and search result information from the client 2, and associates the acquired operation information and search result information in a time-series and hierarchical manner as a search history database. 14 stored.

  The time-series and hierarchical association can also be expressed by a graph represented using nodes and links. A node is an individual operation indicated by operation information. Alternatively, each search result indicated by the search result information is represented, and the link indicates a relationship between the nodes. Search result data representing a series of operations and search results from the LOGIN operation to the LOGOUT operation is called a search result data group, and one search result data group is represented by one graph.

  FIG. 5 is a diagram showing an example of a list 61 of attribute information for each node constituting the graph generated by the hierarchy history storage unit 13. The list 61 of attribute information for each node includes items of node name, link attribute, parent node attribute, and attribute value. The node name is a name indicating the type of the node, the link attribute is an attribute indicating the type of the link when associating with another node, and the parent node attribute is the name of the parent node to which the node is connected. The attribute value is an attribute determined for each type of node.

  The node name includes seven types of “LOGIN”, “LOGOUT”, “SEARCH”, “LIST”, “VIEW”, “PAGE”, and “HELP”. “LOGIN”, “LOGOUT”, “SEARCH”, “VIEW”, “PAGE”, and “HELP” are LOGIN operation, LOGOUT operation, SEARCH operation, and VIEW operation, respectively, among the operations indicated by the operation information received from the client 2. , PAGE operation and HELP operation are supported. “LIST” corresponds to each search result indicated by the search result information received from the client 2.

  There are two types of link attributes: time links and reference links. The time link is a link indicating that the relationship between two nodes is time series. The reference link is a link indicating that a relationship between two nodes is a parent-child relationship in which one node is expanded from the other node, for example, is referred to. The parent node attribute indicates a node immediately before the node in time series in the time link, and indicates a base node in which the node is expanded in the reference link. “NULL” indicates that there is no parent node.

  The link attribute, parent node attribute, and attribute value for each node name are “time link”, “NULL”, and “NULL” for the node name “LOGIN”. Similarly, “LOGOUT” is “time link”, “NULL” and “NULL”, and “SEARCH” is “time link”, “NULL” and “search formula, total number of search results”, “LIST” is “reference link”, “SEARCH, PAGE” and “data ID, cluster ID”, and “VIEW” is “reference link”, “LIST” and “data ID, cluster ID”. Yes, “PAGE” is “time link, reference link”, “SEARCH, PAGE” and “PAGE number”, and “HELP” is “time link”, “NULL” and “NULL”.

  FIG. 6 is a diagram illustrating an example of the search history data 62 stored in the search history database 14. The search history data is stored in units of records, and each record has seven fields of “SEQ”, “time”, “user ID”, “node name”, “time link”, “reference link”, and “attribute value”. Contains items.

  “SEQ” is a number for identifying a record stored in time series. “Time” is the time when the operation information or the search result information is received. The “user ID” is identification information for identifying the user who performed the operation (hereinafter also referred to as “searcher”), and is included in the operation information and the search result information. The “node name” is a node name corresponding to an operation or search result performed by the searcher. “Time link” and “reference link” are the numbers of “SEQ” of the parent node of the node indicated by the node name. If there is no parent, it is “NULL”. The “attribute value” is the same as the attribute value in the list 61 of attribute information for each node shown in FIG.

  The search history data 62 shown in FIG. 6 is a search related to a search performed by a searcher having a user ID “asai” between time “13: 00: 00: 00” and time “15: 00: 00: 00”. It is history data, and “user ID” is “asai”.

  SEQ "1" indicates that an operation corresponding to the node name "LOGIN" was performed at time "2006/08/25 13:00:00". “Time link”, “reference link”, and “attribute value” are all “NULL”. SEQ “2” indicates that an operation corresponding to the node name “HELP” was performed at the time “2006/08/25 13:00:01”. “Time link” is “1”, and “reference link” and “attribute value” are both “NULL”.

  SEQ “3” indicates that an operation corresponding to the node name “SEARCH” was performed at time “2006/08/25 13:03:00”. The “time link” is “2”, the “reference link” is “NULL”, and the “attribute value” is “AQUA *” and “10,000”. “AQUA *” is a search expression, and “10,000” is the total number of search results.

  SEQ “4” indicates that a search result corresponding to the node name “LIST” was obtained at time “2006/08/25 13:03:01”. “Time link” is “NULL”. “Reference link” is “3”, indicating that it is one of the search results indicated by the search result list searched by the SEARCH operation corresponding to the node name “SEARCH” of SEQ “3”. Yes. “Attribute values” are “doc1” and “C1-1”. “Doc1” is a data ID, and “C1-1” is a cluster ID.

  SEQ “5” indicates that a search result corresponding to the node name “LIST” was obtained at time “2006/08/25 13:03:01”. The “time link” is “NULL”, the “reference link” is “3”, and the “attribute value” is “doc2” and “C1-2”. “Doc2” is a data ID, and “C1-2” is a cluster ID.

  SEQ “14” indicates that an operation corresponding to the node name “VIEW” was performed at time “2006/08/25 13:08:00”. “Time link” is “NULL”. “Reference link” is “5”, which indicates that the search result corresponding to the node name “LIST” of SEQ “5” is selected and the detailed data is browsed. “Attribute values” are “doc2” and “C1-2”. “Doc2” is a data ID, and “C1-2” is a cluster ID.

  SEQ “15” indicates that the operation corresponding to the node name “PAGE” was performed at the time “2006/08/25 13:15:00”. “Time link” and “reference link” are both “3”, and the page for displaying the search result list searched by the SEARCH operation corresponding to the node name “SEARCH” of SEQ “3” is switched. Show. The “attribute value” is “10” and “NULL”. “10” is the PAGE number of the switched page.

  SEQ “26” indicates that an operation corresponding to the node name “LOGOUT” was performed at the time “2006/08/25 15:00:00”. “Time link” is “15”, and “reference link” and “attribute value” are both “NULL”.

  FIG. 7 is a diagram showing an example of the graph structure of the graph 71 corresponding to the search history data 62 shown in FIG. A series of search history data of the search history data 62 shown in FIG. 6 is one search history data group, and a graph 71 is a graph for the one search history data group.

  In the graph structure of the graph 71 shown in FIG. 7, the node name “LOGIN”, the node name “HELP”, the node name “SEARCH”, the node name “PAGE”, and the node name “LOGOUT” are connected by a time link in this order. Has been. Further, the node name “SEARCH” and a plurality of node names “LIST”, the node name “LIST” and the node name “VIEW”, the node name “SEARCH” and the node name “PAGE”, and the node name “PAGE” and a plurality of nodes The name “LIST” is connected by a reference link.

  The navigation pattern storage means 15 as extraction means first extracts from the search history database 14 search history data to be extracted for extracting a navigation pattern corresponding to the purpose of navigation. Next, a navigation pattern is extracted from the extracted search history data. The navigation pattern which is an operation procedure is a pattern of a series of operations and search results for assisting a search performed by a searcher in the client 2. The navigation pattern storage means 15 stores the extracted navigation pattern and navigation pattern management information for managing the navigation pattern in the navigation pattern management database 16.

  FIG. 8 is a diagram illustrating an example of a graph extracted by the navigation pattern storage unit 15 for each purpose of navigation. FIG. 8 shows three graphs 72 to 74 corresponding to three navigation applications. The three navigation uses are the type A operation flow, the type B search diversity, and the type C viewpoint change.

  The graph 72 is a graph along the flow of operations, and is a graph obtained by extracting a portion connected by a time link from the graph structure of the graph 71. A graph along the flow of operation is hereinafter referred to as a “time link graph”. The time link graph can be used when navigating to perform a HELP operation immediately after a LOGIN operation, for a user who is unfamiliar with the computer system and does not understand the operation of the system itself.

  In the graph structure of the graph 72 shown in FIG. 8, the node name “LOGIN”, the node name “HELP”, the node name “SEARCH”, the node name “PAGE”, and the node name “LOGOUTO” are connected by a time link in this order. Has been.

  The graph 73 is a graph structure representing the diversity of search, and is a graph obtained by extracting portions connected by reference links from the graph structure of the graph 71. A graph representing the diversity of search is hereinafter referred to as a “reference link graph”. For a user who is familiar with a field related to information desired to be acquired by search and a user who is not familiar with the information, for example, the variation of the document data selected by the VIEW operation corresponding to the node name “VIEW” included in the reference link graph is The users who are familiar are narrowed down to a small number of clusters, while the users who are not familiar are spread over a wide range of clusters. Therefore, navigation is performed for a user who is not well-versed so that document data included in a cluster that matches a search purpose described later is displayed with priority. Further, a cluster including document data selected by a savvy user by the VIEW operation is set as a reference cluster, and the reference cluster and the non-reference cluster are compared and navigated.

