JP3637756B2 - Information search device, information search method, and recording medium - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an information search apparatus, an information search method, and a recording medium, and in particular, an information search apparatus and an information search method for searching for information in a hyper document system configured by nodes as links of information and links between nodes. And a recording medium.
[0002]
[Prior art]
In the hyper document system, information to be handled is divided into several small units (nodes), and they are associated and organized (information indicating such association is called a link). For example, on the Internet, a hyper document system is constructed by WWW (World Wide Web). WWW information is described in HTML (Hyper Text Markup Language). This HTML has no semantic restrictions on links between nodes. As described above, a system that does not have a semantic constraint on the link between nodes has an advantage that the document author (author) can determine the content and the link structure at will. With such a hyper document system, the document reader (reader) can follow the link structure constructed by the document author and access all information provided by the document author.
[0003]
By the way, the amount of information in hyper document systems such as the Internet is enormous. Therefore, in order for the document reader to find necessary information, a system that supports information retrieval is required. There are the following two conventional techniques for supporting such a search.
[0004]
The first conventional technique scans as many nodes as possible in advance (randomly), prepares a search index for each node, and presents what matches a query (a combination of keywords) from the document reader. Is. In addition, as an elemental technology for search index creation and query matching, vector space model (G. Salton & J. Allan, Text Retrieval Using the Vector Processing Model, in Proc. Of SDAIR94) Has been devised.
[0005]
The second conventional technique scans as many nodes as possible in advance (randomly) and assigns them to a tree-structured directory classified by topic. The document reader searches the directory for a topic considered to contain the desired information, and accesses the target information from there. As an elemental technology for realizing this technology, automatic document classification methods using natural language processing (for example, P. Jacobs, Joining Statistics with NLP for Text Categorization, in Proc. Of Applied-ACL92) have been proposed. Yes. Furthermore, an automatic document classification method (United States Patent: 5526443, T. Nakayama (FXPAL), Method and apparatus for highlighting and categorizing documents using coded word tokens, issue date: 1996.6.11) was also devised. Yes.
[0006]
[Problems to be solved by the invention]
However, in these conventional technologies, since one node (for example, one HTML document) is used as one search target unit, it is possible to grasp the essence of a hyper document system in which a concept is expressed by a structure of nodes and links. However, the following problems have occurred.
[0007]
How many nodes are divided and how they are structured depends on the preference of the document author, but in a search that uses a node as a unit, it is structured on the hypernetwork. A node group cannot be regarded as information having a single semantic group. That is, in the search according to the prior art, only the semantically incomplete information fragment is set as the search target, and the context is not reflected in the search.
[0008]
For example, when information with one semantic group created by one document author is divided into a plurality of HTML documents and expressed by performing a document search according to the conventional technique, each HTML document is individually searched. It becomes a target. Here, when the document reader searches for information similar to “concept A”, even though the information created by the document author is generally similar to “concept A”, the divided individual nodes are “concept A”. If it is not similar to this information, this information (or some nodes) is not detected.
[0009]
In addition, if one node is a search target unit, there is a problem that a concept representing a search request cannot be expressed by a structure on a hyper network.
Furthermore, when the document author divides information having a single semantic group into a plurality of nodes and structures them, the conventional search outputs each node individually, resulting in redundancy. is there. When a series of HTML documents created by a single document author as information having a single semantic group is output individually, the amount of search results increases, and the document reader for searching for documents that meet the purpose Labor increases.
[0010]
Therefore, starting from a certain node, the features of Nth-order nodes (N = 2, 3,...) Linked to that node are compared to determine their similarity, and determined to be similar. By combining the Nth-order node with the starting node, it is possible to search using information having a semantic unit as one unit.
[0011]
However, according to such a method, information retrieval that captures the essence of the hyper document system can be performed, but an inclusion relationship may occur between “semantic groups”. There is a problem of becoming.
[0012]
An example of such a case is shown in FIG. In this figure, each parallelogram indicates a node, and each arrow indicates a link. In this example, the semantic group starting from node A (the part surrounded by a solid line) includes the semantic group starting from node B (the part surrounded by a dotted line). Therefore, when both are given as search results for a certain search request, the act of reading the latter is redundant for a document reader that has read (brows) the former first. However, there is also a problem in that it cannot be determined whether it is redundant unless it is actually accessed.
[0013]
The present invention has been made in view of these points, and an object of the present invention is to provide an information search apparatus that can efficiently search for information having a semantic unit as a unit.