  The graph structure of the graph 73 shown in FIG. 8 includes a node name “SEARCH” and a plurality of node names “LIST”, a node name “LIST” and a node name “VIEW”, a node name “SEARCH” and a node name “PAGE”, The node name “PAGE” and a plurality of node names “LIST” are connected by a reference link.

  The graph 74 is a graph representing a change in viewpoint, and is a graph obtained by extracting the portion connected by the time link and the reference link from the graph structure of the graph 71. The graph representing the change in viewpoint is a graph in which nodes having a node name “SEARCH” (hereinafter referred to as “SEARCH node”) are connected by a time link in addition to the reference link graph. A graph representing a change in viewpoint is hereinafter referred to as a “viewpoint link graph”. The viewpoint link graph is used when navigating an operation procedure for searching for information that the user is interested in. For example, a viewpoint link graph for the purpose of tackling a problem in product development is used to display and navigate a document name included in a cluster to be referred to next.

  In the graph structure of the graph 74 shown in FIG. 8, a reference link graph is shown on one surface for each SEARCH node, and a plurality of surfaces are shown in time series from the top surface.

  As a method for extracting correlation rules between extracted graphs, the above-described graph mining technique may be used.

  FIG. 9 is a diagram illustrating an example of an event link graph 75 to which the viewpoint link graph illustrated in FIG. 8 is applied. The event link graph 75 is a graph in which the concept of the viewpoint link graph shown in FIG. 8 is developed and applied to a product development process in the manufacturing industry, for example, and the time axis scale is scaled up to the entire product development process.

  The SEARCH node in the viewpoint link graph is expressed by one virtual node (hereinafter referred to as “event node”) for each period of one day, one week, or one month. Then, a node having a node name “VIEW” connected by each SEARCH node and the reference link (hereinafter, also referred to as “VIEW node”) is represented by a graph as a node connected to the event node. This graph is called an “event link graph”.

  The event link graph 75 shown in FIG. 9 is a graph for a product development process that spans three months to one year from planning to development through shipment to the market. By analyzing the event link graph using the above-described graph mining technique, it is possible to extract a graph that represents the feature of the viewpoint change in the product development process. If detailed analysis is required, the event link graph may be analyzed using a scaled-down viewpoint link graph.

  FIG. 10 is a diagram showing the relationship between the navigation pattern extracted by the navigation pattern storage means 15 and the operation by the user.

  User X refers to the document data in the order of data ID “a1”, data ID “a2”, and data ID “a3”, and user Y has data ID “b1”, data ID “b2”, and data ID “ Document data is referred to in the order of “b3”. The user X and the user Y refer to the document data included in the cluster having the same cluster ID “Ci” for the data ID “a2” and the data ID “b2”. The document data included in different clusters, that is, the document data with the data ID “a3” and the document data with the data ID “b3” are referred to.

  If the document data included in the cluster with the cluster ID “Cj” is recommended next to the document data included in the cluster with the cluster ID “Ci”, the navigation includes the cluster data with the cluster ID “Ci” in the navigation. Next to the document data to be displayed, the user suggests or navigates to refer to the document data included in the cluster having the cluster ID “Cj”.

  FIG. 11 is a diagram illustrating an example of a list 63 of frequent patterns extracted by the navigation pattern storage unit 15. The frequent pattern list 63 shown in FIG. 11 is a partial graph having a large number of frequent patterns extracted from all type A graphs using the above-described graph mining technique, that is, a type A graph.

  The frequent pattern list 63 includes items of a graph ID, a pattern size, and a frequent pattern, that is, a partial graph. The graph ID is a number for identifying the frequent pattern. The frequent pattern is a pattern included in a number of graphs that are greater than or equal to a predetermined threshold, for example, greater than or equal to the minimum support, among patterns extracted from the graph. The pattern size is the number of node names included in the pattern. In the item of the frequent pattern, a series of node names included in the subgraph that is the frequent pattern is shown in time series.

  In the frequent pattern list 63 shown in FIG. 11, the pattern size of the graph ID “1” is “1”, and the frequent pattern is only “LOGIN”. The pattern size of the graph ID “2” is “2”, and “LOGIN” and “HELP” are shown in time series as frequent patterns. The pattern size of the graph ID “3” is “3”, and “LOGIN”, “HELP”, and “SEARCH” are shown in time series as frequent patterns. The pattern size of the graph ID “4” is “4”, and “LOGIN”, “HELP”, “SEARCH”, and “PAGE” are shown in time series as frequent patterns. The pattern size of the graph ID “5” is “5”, and “LOGIN”, “HELP”, “SEARCH”, “PAGE”, and “LOGOUT” are shown in time series as frequent patterns. The pattern size of the graph ID “6” is “1”, and the frequent pattern is only “HELP”. The pattern size of the graph ID “7” is “2”, and “HELP” and “SEARCH” are shown in time series as frequent patterns.

  Since the largest pattern size among the frequent patterns shown in FIG. 11 is “5”, the maximum frequent pattern is the frequent pattern with the graph ID “5”. When extracted using the graph mining technique, all the partial graphs included in the maximum frequent pattern are also extracted as the frequency pattern, so it is sufficient to extract only the maximum frequent pattern.

  The navigation pattern storage means 15 associates the extracted maximum frequent pattern with the search history data group to be extracted from the search history data stored in the search history database 14 and extracts the navigation pattern. It is stored in the navigation pattern management database 16 as a pattern.

  FIG. 12 is a diagram showing an example of the navigation pattern 64 stored in the navigation pattern management database 16. The navigation pattern 64 includes items of “number of rows”, “type”, “value”, “attribute 1”, and “attribute 2”.

  The navigation pattern 64 is composed of a plurality of records, and one record is represented by one line. “Number of rows” is a number for identifying a record. “Type” indicates the type of record, and there are two types, “pattern size” and “node name”. “Value” indicates the number of records whose type is “node name” among the records included in the navigation pattern 64 when the type is “pattern size”, and “node” when the type is “node name”. The name is shown.

  Attribute 1 and attribute 2 indicate the search history data name for identifying the search history data group from which the graph including the navigation pattern is extracted, and the SEQ number in the search history data group. The search history data name is identification information for identifying each search history data group, and is represented by, for example, “log1”, “log2”, and the like. Log1 is the search history data name of the search history data group composed of the search history data 62 shown in FIG.

  Although not shown in the search history data 62 shown in FIG. 6, the search history data name for each search history data group is stored in the search history database 14. Although only the attribute 1 and the attribute 2 are shown in the navigation pattern 64 shown in FIG. 12, if there are a plurality of search history data groups from which a graph including the navigation pattern is extracted, the attribute is shown by that number.

  In the navigation pattern 64 shown in FIG. 12, a record of the type “pattern size” is shown in the number of rows “1”, and the “value” is “5”. Both attribute 1 and attribute 2 are “NULL”, indicating that there is no corresponding information. A record of type “node ID” is shown in the number of rows “2” to the number of rows “6”, and “value”, “attribute 1”, and “attribute 2” have a value “LOGIN” for the number of rows “2”. , Attribute 1 “log1” and “SEQ = 1”, and attribute 2 “log2” and “SEQ = 1”. For the number of rows “3”, the values are “HELP”, attributes 1 “log 1” and “SEQ = 2”, and attributes 2 “log 2” and “SEQ = 2”. For the number of rows “4”, the values are “SEARCH”, attributes 1 “log” and “SEQ = 3”, and attributes 2 “log 2” and “SEQ = 3”. For the number of rows “5”, the values are “PAGE”, attributes 1 “log 1” and “SEQ = 15”, and attributes 2 “log 2” and “SEQ = 9”. For the number of rows “6”, the values are “LOGOUT”, attributes 1 “log 1” and “SEQ = 26”, and attributes 2 “log 2” and “SEQ = 18”.

  The navigation pattern 64 shown in FIG. 12 uses the maximum frequent pattern of the graph ID “5” in the frequent pattern list 63 shown in FIG. 11 as the navigation pattern. That is, although the navigation pattern is for the type A graph, the type B and type C graphs can also be represented by nodes and links, and therefore stored in the navigation pattern management database 16 with the same navigation pattern. Can do.

  FIG. 13 is a diagram showing an example of the navigation pattern management table 65 stored in the navigation pattern management database 16. The navigation pattern management table 65 is a table for managing navigation patterns stored in the navigation pattern management database 16 and stores navigation pattern management information in units of records.

  The navigation pattern management information includes items of “management ID”, “subject”, “specialty”, “minimum support”, “type”, “purpose”, and “navigation ID”. In the navigation pattern management table 65 shown in FIG. 13, in order to use the navigation pattern more effectively, the “type” of the graph and the “purpose” of the search are added to the items.

  “Management ID” is a number for identifying a record of navigation pattern management information. “Target person” is information indicating the degree of skill in the design work of the searcher. For example, a highly skilled engineer is represented as “skilled engineer”. “Professional” is information indicating another design process specialized by the searcher, and is represented by, for example, “functional design” and “circuit design”. The “minimum support level” displays a predetermined threshold value in% when extracting a frequent pattern using the graph mining technique. “Type” indicates the type of graph when the navigation pattern is extracted, and there are three types “A”, “B”, and “C”. “A” indicates type A, “B” indicates type B, and “C” indicates type C. “Purpose” is information indicating the purpose of searching for desired information. “Navigation ID” is navigation pattern identification information for identifying a navigation pattern, and is represented by, for example, the address of the first record of the navigation pattern stored in the navigation pattern management database 16.

  The items of “target person” and “specialty” in the navigation pattern management information can be acquired in cooperation with, for example, a personnel information database used in a company, or “target person” and “specialty”. ”And“ purpose ”can be explicitly input by a user who uses the system. Alternatively, it is also possible to provide a user with a system that narrows down the purpose of data search to a specific application, and confirms the “purpose” item by using the system.

  In the navigation pattern management table 65 shown in FIG. 13, for the management ID “1”, the target person “skilled engineer”, specialized “functional design”, minimum support level “80”%, type “A”, purpose “product” The mechanism of the hinge part of S is examined ", and the navigation ID" Navi1 "is shown. For the management ID “2”, the subject “skilled engineer”, specialized “functional design”, minimum support level “80”%, type “B”, purpose “examine mechanism of hinge part of product S”, and The navigation ID “Navi2” is shown. For management ID “3”, subject “expert engineer”, specialized “functional design”, minimum support level “80”%, type “C”, purpose “examine mechanism of hinge part of product S”, and The navigation ID “Navi3” is shown. For the management ID “4”, the subject “skilled engineer”, specialized “circuit design”, minimum support level “80”%, type “A”, purpose “examine heat generation problem of circuit board T”, and navigation The ID “Navi4” is shown. For the management ID “5”, the subject “skilled engineer”, specialized “circuit design”, minimum support level “80”%, type “B”, purpose “examine heat generation problem of circuit board T”, and navigation The ID “Navi5” is shown. For the management ID “6”, the subject “skilled engineer”, specialized “circuit design”, minimum support level “80”%, type “C”, purpose “examine heat generation problem of circuit board T”, and navigation The ID “Navi6” is shown.

  When the navigation means 17 serving as a search support means receives the navigation request 55 from the client 2, it starts navigation. First, the navigation pattern storage means 15 is notified that the navigation request 55 has been received, and the navigation pattern management database 16 is updated. The navigation request 55 includes a user ID, a search purpose, and a navigation method. The navigation method is used for navigation, for example.

  When the navigation pattern management database 16 is updated, a navigation method corresponding to a navigation method included in the navigation request 55, that is, a navigation application, is read from the navigation pattern management database 16, and the navigation information 56 is based on the read navigation pattern. And the generated navigation information 56 is transmitted to the client 2.

  Thus, the search navigation apparatus 1 can be used as search navigation when searching for data on various networks such as the Internet or an intranet. In particular, by applying it to an intranet within a company, it is possible to promote the reuse of a huge database of know-how accumulated in the company.

  FIG. 14 is a flowchart showing a processing procedure of cluster division processing executed by the cluster information storage unit 11. When the search navigation apparatus 1 is turned on and becomes operable, and the search object information stored in the search object database 5 such as document data can be acquired, the process proceeds to step A1.

  In step A1, the number n of clusters to be divided is set. n is a natural number less than N, and N is the number of document data stored in the search target database 5. In step A2, the document data stored in the search target database 5 is divided into n clusters using the clustering method described above. The clustering method used at this time is not limited to the clustering method described above, and other clustering methods may be used. In step A3, the variable i is set to “1”.

  In step A4, a cluster information management table (referred to as “cluster information table” in the figure) 12 of data belonging to cluster i is set. That is, the data ID, the cluster ID, and the layer ID are set in the cluster information management table 12 for the document classified into the cluster whose cluster ID is Ci. The data ID of each document data is set as the data ID, and the cluster ID “Ci” and the layer ID of the cluster into which the document is classified are set as the cluster ID of the first hierarchy. The layer ID set in step A4 is the layer ID of the first hierarchy, for example “L1”.

  In step A5, it is determined whether or not the variable i is equal to the number of clusters “n”. If the variable i is equal to the number of clusters “n”, the process proceeds to step A6. If the variable i is not equal to the number of clusters “n”, the process proceeds to step A17. In step A6, the variable i is set to “1”.

  In step A7, it is determined whether or not the cluster having the cluster ID Ci is to be subdivided, that is, further divided. When subdividing, the process proceeds to step A8, and when not subdividing, the process proceeds to step A15. The determination of whether or not to divide is performed by, for example, displaying a message inquiring whether or not to divide again on the client 2 and an input response to the displayed message, for example, a response instructing to divide or Judgment is made based on a response instructing not to subdivide.

  In step A8, the number k of clusters to be subdivided is set. k is a natural number less than n. In step A9, it is determined whether or not to re-divide by changing the layer. When the layer is changed and the subdivision is performed, the process proceeds to step A10, and when the layer is changed and the subdivision is not performed, the process proceeds to step A11. The determination of whether or not to re-divide by changing the layer is performed by, for example, displaying a message for inquiring whether or not to re-divide by changing the layer on the client 2, and an input response to the displayed message, For example, the determination is made based on a response instructing to change the layer and re-divide or a response instructing to change and not re-divide the layer.

  In step A10, the similarity measure is changed to a data set of cluster i. That is, the similarity measure used in the clustering method is changed to a similarity measure for document data included in a cluster whose cluster ID is Ci. In step A11, the document data included in the cluster whose cluster ID is Ci is divided into k clusters using the clustering method described above.

  In step A12, the variable j is set to “1”. In step A13, the cluster information management table 12 of data belonging to the cluster j is set. That is, among the subdivided clusters, for a document classified as a cluster whose cluster ID is Ci-j, the cluster ID “Ci-j” and the layer ID are added to the second layer cluster ID of the cluster information management table 12. Set. The layer ID set in step A13 is the layer ID of the first layer, and if it is the cluster ID “C1-1”, for example, “LC1”.

  In step A14, it is determined whether or not the variable j is equal to the number of clusters “k” in the subdivision. If the variable j is equal to the number of clusters “k”, the process proceeds to step A15. If the variable j is not equal to the number of clusters “k”, the process proceeds to step A18. In step A15, it is determined whether or not the variable i is equal to the number of clusters “n”. If the variable i is equal to the number of clusters “n”, the process proceeds to step A16. If the variable i is not equal to the number of clusters “n”, the process proceeds to step A19.

  In step A16, it is determined whether or not to end. For example, a message for inquiring whether or not to end the cluster division process is displayed on the client 2. When the response to the displayed message is “YES”, it is determined to end, and the cluster division process is ended. If the response to the displayed message is “NO”, it is determined not to end, and the process returns to step A6. In step A17, “1” is added to the variable i, and the process returns to step A4. In step A18, “1” is added to the variable j, and the process returns to step A13. In step A19, "1" is added to the variable i, and the process returns to step A7.

  In the flowchart shown in FIG. 14, the layer only shows processing up to the second layer. However, when there is a layer after the third layer, re-division is performed by processing in the same manner as the second layer.

  FIG. 15 is a flowchart showing a processing procedure of cluster ID detection processing executed by the hierarchy history storage unit 13. When the search result information is received from the client 2, the process proceeds to step B1.

  In step B1, the variable i is set to “1”. In step B2, the i-th data ID is read from the cluster information management table 12. In step B 3, it is determined whether or not the data ID included in the search result indicated by the search result information received from the client 2 matches the data ID read from the cluster information management table 12. If they match, the process proceeds to step B4. If they do not match, the process proceeds to step B5.

  In step B4, the address of the record in which the matched data ID is registered among the addresses of the records in which the data IDs are registered in the cluster information management table 12 is stored in the storage device, and the cluster ID detection process is terminated. To do. In step B5, it is determined whether or not the variable i is equal to the number of document data “N”. If the variable i is not equal to the number of document data “N”, the process proceeds to step B6. If the variable i is equal to the number of document data “N”, the cluster ID detection process is terminated. In Step B6, “1” is added to the variable i, and the process returns to Step B2.