[0014]
Furthermore, another object of the present invention is to provide an information retrieval method capable of efficiently retrieving information with information having a semantic unit as one unit.
Furthermore, another object of the present invention is to provide a recording medium on which an information retrieval program for efficiently retrieving information with a semantic unit as a unit is recorded.
[0015]
[Means for Solving the Problems]
In the present invention, in order to solve the above-mentioned problem, in the information retrieval apparatus for retrieving information for a hyper document system composed of nodes as a unit of information and links between the nodes, the combination of links and the similarity of contents Node group configuring means that configures a node group based on the configuration, configuration node storage means for storing individual nodes that configure the node group, and information for retrieving information from the hyperdocument system in response to a predetermined search request An inclusion relation analyzing means for analyzing a mutual inclusion relation among a plurality of node groups obtained as a search result of the information searching means based on information stored in the component node storage means; and the inclusion In accordance with the analysis result of the relationship analysis unit, a node group exclusion unit that excludes a node group included in another node group; Information retrieval apparatus characterized by having a search result output means for outputting the remaining nodes as the search results without being is provided.
[0016]
Here, the node group configuring means configures the node group based on the link connection and the similarity of contents. The configuration node storage means stores individual nodes constituting the node group. The information search means searches for information from the hyper document system in response to a predetermined search request. The inclusion relation analyzing means analyzes the mutual inclusion relation of the plurality of node groups obtained as a search result of the information search means based on the information stored in the constituent node storage means. The node group exclusion unit excludes a node group included in another node group according to the analysis result of the inclusion relationship analysis unit. The search result output means outputs the node group remaining without being excluded as a search result.
[0017]
In addition, in an information search device for searching information in a hyper document system configured by nodes and links between nodes, which are units of information, the nodes to be searched based on the combination of links and the similarity of contents A node group configuration unit that configures a group, a configuration node storage unit that stores individual nodes belonging to each of the node groups configured by the node group configuration unit, and a mutual inclusion relationship between the node groups, the configuration node Inclusion relation analysis means for analyzing based on information stored in the storage means; similarity calculation means for calculating similarity between a set of nodes whose inclusion relation is recognized by the inclusion relation analysis means; If the similarity calculated by the similarity calculation means is greater than or equal to a predetermined threshold, the included node group is excluded and excluded Information retrieval apparatus is provided, characterized in that it comprises a search object narrowing means to narrow the remaining nodes without as searched, a.
[0018]
Here, the node group configuration means configures a node group to be searched based on the link connection and the similarity of contents. The configuration node storage unit stores individual nodes belonging to each of the node groups configured by the node group configuration unit. The inclusion relation analyzing means analyzes the mutual inclusion relation of the node groups based on information stored in the constituent node storage means. The similarity calculation means calculates the similarity between a set of nodes whose inclusion relation is recognized by the inclusion relation analysis means. When the similarity calculated by the similarity calculation means is equal to or greater than a predetermined threshold, the search target narrowing means excludes the included node group and searches the remaining node group for the search. Refine as.
[0019]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a diagram showing a configuration example of an embodiment of an information search apparatus of the present invention. In this figure, the node group constructing means 1 is a node group (for example, a node starting from the node A or B in FIG. 10) according to the similarity of contents from information distributed in the hyper document system 10. Group).
[0020]
The configuration node storage unit 2 stores the path names of (different) nodes belonging to the node group configured by the node group configuration unit 1. The path name used here is an absolute address uniquely given to a node. For example, in the case of WWW (World Wide Web), it indicates a URL (Uniform Resource Locator).
[0021]
The information search means 3 acquires a node corresponding to the search request from the nodes arranged in the hyper document system 10 and outputs it as a search result candidate. The inclusion relation analyzing unit 4 inputs a node obtained as a result of the information search, acquires a node group starting from the node from the constituent node storage unit 2, and analyzes the mutual inclusion relation of the obtained node group To do.
[0022]
The node group exclusion unit 5 excludes a node group included in another node group according to the analysis result by the inclusion relationship analysis unit 4.
The search result output means 6 outputs the node group remaining without being excluded by the node group exclusion means 5 as the final search result.
[0023]
Next, a configuration example of the node group configuring unit 1 will be described with reference to FIG. As shown in FIG. 2, the node group constituting unit 1 includes a starting node feature extracting unit 1a, a related node acquiring unit 1b, a related node feature extracting unit 1c, a similarity determining unit 1d, and a node combining unit 1e.
[0024]
When the start node is input, the start node feature extraction unit 1a analyzes the content of the start node and extracts the feature. The extracted feature is passed to the similarity determination unit 1d as a starting node feature profile. Here, the information related to the characteristics of a node refers to a set of a pair of a word that characterizes the contents of the node and a value indicating its importance. For example, the origin node feature profile is created by weighting based on the appearance frequency, appearance position, and part-of-speech information regarding each word appearing at the origin node.
[0025]
When the origin node is input, the related node acquisition unit 1b acquires a secondary node linked from the node, and also selects a node (related node) further linked from the acquired node. Obtain sequentially. Then, the same processing is performed until there are no links to other nodes. Each node after the secondary node acquired at this time is set as the Nth order node (N = 2, 3, 4,...).
[0026]
The related node feature extraction unit 1c extracts the features of the Nth-order nodes extracted by the related node acquisition unit 1b, and creates a related node feature profile. The created related node feature profile is passed to the similarity determination unit 1d.
[0027]
Based on the related node feature profile created by the related node feature extraction unit 1c, the similarity determination unit 1d performs similarity determination processing for all the accessible nodes by following links from the start node. Then, the contents of the similar nodes are passed to the node composition unit 1e. The node synthesis unit 1e synthesizes all the nodes extracted by the similarity determination unit 1d with the origin node and outputs a synthesis node (node group). As a result, the node group constituting unit 1 outputs a plurality of path names corresponding to the composite node (node group) included by the same concept.
[0028]
Next, the operation of the embodiment shown in FIG. 1 will be described.
The node group constituting unit 1 generates and outputs a node group from information arranged in the hyper document system 10. That is, the node group constituting unit 1 scans the nodes arranged in the hyper document system 10 at random, and executes the following processing using the acquired node as a starting node.
[0029]
The related node acquisition unit 1b sequentially acquires N-th order nodes related to (linked to) the given starting node node. Then, the origin node feature extraction unit 1a and the related node feature extraction unit 1c perform feature extraction processing according to the processing shown in FIG. Since the processes performed by the origin node feature extraction unit 1a and the related node feature extraction unit 1c are the same, only the process performed by the origin node feature extraction unit 1a will be described in the following description.
[S1] A starting node is given, and the information source is input to the starting node feature extraction unit 1a.
[S2] The tag defined in the hyper document system description language (for example, HTML) is removed from the information source.
[S3] A word is extracted from the remaining text using a known morphological analysis technique.
[S4] Only important words are extracted from the set of words obtained in step S3. Here, the important word is a word characterizing the contents of the information source, and is extracted by, for example, a method in which only a noun is regarded as an important word.
[S5] The important word obtained in step S4 is weighted in consideration of the appearance frequency and the appearance position. That is, the higher the frequency of appearance, the higher the importance. Also, the importance is increased as the appearance position is closer to the head of the document.
[S6] Finally, a list composed of pairs of important words and their weights is created, and this is used as a starting node feature profile.
[0030]
The feature profile (single) of the starting node obtained in this way is passed to the similarity determination unit 1d.
As described above, the origin node is also passed to the related node acquisition unit 1b, and the related node acquisition unit 1b searches for link information included in the information source of the received node, and the link destination node. As a secondary node. For example, if the origin node is created in HTML, the URL (Uniform Resource Locator) in the area enclosed by the anchor tag (<A>... </A>) is extracted and specified by the URL. A document (secondary node) is acquired.
[0031]
Then, by repeating the same processing, Nth-order nodes (related nodes) are sequentially acquired and passed to the related node feature extraction unit 1c as a set of related nodes. The related node feature extraction unit 1c creates a related node feature profile for each related node by executing the same process as described above. The characteristic profiles (generally a plurality) of the related nodes are passed to the similarity determination unit 1d. Thereby, the origin node feature profile and a plurality of related node feature profiles are passed to the similarity determination unit 1d.
The similarity determination unit 1d executes the process shown in FIG.
[S21] An initialization of N = 2 is performed.
[S22] It is determined whether or not an Nth order node exists. If it exists, the process proceeds to step S23, and if not, the process ends.
[S23] The initialization p = 1 is performed. The number of Nth order nodes is m.
[S24] The magnitudes of p and m are compared. If p> m, the process proceeds to step S29. Otherwise, the process proceeds to step S25.