  FIG. 16 is a flowchart showing a processing procedure of search history storage processing executed by the hierarchy history storage means 13. When the operation information or the search result information is received from the client 2, the process proceeds to Step C1. In the case of search result information, after performing the cluster ID detection process shown in FIG. 15, the process proceeds to step C1.

  In step C1, “1” is added to “SEQ”, which is the SEQ number. In step C2, the current time is acquired from the timing device, and the acquired time is set to “time” of the record indicated by “SEQ” in the records of the search history database 14, and the received operation information or search result When the user ID included in the information is set in the “user ID” of the record indicated by “SEQ” and the received information is operation information, the node ID corresponding to the operation indicated by the received operation information is received. Is the search result information, the node ID “LIST” is set to “node ID” of the record indicated by “SEQ”.

  In step C3, the “time link”, “reference link”, and “attribute value” of the record indicated by “SEQ” are initialized. In Step C4, it is stabilized whether or not the node ID set in “Node ID” of the record indicated by “SEQ” is “LOGIN”. If it is “LOGIN”, the process proceeds to Step C12, and if it is not “LOGIN”, the process proceeds to Step C5.

  In step C5, it is stabilized whether or not the node ID set in the “node ID” of the record indicated by “SEQ” is “SEARCH”. If it is “SEARCH”, the process proceeds to Step C14, and if it is not “SEARCH”, the process proceeds to Step C6. In Step C6, it is stabilized whether or not the node ID set in “Node ID” of the record indicated by “SEQ” is “PAGE”. If it is “PAGE”, the process proceeds to Step C15, and if it is not “PAGE”, the process proceeds to Step C7.

  In step C7, it is stabilized whether or not the node ID set in “node ID” of the record indicated by “SEQ” is “HELP”. If it is “HELP”, the process proceeds to Step C16, and if it is not “HELP”, the process proceeds to Step C8. In Step C8, it is stabilized whether or not the node ID set in “Node ID” of the record indicated by “SEQ” is “LOGOUT”. If it is “LOGOUT”, the process proceeds to Step C17, and if it is not “LOGOUT”, the process proceeds to Step C9.

  In step C9, the cluster ID of the data ID is acquired. Specifically, the cluster ID of the record indicated by the address stored in the storage device in step B4 of the flowchart shown in FIG. 15 is acquired from the cluster information management table 12. In step C10, it is stabilized whether or not the node ID set in the “node ID” of the record indicated by “SEQ” is “LIST”. If it is “LIST”, the process proceeds to Step C18, and if it is not “LIST”, the process proceeds to Step C11. In step C11, it is stabilized whether or not the node ID set in the “node ID” of the record indicated by “SEQ” is “VIEW”. If it is “VIEW”, the process proceeds to step C19. If it is not “VIEW”, the search history storage process is terminated.

  In Step C12, it is determined whether or not the value of the variable “SEQ” is equal to “1”. If the value of the variable “SEQ” is equal to “1”, the process proceeds to step C20, and if the value of the variable “SEQ” is not equal to “1”, the process proceeds to step C13. In step C13, the SEQ number of the previous record, that is, the last record is set in the “time link” of the record indicated by “SEQ”, and the process proceeds to step C20. In step C14, the value of the variable “parent SEQ” is set in “time link” of the record indicated by “SEQ”, the search expression and the total number of search results are set in “attribute value”, and the process proceeds to step C20. The variable “parent SEQ” indicates the SEQ number of the record of the parent node ID. The search formula and the total number of search results are included in the operation information related to the SEARCH operation.

  In step C15, the value of the variable “parent SEQ” is set in “time link” and “reference link” of the record indicated by “SEQ”, the PAGE number is set in “attribute value”, and the process proceeds to step C20. The PAGE number is included in the operation information related to the PAGE operation. In step C16, the value of the variable “parent SEQ” is set in the “time link” of the record indicated by “SEQ”, and the process proceeds to step C20. In step C17, the value of the variable “parent SEQ” is set in the “time link” of the record indicated by “SEQ”, and the search history storage process is terminated.

  In step C18, the value of the variable “parent SEQ” is set in “reference link” of the record indicated by “SEQ”, and the data ID and the cluster ID acquired in step C9 are set in “attribute value”. The search history storage process is terminated. The data ID is included in operation information related to the LIST operation and the VIEW operation. In step C19, the SEQ number of the record of the node ID of the parent LIST is set in the “reference link” of the record indicated by “SEQ”, the data ID in the “attribute value”, and the cluster ID acquired in step C9. Set the search history storage process. In step C20, the number “SEQ” is set in the variable “parent SEQ”, and the search history storage process ends. Steps C1 to C20 are storage steps.

  FIG. 17 is a flowchart showing a processing procedure of navigation pattern extraction processing executed by the navigation pattern storage means 15. When the navigation means 17 is notified that the navigation request 55 has been received from the client 2, the process proceeds to step D1.

  In step D1, target data is extracted. That is, the search history data to be extracted for extracting the navigation pattern is extracted from the search history database 14. The search history data may be extracted by grouping the search history data in advance, and extracting the search history data in groups from the grouped groups.

  For example, the search history data is divided into groups such as skilled engineers, middle-level engineers, and young engineers in product development, and the search history data is extracted in units of groups. At this time, the skill level of the engineer may be acquired in cooperation with a personnel information database used in the company. The skill level is an index indicating the level of skill and is represented by the years of experience in a specific technical field. For example, engineers having 15 years of experience or more are grouped as skilled engineers, engineers having 5 or more years but less than 15 years as medium-sized engineers, and engineers having less than 5 years as young engineers. Alternatively, search history data of searches performed by engineers who have received awards from inside and outside the company for product development achievements in the past can be grouped as a group of excellent engineers. It is also possible to group by department.

  In step D2, an analysis application is selected. Specifically, a navigation method included in the navigation request 55 received from the client 2, that is, a type corresponding to the purpose of navigation is selected from types A to C, and a graph of the selected type is generated. In step D3, the variable i is set to “1”. In step D4, it is determined whether or not the selected type is type A. If it is type A, the process proceeds to step D13, and if it is not type A, the process proceeds to step D5.

  In step D5, it is determined whether or not the selected type is type B. If it is type B, the process proceeds to step D14. If it is not type B, the process proceeds to step D6. In step D6, it is determined whether or not the selected type is type C. If it is type C, the process proceeds to step D15, and if it is not type C, the process proceeds to step D7.

  In step D7, the generated graphs are merged. In Step D8, it is determined whether or not the variable i is equal to “N”. N is the total number of target data extracted in step D1. If the variable i is equal to “N”, the process proceeds to step D9. If the variable i is not equal to “N”, the process proceeds to step D16.

  In step D9, a minimum support level for extracting a frequent pattern using a graph mining technique is set. The minimum support level can also be set automatically. If the minimum support level is set to an unlimitedly small value from 100% to 0%, a frequent pattern having a small appearance frequency and a small pattern size is obtained. As a result, the number of search spaces becomes enormous and the calculation cost also increases. In order to avoid an increase in calculation cost, a time limit for calculation time to be calculated is set in advance using graph mining technology, and the minimum support level immediately before exceeding the limit time is set to the minimum support level that can be set. It may be a limit value.

  In step D10, a frequent pattern having a support level exceeding the minimum support level is extracted from the merged graph using a graph mining technique. For example, in order to extract a frequent pattern of operations similarly performed by 80% or more of skilled engineers, search history data of all skilled engineers is extracted from the search history database 14, and a graph is drawn based on the extracted search history data. After generating, the minimum support degree is set to 80%, and the graph mining technique is used to extract the data.

  In step D11, the maximum frequent pattern having the largest pattern size is extracted from the extracted frequent patterns. In step D12, the extracted maximum frequent pattern is stored in the navigation pattern management database 16 in association with the search history data from which the maximum frequent pattern is extracted, and the navigation pattern extraction process is terminated.

  In Step D13, a time link graph is generated, and the process proceeds to Step D7. In step D14, a reference link graph is generated, and the process proceeds to step D7. In Step D15, a viewpoint link graph is generated, and the process proceeds to Step D7. In step D16, “1” is added to the variable i, and the process returns to step D4. Steps D1 to D16 are extraction steps.

  FIG. 18 is a flowchart showing a processing procedure of navigation processing executed by the navigation means 17. When the navigation pattern storage means 15 informs that the update of the navigation pattern management database 16 is completed, the process proceeds to step E1.

  In step E1, user information of the user indicated by the user ID included in the received navigation request 55, for example, information such as skill level and specialty, is acquired from, for example, a personnel information database used in the company, and further included in the navigation request 55. Get search purpose. Alternatively, information such as the skill level and specialty of the user may be input by the client 2 and included in the navigation request 55 and transmitted.

  In step E2, the navigation method included in the navigation request 55 is acquired. There are three types of navigation systems, that is, navigation applications, types A to C. Type A is navigation for command operations corresponding to the flow of operations, and Type B is navigation for increasing the priority of clusters matching the search purpose corresponding to search diversity and reflecting them in the search result list. Type C is navigation for navigating related data in order in accordance with a search purpose corresponding to a change in viewpoint. The user can select one or a plurality of navigation methods on the client 2.

  In step E3, a navigation pattern that matches the skill level and specialty of the user, the search purpose, and the navigation method is extracted from the navigation pattern management database 16.

  In step E4, it is determined whether or not the navigation method is type A. If it is type A, the process proceeds to step E7, and if it is not type A, the process proceeds to step E5. In step E5, it is determined whether or not the navigation method is type B. If it is type B, the process proceeds to step E8. If it is not type B, the process proceeds to step E6. In step E6, it is determined whether or not the navigation method is type C. If it is type C, the process proceeds to step E9. If it is not type C, the navigation process is terminated.

  In step E7, a type A process for navigating the command operation is called, and when the type A process ends, the process proceeds to step E5. In step E8, the type B process that raises the priority of the cluster that matches the search purpose and reflects it in the search result list is called. When the type B process ends, the process proceeds to step E6. In step E9, a type C process for navigating related data in order according to the search purpose is called, and when the type C process ends, the navigation process ends. Steps E1 to E9 are search support steps.

  FIG. 19 is a flowchart showing the processing procedure of the type A processing called from the navigation processing. When the type A process is called from step E7 in the flowchart shown in FIG. 18, the process proceeds to step F1.

  In step F1, the variable i is set to “1”. In step F2, the i-th node name is read from the navigation pattern extracted in step E3 of the flowchart shown in FIG. In step F3, it is determined whether or not the read node name is “LOGIN”. If the read node name is “LOGIN”, the process proceeds to step F4. If the read node name is not “LOGIN”, the process proceeds to step F6.

  In step F4, it is determined whether or not the (i + 1) th node name exists. If the (i + 1) -th node name exists, the process proceeds to step F5. If the (i + 1) -th node name does not exist, the type A process ends. In step F5, “1” is added to the variable i, and the process returns to step F2. In step F6, it is determined whether or not the i-th node name is “SEARCH”. If it is “SEARCH”, the process proceeds to step F9, and if it is not “SEARCH”, the process proceeds to step F7.

  In step F7, the user is instructed to execute the command of the i-th node name. That is, navigation information 56 indicating that the operation corresponding to the i-th node name is recommended is transmitted to the client 2 and displayed. In step F8, it is determined whether or not the command has been executed. When the operation information is received from the client 2, it is determined that the command is executed, and the process proceeds to Step F4. When the operation information is not received from the client 2, it is determined that the command is not executed, and the type A process is terminated.

  In step F9, the variable j is set to “1”. In step F10, the search history data name (referred to as “logID” in the figure) and the SEQ number are read from the jth attribute of the i-th node name. In step F11, a search expression is read out from the search history data stored in the search history database 14 from the search history data attribute value indicated by the read search history data name and SEQ number.

  In step F12, the user is suggested to execute SEARCH using the retrieved search expression. That is, the navigation information 56 that recommends execution of the SEARCH operation to which the read search formula is substituted is transmitted to the client 2 and displayed. In step F13, it is determined whether or not the command has been executed. When the operation information is received from the client 2, it is determined that the command is executed, and the process proceeds to Step F4. When the operation information is not received from the client 2, it is determined that the command is not executed, and the process proceeds to Step F14.

  In step E14, it is determined whether or not the (j + 1) th attribute exists. If the (j + 1) th attribute exists, the process proceeds to step F15. If the (j + 1) th attribute does not exist, the type A process is complete | finished. In Step F15, “1” is added to the variable j, and the process returns to Step F10.

  For example, if the navigation pattern extracted in step E3 of the flowchart shown in FIG. 18 is a navigation pattern registered as type A, the navigation pattern is a sequence of commands when a skilled engineer uses a search device. That is, the operation flow is registered. Therefore, by operating according to the processing procedure of the type A process shown in FIG. 19, the user can perform an operation according to the navigation pattern of the operation flow performed by the skilled engineer.

  Thus, since the navigation information 56 is operation information indicating an operation to be performed next in a series of operations indicated by the operation procedure extracted by the navigation pattern storage unit 15, that is, the navigation pattern, it is indicated by the navigation information 56. The target information can be retrieved in a shorter time by simply performing the operation.

  Thus, the operation includes a SEARCH operation for instructing the search device to perform a search, and the operation information includes search condition information indicating the search condition when the operation is a SEARCH operation. Then, when the operation to be performed next is the SEARCH operation among the operation procedures extracted by the navigation unit 17 by the navigation pattern storage unit 15, that is, the operation indicated by the navigation pattern, the operation information including the search condition information of the SEARCH operation However, since the navigation information 56 is generated, the SEARCH operation can be searched under the same search conditions as in the previous search.

  FIG. 20 is a flowchart showing a processing procedure of type B processing called from the navigation processing. When the type B process is called from step E8 in the flowchart shown in FIG. 18, the process proceeds to step G1.

  In step G1, variables i, j, and k are each set to “1”. In step G2, the i-th node name is read from the navigation pattern extracted in step E3 of the flowchart shown in FIG. In step G3, it is determined whether or not the read node name is “LIST”. If it is “LIST”, the process proceeds to Step G8, and if it is not “LIST”, the process proceeds to Step G4.

  In step G4, it is determined whether or not the (i + 1) th node name exists. If the (i + 1) th node name exists, the process proceeds to step G7, and if the (i + 1) th node name does not exist, the process proceeds to step G5. In step G5, it is determined whether or not SEARCH has been executed. When the operation information indicating the SEARCH operation is acquired from the client 2, it is determined that the SEARCH is executed, and the process proceeds to step G16. When the operation information indicating the SEARCH operation is not acquired from the client 2, it is determined that the SEARCH is not executed, and the step G6 is performed. Proceed to

  In step G6, it is determined whether or not the next command has been executed. When the next operation information is received from the client 2, it is determined that the next command has been executed, and the process returns to step G5. When the next operation information is not received from the client 2, it is determined that the next command is not executed, and type B The process ends. In step G7, “1” is added to the variable i, and the process returns to step G2.

  In step G8, it is determined whether or not the (i + 1) th node name exists. If the (i + 1) th node name exists, the process proceeds to step G9, and if the (i + 1) th node name does not exist, the process proceeds to step G5. In step G9, the (i + 1) th node name is read from the navigation pattern extracted in step E3 of the flowchart shown in FIG.

  In step G10, it is determined whether or not the (i + 1) th node name is “VIEW”. If it is “VIEW”, the process proceeds to Step G13, and if it is not “VIEW”, the process proceeds to Step G11. In step G11, among the search history data stored in the search history database 14, the search history data name stored in the attribute 1 of the i-th node name and the cluster ID of the attribute value of the search history data indicated by the SEQ number Are registered in the k-th non-reference array (referred to as “non-reference (k)” in FIG. 20), which is a one-dimensional array. In Step G12, “1” is added to the variable k, and the process proceeds to Step G4.

  In step G13, among the search history data stored in the search history database 14, the search history data attribute value of the search history data name and the SEQ number stored in the attribute 1 of the (i + 1) th node name are set. The cluster ID is registered in the j-th reference array (referred to as “reference (j)” in FIG. 20), which is a one-dimensional array. In step G14, “1” is added to the variable j. In Step G15, “1” is added to the variable i, and the process proceeds to Step G4.

  In step G16, a search result list is acquired. That is, in step G5, search result list information representing a search result list by the SEARCH operation indicated by the operation information acquired from the client 2 is acquired from the client 2. The number of search results included in the search result list is M. M is a natural number. In step G17, the variable i is set to “1”. In step G18, the cluster ID of the data ID included in the i-th search result is acquired from the search results included in the search result list indicated by the acquired search result list information.

  In step G19, it is determined whether or not the acquired cluster ID is included in the reference array. If the acquired cluster ID is included in the reference array, the process proceeds to step G20, and if the acquired cluster ID is not included in the reference array, the process proceeds to step G23. In step G20, the i-th search result is moved to the top of the search result list. For example, the i-th search result is classified into priority groups to be displayed with priority. In step G21, it is determined whether or not the variable i is smaller than “M”. If the variable i is smaller than “M”, the process returns to step G18. If the variable i is not smaller than “M”, the process proceeds to step G22.