[S25] The similarity between the starting node and the p-th Nth-order node is calculated by the above-described method (known vector inner product calculation method).
[S26] The similarity value obtained in step S25 is compared with a threshold value. If similarity> threshold value, the process proceeds to step S27. Otherwise, the process proceeds to step S28. Here, the threshold value is a preset value, and the allowable range of similarity is adjusted according to the magnitude.
[S27] The pth Nth order node is stored as a candidate node to be combined with the starting node.
[S28] Add 1 to the value of p, and proceed to step S24.
[S29] 1 is added to the value of N, and the process proceeds to step S22.
[0032]
As a result, from all the nodes that can be accessed by following the link from the origin node, those having contents similar to the origin node are extracted.
Here, consider a case where the node 11 is input as a starting node (see FIG. 2). The node 11 is linked to two nodes 12 and 13. These nodes 12 and 13 become secondary nodes. Links are also established from the nodes 12 and 13 to other nodes, and finally links to the nodes 14 to 16 are established.
[0033]
When the node 11 is input to the node group constituting unit 1, the content is analyzed by the starting node feature extracting unit 1a, and the feature of the node 11 is passed to the similarity determining unit 1d as the starting node feature profile. The related node acquisition unit 1 b extracts the node path names of the nodes 12 and 13 linked from the node 11 and acquires the nodes 12 and 13.
[0034]
Further, all the N-th nodes 14 to 16 are obtained as accessible nodes by following the links from the nodes 12 and 13. The acquired node is passed to the related node feature extraction unit 1c. Then, the related node feature extraction unit 1c extracts the features of the contents of each node, and creates a related node feature profile. Then, all the related nodes having contents similar to the node 11 are extracted by the similarity determination unit 1d. Then, all the extracted nodes are combined with the starting node by the node combining unit 1e, and a combined node is generated.
[0035]
The above process is repeated starting from different nodes, and as a result, a plurality of composite nodes (node groups) composed of a plurality of nodes having similarity are generated and output.
[0036]
The composite node generated as described above is supplied to the configuration node storage unit 2.
The configuration node storage unit 2 stores a path name corresponding to each node constituting the composite node generated as described above.
[0037]
The information search means 3 searches the hyper document system 10 for a node corresponding to the input search request. As this search method, for example, as many nodes as possible are scanned from the hyper document system 10, a search index for each node is prepared, and an output corresponding to the input search request (a combination of keywords) is output. You can do that.
[0038]
The inclusion relation analysis unit 4 inputs a node that is a candidate for a search result supplied from the information search unit 3, and a synthesis node {G ₁ , G ₂ , G _Three , ..., G _N } Is searched from the constituent node storage means 2. Then, the inclusion relationship of the obtained composite nodes is analyzed according to the flowchart shown in FIG. The following processes are all executed by the inclusion relation analyzing unit 4.
[S41] A composite node including the node supplied from the information search means 3 is searched from the constituent node storage means 2, and a node group string {G ₁ , G ₂ , G _Three , ..., G _N }.
[S42] The variable i is initialized to the value “1”, and the contents of the excluded node group list are initialized to an empty (= NULL) state.
[S43] It is determined whether or not the i-th node group (composite node) is included in the excluded node group list. If it is included, the process proceeds to step S49. Proceed to S44.
[S44] A value (i + 1) is substituted into the variable k.
[S45] i-th synthesis node G _i Are all k nodes (= i + 1) th synthesis node G _k Is included in all the nodes constituting the node, and if included, the process proceeds to step S47, where it is not included, or G _i And G _k If they are equal, the process proceeds to step S46.
[S46] i-th synthesis node G _i Are all k nodes (= i + 1) th synthesis node G _k Whether or not they are equal to each other is determined. If they are included or equal, the process proceeds to step S48. If not, the process proceeds to step S50. move on.
[S47] G _i Is added to the excluded node group list.
[S48] G _k Is added to the excluded node group list.
[S49] The value of variable i is incremented by one.
[S50] The variable k is incremented by 1.
[S51] It is determined whether or not the value of the variable i is equal to the value of (N-1). If they are equal, the process proceeds to step S53, and if they are not equal, the process returns to step S43.
[S52] It is determined whether or not the value of the variable k is greater than the value of N. If it is determined that the value is larger, the process proceeds to step S49. If not, the process returns to step S45.
[S53] An excluded node group list is output.