  In step G22, a search result list is presented to the user, and the process proceeds to step G6. For example, when classified into the priority group and the non-priority group, the navigation information 56 displays a search result list arranged so that the search results classified into the priority group are displayed in preference to the search results classified into the non-priority group. Is transmitted to the client 2 and displayed, and the process proceeds to step G6. In step G23, it is determined whether or not the acquired cluster ID is included in the non-reference array. If the acquired cluster ID is included in the non-reference array, the process proceeds to step G24. If the acquired cluster ID is not included in the non-reference array, the process proceeds to step G21. In step G24, the i-th search result is moved to the lower part of the search result list, and the process proceeds to step G21. For example, the i-th search result is classified into a non-priority group that is not preferentially displayed, and the process proceeds to step G21.

  For example, if the navigation pattern extracted in step E3 of the flowchart shown in FIG. 18 is a navigation pattern registered as type B, the navigation pattern matches a theme by a skilled engineer familiar with the search target theme. The cluster when the data to be matched and the data that does not match is sorted is registered. Therefore, by operating according to the processing procedure of the type B process shown in FIG. 20, the priority of the cluster corresponding to the theme can be increased and reflected in the search result list, and the user has no difficulty in the desired theme. You can search for information.

  Thus, the search target information includes a data ID for identifying each search target information and detailed data representing the details of the search target information, and the operation is a SEARCH for instructing the search device to perform a search. And a VIEW operation for instructing selection of a search result for displaying detailed data from search results indicating search target information searched by the search device that has received an instruction for the operation and SEARCH operation. The search result information representing the result includes the data ID of the search target information indicated by the search result.

  A plurality of search target information stored in the search device is grouped according to a predetermined classification condition for classifying into a plurality of clusters by the cluster information storage means 11, and searched by the search device by the hierarchy history storage means 13. Search result information representing the search results obtained is further acquired from the search device, the acquired search result information is sequentially stored in time series together with the operation information, and the search result information is grouped by the cluster information storage means 11 Among the cluster IDs for identifying the acquired clusters, the cluster ID of the cluster including the search target information indicated by the data ID included in the acquired search result information is added to the search result information and stored.

  Then, when the operation procedure, that is, the navigation pattern is extracted by the navigation pattern storage means 15, the operation result indicated by the operation result information is included and extracted. The navigation unit 17 extracts the cluster ID added to the search result information of the search result selected by the VIEW operation among the search results included in the navigation pattern extracted by the navigation pattern storage unit 15.

  Furthermore, when the operation indicated by the operation information acquired by the hierarchy history storage unit 13 from the search device by the navigation unit 17 is a SEARCH operation, among the search result information acquired by the hierarchy history storage unit 13 from the search device, The search result indicated by the search result information related to the SEARCH operation includes a cluster ID of a cluster including search target information indicated by the data ID included in the search result information of each search result, a priority group included in the extracted cluster ID, A search that represents a search result that is classified into a non-priority group that is not included in the extracted cluster ID and arranged so that the search result classified into the priority group is displayed with priority over the search result classified into the non-priority group. Result information is generated as navigation information 56.

  Therefore, it is possible to preferentially display previously searched search target information among the search target information indicated by the search result information.

  FIG. 21 is a flowchart showing the processing procedure of the type C processing called from the navigation processing. When the type C process is called from step E9 in the flowchart shown in FIG. 18, the process proceeds to step H1.

  In step H1, variables i, j, and k are set to “1”, respectively. In step H2, the i-th node name is read from the navigation pattern extracted in step E3 of the flowchart shown in FIG. In step H3, it is determined whether or not the read node name is “SEARCH”. If it is “SEARCH”, the process proceeds to step H9. If it is not “SEARCH”, the process proceeds to step H4.

  In step H4, it is determined whether or not the (i + 1) th node name exists. If the (i + 1) th node name exists, the process proceeds to step H8, and if the (i + 1) th node name does not exist, the process proceeds to step H5. In Step H5, the constant L is set to “j−1”. In step H6, it is determined whether SEARCH has been executed. When the operation information indicating the SEARCH operation is acquired from the client 2, it is determined that SEARCH has been executed, and the process proceeds to step H17. When the operation information indicating the SEARCH operation is not acquired from the client 2, it is determined that the SEARCH is not executed, and step H7 is performed. Proceed to

  In step H7, it is determined whether or not the next command has been executed. When the next operation information is received from the client 2, it is determined that the next command has been executed, and the process returns to step H6. When the next operation information is not received from the client 2, it is determined that the next command is not executed, and type C The process ends. In Step H8, “1” is added to the variable i, and the process returns to Step H2.

  In step H9, it is determined whether or not the (i + 1) th node name exists. If the (i + 1) th node name exists, the process proceeds to step H10. If the (i + 1) th node name does not exist, the process proceeds to step H4. In Step H10, the (i + 1) th node name is read from the navigation pattern extracted in Step E3 of the flowchart shown in FIG.

  In Step H11, it is determined whether or not the (i + 1) th node name is “VIEW”. If it is “VIEW”, the process proceeds to Step H12, and if it is not “VIEW”, the process proceeds to Step H15. In step H12, among the search history data stored in the search history database 14, the attribute value of the search history data indicated by the search history data name and the SEQ number stored in the attribute 1 of the (i + 1) th node name is set. The cluster ID is registered in the k-th column (referred to as “reference (j, k)” in FIG. 21) of the j-th row of the reference two-dimensional array that is a two-dimensional array.

  In Step H13, “1” is added to the variable k. In Step H14, “1” is added to the variable i, and the process returns to Step H9. In Step H15, “1” is added to the variable j. In Step H16, “1” is added to the variable i, and the process proceeds to Step H4.

  In step H17, a data ID is selected from the search result list. That is, the operation information of the VIEW operation in which one search result is selected from the search result list by the SEARCH operation indicated by the operation information acquired from the client 2 is acquired from the client 2, and the data ID is acquired from the acquired operation information. The operation information of the VIEW operation includes information on a search result selected by the VIEW operation, and the search result includes a data ID. In step H18, the variable j is set to “1”. In step H19, a cluster ID corresponding to the acquired data ID is acquired from the cluster information management table 12.

  In step H20, the variable k is set to “1”. In step H21, it is determined whether or not the acquired cluster ID is equal to the cluster ID of the kth column of the jth row of the reference two-dimensional array. When the acquired cluster ID is equal to the cluster ID of the kth column of the jth row of the reference two-dimensional array, the process proceeds to step H26. If the acquired cluster ID is not equal to the cluster ID of the kth column of the jth row of the reference two-dimensional array, the process proceeds to step H22.

  In step H22, it is determined whether or not the (k + 1) th column in the jth row of the reference two-dimensional array exists. If the (k + 1) th column of the jth row of the reference two-dimensional array exists, the process proceeds to step H24, and if the (k + 1) th column of the jth row of the reference two-dimensional array does not exist, Proceed to step H23. In step H23, it is determined whether or not the variable j is smaller than a constant L. If the number j is smaller than the constant L, the process proceeds to step H25, and if the number j is not smaller than the constant L, the process proceeds to step H7.

  In Step H24, “1” is added to the variable k, and the process returns to Step H21. In Step H25, “1” is added to the variable j, and the process returns to Step H19. In step H26, it is determined whether or not the number j is smaller than a constant L. If the number j is smaller than the constant L, the process proceeds to step H27, and if the variable j is not smaller than the constant L, the process proceeds to step H7.

  In Step H27, the cluster ID is acquired from the (j + 1) th row of the reference two-dimensional array (referred to as “reference (j + 1,)” in FIG. 21). In step H28, a data ID having the same cluster ID as the acquired cluster ID is acquired from the cluster information management table 12. In step H29, the next candidate list is presented to the user, and the process returns to step H17. That is, the navigation information 56 which shows the search result including the data ID acquired in step H28 as a next candidate list is generated, the generated navigation information 56 is transmitted to the client 2 for display, and the process returns to step H17.

  That is, based on the navigation pattern, from the search results included in the search result list displayed by the SEARCH operation, every time browsing is performed by the VIEW operation, the cluster IDs of the browsed search results are stored in time series. . Thereafter, when a SEARCH operation is performed, for each search result in the search result list by the SEARCH operation, the cluster ID of the cluster including the document data of the data ID included in the search result matches the stored cluster ID. The search result including the data ID of the document data included in the cluster of the cluster ID stored next to the cluster ID is displayed as the next search result list.

  In other words, the document data belonging to the same cluster as the document data cluster displayed in the next search after browsing the document data of the same category as the displayed document data is displayed and navigated as the next search result list. Thereafter, the same navigation is repeated each time document data is selected from the search result list.

  As described above, when the navigation pattern extracted in step E3 of the flowchart shown in FIG. 18 is a navigation pattern registered as type C, the navigation pattern is focused on when dealing with countermeasures related to product development. Data search history is registered. Therefore, by operating according to the processing procedure of the type C process shown in FIG. 21, it is possible to trace the point of focus when the problem has been tackled in the past, and the user can obtain a hint for solving the problem.

  Thus, the search target information includes a data ID for identifying each search target information and detailed data representing the details of the search target information, and the operation is a SEARCH for instructing the search device to perform a search. And a VIEW operation for instructing selection of a search result for displaying detailed data from search results indicating search target information searched by the search device that has received an instruction for the operation and SEARCH operation. The search result information representing the result includes the data ID of the search target information indicated by the search result.

  A plurality of search target information stored in the search device is grouped according to a predetermined classification condition for classifying into a plurality of clusters by the cluster information storage means 11, and searched by the search device by the hierarchy history storage means 13. Search result information representing the search results obtained is further acquired from the search device, the acquired search result information is sequentially stored in time series together with the operation information, and the search result information is grouped by the cluster information storage means 11 Among the cluster IDs for identifying the acquired clusters, the cluster ID of the cluster including the search target information indicated by the data ID included in the acquired search result information is added to the search result information and stored.

  When an operation procedure, that is, a navigation pattern is extracted by the navigation pattern storage unit 15, the operation result indicated by the operation result information is included and extracted, and the navigation unit 17 extracts the navigation pattern extracted by the navigation pattern storage unit 15. Among the included search results, the cluster ID added to the search result information of the search result selected by the VIEW operation is sequentially extracted for each VIEW operation.

  Furthermore, when the operation indicated by the operation information acquired by the hierarchy history storage unit 13 from the search device by the navigation unit 17 is a SEARCH operation, among the search result information acquired by the hierarchy history storage unit 13 from the search device, When the cluster ID of the cluster including the search target information indicated by the data ID included in the search result information related to the SEARCH operation is included in the sequentially extracted cluster ID, it is included in the cluster of the cluster ID extracted next to the cluster ID. Search result information representing the search result including the data ID of the search target information to be generated is generated as the navigation information 56.

  That is, when the cluster including the search target information obtained as a result of the search is a cluster including the search target information whose detailed data is referred to in the previous search, the SEARCH operation for searching the search target information of the cluster The search target information included in the cluster of search target information obtained as a result of the next SEARCH operation is displayed. Therefore, it is possible to display the search target information of the cluster of the search target information searched by the SEARCH operation performed immediately after the search target information is referred to by the SEARCH operation performed previously. That is, it is possible to search with the same focus as the focus in the search process performed in the past.

  According to the present invention, the user's operation procedure when searching for information is stored in the search history database 14 in a hierarchical relationship, thereby enabling “operation flow”, “variety of search”, and “change in viewpoint”. It is possible to generate a graph representing the user's operation procedure from various viewpoints such as “”. By obtaining a correlation rule between graphs of the generated graph, a navigation pattern corresponding to each application can be extracted. Therefore, an inexperienced person can reach information that can be searched in a short time by an experienced person by operating according to the navigation based on the navigation pattern.

  Further, in storing a plurality of pieces of search target information to be searched, and assisting a search apparatus that searches for search target information satisfying a search condition for narrowing down the search target information from among the plurality of stored search target information. The hierarchy history storage unit 13 acquires operation information representing an operation performed on the search device from the search device, and the acquired operation information is sequentially stored in time series.

  Then, the navigation pattern storage means 15 performs an operation that satisfies a predetermined extraction condition from among a predetermined operation sequence constituted by operations indicated by operation information stored in time series by the hierarchy history storage means 13. A procedure, that is, a navigation pattern is extracted, and navigation information 56 for supporting an operation to be performed next is generated by the navigation means 17 according to the navigation pattern extracted by the navigation pattern storage means 15, and the generated navigation information 56 is generated. Are sequentially transmitted to the search device, so that search of information is supported.

  Therefore, among the operation procedures operated in the past, the operation procedure satisfying the predetermined extraction condition, that is, the navigation information 56 based on the navigation pattern is transmitted to the search device, so that the user is received and displayed by the search device. By operating according to the navigation information 56, even an inexperienced person who is not familiar with words related to information to be searched can search for target information in a relatively short time comparable to that of an experienced person.

  As described above, the predetermined extraction condition is an operation procedure including a predetermined series of operations constituted by operations indicated by the operation information stored in time series by the hierarchy history storage unit 13, and a searcher who performs a search. The operation procedure that is operated by a searcher whose skill level is equal to or higher than a predetermined skill level and the number of times that the operation procedure is executed is equal to or more than a predetermined number of times, that is, an operation procedure that is executed most frequently, that is, a frequent pattern. Maximum frequent pattern.

  Therefore, when more than a certain number of operations are required, information can be searched using the same operation procedure that was most frequently operated by highly skilled searchers, and the desired information can be searched in a shorter time. can do.

  In this manner, the navigation pattern storage means 15 extracts an operation procedure, that is, a navigation pattern for each of a plurality of predetermined search purposes indicating the purpose of search, and the navigation means 17 extracts each search purpose extracted by the navigation pattern storage means 15. Navigation information 56 is generated in accordance with a navigation pattern that matches the search purpose specified by the search device. Therefore, since an operation procedure according to the purpose of the search can be shown, if the purpose is clear, the target information can be retrieved more quickly.

Further, in storing a plurality of pieces of search target information to be searched, and assisting a search apparatus that searches for search target information satisfying a search condition for narrowing down the search target information from among the plurality of stored search target information. ,
In step C1 to step C20 of the flowchart of the search history storage process shown in FIG. 16, operation information representing an operation performed on the search device is acquired from the search device, and the acquired operation information is sequentially stored in time series. To do.

  Steps D1 to D16 in the flowchart of the navigation pattern extraction process shown in FIG. 17 are previously configured by operations indicated by the operation information stored in steps C1 to C20 in the flowchart of the search history storage process shown in FIG. An operation procedure that satisfies a predetermined extraction condition, that is, a navigation pattern is extracted from an operation procedure including a predetermined series of operations.

  In steps E1 to E9 of the flowchart of the navigation process shown in FIG. 18, the next operation to be performed is supported according to the navigation patterns extracted in steps D1 to D16 of the flowchart of the navigation pattern extraction process shown in FIG. The navigation information 56 is generated, and the generated navigation information 56 is sequentially transmitted to the search device, thereby supporting the information search.

  Therefore, if the information search support method according to the present invention is applied, among the operation procedures operated in the past, the operation procedure satisfying the predetermined extraction condition, that is, the navigation information 56 based on the navigation pattern is transmitted to the search device. Therefore, if the user operates according to the navigation information 56 received and displayed by the search device, even the inexperienced person who is not familiar with the words related to the information to be searched can target information in a relatively short time as the experienced person. Can be searched.

  The program for controlling the search navigation apparatus 1 is also a program for causing a computer to execute each step of the information search support method according to the present invention. Therefore, the present invention can be provided as a program for causing a computer to execute each step of the information search support method.

  In the above-described embodiment, the program is stored in a storage device of a computer such as a semiconductor memory or a hard disk device. However, the program is not limited to these storage devices, and a computer-readable recording medium. May be recorded. The recording medium may be a recording medium that can be read by providing a program reading device as an external storage device (not shown) and inserting the recording medium therein, or may be a storage device of another device. .

  Any recording medium may be used as long as the stored program is accessed from a computer and executed. Alternatively, any recording medium may be configured such that the program is read, the read program is stored in the program storage area of the storage device, and the program is executed. Further, it may be downloaded from another device via a communication network and stored in the program storage area. The download program is stored in advance in a storage device of a computer, or installed in a program storage area from another recording medium.

Recording media configured to be separable from the main body include, for example, tape-based recording media such as magnetic tape and cassette tape, magnetic disks such as flexible disks and hard disks, CD-ROM (Compact Disk Read Only Memory), and MO (Magneto Optical). disk), MD (Mini Disc), DVD (Digital Versatile Disk) and other optical disk recording media, IC (Integrated Circuit) cards (including memory cards), optical cards and other card recording media, or masks ROM, EPROM (Erasable Programmable
Read Only Memory), EEPROM (Electrically Erasable Programmable Read Only
Memory) or a recording medium that carries a fixed program including a semiconductor memory such as a flash ROM. Therefore, the present invention can be provided as a computer-readable recording medium in which a program for causing a computer to execute each step of the information search support method is recorded.

It is a figure which shows the structure of the search navigation apparatus 1 which is one Embodiment of this invention. 3 is a conceptual diagram of a cluster divided by a cluster information storage unit 11. FIG. It is the conceptual diagram which divided | segmented the cluster shown in FIG. 2 by the layer. It is a figure which shows an example of the cluster information management table 12 produced by the cluster information storage means 11. FIG. It is a figure which shows an example of the list 61 of the attribute information for every node which comprises the graph produced | generated by the hierarchy log | history storage means. It is a figure which shows an example of the search history data 62 memorize | stored in the search history database. It is a figure which shows an example of the graph structure of the graph 71 corresponding to the search history data 62 shown in FIG. It is a figure which shows the example of the graph extracted according to the use of the navigation by the navigation pattern storage means. It is a figure which shows an example of the event link graph 75 which applied the viewpoint link graph shown in FIG. It is a figure which shows the relationship between the navigation pattern extracted by the navigation pattern storage means 15, and operation by a user. It is a figure which shows an example of the list 63 of the frequent pattern extracted by the navigation pattern storage means 15. FIG. It is a figure which shows an example of the navigation pattern 64 memorize | stored in the navigation pattern management database. It is a figure which shows an example of the navigation pattern management table 65 memorize | stored in the navigation pattern management database. 4 is a flowchart showing a processing procedure of cluster division processing executed by cluster information storage means 11. It is a flowchart which shows the process sequence of the cluster ID detection process performed by the hierarchy log | history storage means. 7 is a flowchart showing a processing procedure of search history storage processing executed by a hierarchy history storage means 13; 4 is a flowchart showing a processing procedure of navigation pattern extraction processing executed by a navigation pattern storage means 15. 4 is a flowchart showing a processing procedure of navigation processing executed by a navigation means 17. It is a flowchart which shows the process sequence of the type A process called from a navigation process. It is a flowchart which shows the process sequence of the type B process called from a navigation process.

It is a flowchart which shows the process sequence of the type C process called from a navigation process.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Search navigation apparatus 2 Client 3 Search server 4 Data server 5 Search target database 11 Cluster information storage means 12 Cluster information management table 13 Hierarchy history storage means 14 Search history database 15 Navigation pattern storage means 16 Navigation pattern management database 17 Navigation means 51 Search Request 52 Search result list 53 Data acquisition request 54 Detailed data 55 Navigation request 56 Navigation information

Claims (10)

Information search support that supports a search device that stores a plurality of search target information as search targets and searches the search target information satisfying a search condition for narrowing down the search target information from the stored plurality of search target information A device,
Storage means for acquiring operation information representing an operation performed on the search device from the search device, and sequentially storing the acquired operation information in time series;
An extraction means for extracting an operation procedure satisfying a predetermined extraction condition from an operation procedure including a predetermined series of operations configured by operations indicated by operation information stored in time series in the storage means;
Search that supports search of information by generating operation support information for supporting an operation to be performed next in accordance with the operation procedure extracted by the extraction unit, and sequentially transmitting the generated operation support information to the search device. An information search support apparatus comprising: support means.   The operation procedure that satisfies the predetermined extraction condition is an operation that is operated by a searcher whose degree of proficiency of a searcher performing a search is equal to or higher than a predetermined skill level, and the number of times that the operation procedure is executed is equal to or greater than a predetermined number of times. The information search support apparatus according to claim 1, wherein the operation search procedure is the most frequently executed procedure. The extraction means extracts an operation procedure for each of a plurality of predetermined search purposes indicating the purpose of search,
The search support means generates operation support information in accordance with an operation procedure that matches a search purpose instructed from the search device among operation procedures for each search purpose extracted by the extraction means. Item 3. The information search support apparatus according to Item 1 or 2.   The operation support information is operation information representing an operation to be performed next in a series of operations indicated by the operation procedure extracted by the extraction unit. Information retrieval support device described in one. The operation includes a search operation for instructing the search device to perform a search,
When the operation is the search operation, the operation information includes search condition information representing the search condition,
The search support means, when an operation to be performed next among the operations indicated by the operation procedure extracted by the extraction means is the search operation, operation information including search condition information of the search operation is displayed as operation support information. The information search support apparatus according to claim 4, wherein the information search support apparatus is generated as follows. The search target information includes search target information identification information for identifying each search target information, and detailed information representing details of the search target information,
The search includes a search operation for instructing the search device to perform a search, and a search result that displays detailed information from a search result indicating search target information searched by the search device that has received the search operation instruction. Including a selection operation to direct the selection of results,
The search result information representing the search result includes search target information identification information of the search target information indicated by the search result,
Further comprising grouping means for grouping a plurality of pieces of search target information stored in the search device according to a predetermined classification condition for classifying into a plurality of groups,
The storage means further acquires search result information representing a search result searched by the search device from the search device, sequentially stores the acquired search result information together with the operation information in time series. The group identification information of the group including the search target information indicated by the search target information identification information included in the acquired search result information among the group identification information for identifying the group grouped by the grouping means, Add to search result information and store it,
The extraction means extracts the operation procedure including the operation result indicated by the operation result information when extracting the operation procedure,
The search support means includes
Among the search results included in the operation procedure extracted by the extraction means, extract the group identification information added to the search result information of the search result selected by the selection operation,
When the operation indicated by the operation information acquired by the storage unit from the search device is a search operation, the search result indicated by the search result information related to the search operation among the search result information acquired by the storage unit from the search device The group identification information of the group including the search target information indicated by the search target information identification information included in the search result information of each search result includes the priority group included in the extracted group identification information, and the extracted group identification information. Search result information that represents search results that are classified into non-priority groups that are not included in and arranged so that the search results classified as priority groups are displayed in preference to the search results classified as non-priority groups. The information search support device according to claim 1, wherein the information search support device is generated as support information. The search target information includes search target information identification information for identifying each search target information, and detailed information representing details of the search target information,
The search includes a search operation for instructing the search device to perform a search, and a search result that displays detailed information from a search result indicating search target information searched by the search device that has received the search operation instruction. Including a selection operation to direct the selection of results,
The search result information representing the search result includes search target information identification information of the search target information indicated by the search result,
Further comprising grouping means for grouping a plurality of pieces of search target information stored in the search device according to a predetermined classification condition for classifying into a plurality of groups,
The storage means further acquires search result information representing a search result searched by the search device from the search device, sequentially stores the acquired search result information together with the operation information in time series. The group identification information of the group including the search target information indicated by the search target information identification information included in the acquired search result information among the group identification information for identifying the group grouped by the grouping means, Add to search result information and store it,
The extraction means extracts the operation procedure including the operation result indicated by the operation result information when extracting the operation procedure,
The search support means includes
Among the search results included in the operation procedure extracted by the extraction means, group identification information added to the search result information of the search result selected by the selection operation is sequentially extracted for each selection operation,
When the operation indicated by the operation information acquired by the storage unit from the search device is a search operation, the search included in the search result information related to the search operation among the search result information acquired by the storage unit from the search device When the group identification information of the group including the search target information indicated by the target information identification information is included in the sequentially extracted group identification information, the search target information included in the group of the group identification information extracted next to the group identification information The information search support device according to claim 1, wherein search result information representing a search result including the search target information identification information is generated as operation support information. Information search support that supports a search device that stores a plurality of search target information as search targets and searches the search target information satisfying a search condition for narrowing down the search target information from the stored plurality of search target information A method,
A storage step of acquiring operation information representing an operation performed on the search device from the search device and sequentially storing the acquired operation information in time series,
An extraction step for extracting an operation procedure satisfying a predetermined extraction condition from an operation procedure including a predetermined series of operations configured by operations indicated by operation information stored in time series in the storage step;
Search that supports search of information by generating operation support information for supporting an operation to be performed next in accordance with the operation procedure extracted in the extraction step, and sequentially transmitting the generated operation support information to the search device. An information search support method comprising a support step. In supporting a search device that stores a plurality of search target information that is a search target, and searches for search target information that satisfies a search condition for narrowing down the search target information from the stored plurality of search target information.
A storage step of acquiring operation information representing an operation performed on the search device from the search device and sequentially storing the acquired operation information in time series,
An extraction step for extracting an operation procedure satisfying a predetermined extraction condition from an operation procedure including a predetermined series of operations configured by operations indicated by operation information stored in time series in the storage step;
Search that supports search of information by generating operation support information for supporting an operation to be performed next in accordance with the operation procedure extracted in the extraction step, and sequentially transmitting the generated operation support information to the search device. A program for causing a computer to execute the support step.   10. A computer-readable recording medium on which the program according to claim 9 is recorded.

Images