[0039]
Through the above processing, a composite node included in another node is output as an object to be excluded.
The node group exclusion unit 5 refers to the excluded node group list output from the inclusion relationship analysis unit 4 and excludes unnecessary items from the search result synthesis node.
[0040]
The search result output means 6 outputs a node group from which unnecessary nodes are excluded as a search result.
According to the above-described embodiment, only synthesized nodes (node groups) that are not included in other nodes can be obtained as search results. For example, when the node A and the node B shown in FIG. 10 are obtained as a search result, only the node group starting from the node A is selected and output.
[0041]
In the above embodiment, the information retrieval unit 3 retrieves information from the hyper document system 10, but information necessary for retrieval together with the node name with respect to the constituent node storage unit 2 (for example, each It is also possible to search from the constituent node storage means 2 by storing the node search index).
[0042]
Next, a second embodiment of the information retrieval apparatus of the present invention will be described with reference to FIG.
The second embodiment shown in FIG. 6 excludes the node group included in the search request from the search result when the search request is a node group configured based on the combination of the link and the similarity of the contents. 1 shows an embodiment in which a function is added.
[0043]
In the figure, parts corresponding to those in FIG. 1 are denoted by the same reference numerals, and description thereof is omitted.
The search request constituent node storage unit 21 stores and holds the path names of the nodes constituting the search request. The search request inclusion relation analyzing means 22 excludes search result candidates (node groups) included in the search request node group. That is, the pair search request inclusion relation analyzing means 22 refers to the information held in the search request constituent node storage means 21, the information held in the constituent node storage means 2, and the search result of the information search means 3. Thus, the mutual inclusion relationship between the search request and the search result candidate is analyzed. The pair search request node group exclusion unit 23 refers to the analysis result of the pair search request inclusion relation analysis unit 22 and excludes the node group included in the search request from the search result candidates.
[0044]
The search result output means 25 outputs a node group that has not been excluded by either the pair search request node group exclusion means 23 or the node group exclusion means 5 as a final search result.
[0045]
Next, the operation of the above embodiment will be described. The operations of the node group configuration unit 1, the configuration node storage unit 2, the inclusion relationship analysis unit 4, and the node group exclusion unit 5 are the same as those in FIG.
[0046]
When the document reader inputs a search request, the search request constituent node storage unit 21 stores the path names of the individual nodes constituting the search request. The search request inclusion relation analyzing means 22 extracts, by analysis, those included in the search request from the search result candidates according to the processing shown in FIG.
[S71] A node group sequence {G obtained by acquiring from the constituent node storage unit 2 a composite node starting from the node obtained by the information search unit 3 ₁ , G ₂ , G _Three , ..., G _N } And the node group G of the search request _Q Enter.
[S72] The variable i is initialized to the value “1”, and the contents of the excluded node group list are made empty (NULL).
[S73] i-th node group G _i All (different) nodes that make up G _Q Is included in the set of nodes constituting the node, and if included, the process proceeds to step S74, and if not included, the process proceeds to step S75.
[S74] Node group sequence G _i To the excluded node group list.
[S75] The value of the variable i is incremented by “1”.
[S76] It is determined whether or not the value of the variable i is equal to N. If equal, the process proceeds to step S77. If not equal, the process returns to step S73.
[S77] An excluded node group list is output.
[0047]
According to the above processing, the node group sequence {G ₁ , G ₂ , G _Three , ..., G _N }, Search request G _Q Are included and output as an excluded node group list.
[0048]
The search request node group exclusion means 23 excludes those included in the excluded node group list from the search result candidates and outputs the result.
The search result output means 25 includes a node group output from the pair search request node group exclusion means 23 (a node group not included in the search request node group) and a node group output from the node group exclusion means 5 (others). Node groups not included in the node group) and the node groups included in both are output as search results.
[0049]
According to the embodiment as described above, when a search request including a plurality of node groups is input, among the obtained search result candidates, a node group included in another node group, and The node group included in the search request is excluded, and the final search result is output. Accordingly, it is possible to prevent the input node group itself from being output as a search result, and to prevent output of mutually overlapping information as a search result, so that efficient information search is performed. It becomes possible.
[0050]
In the above embodiment, the exclusion of the node group included in the search request and the exclusion of the node group included in the other node group are performed in parallel, but these are sequentially executed. You may do it. For example, after the node group included in the search request is excluded, the node group included in another node group is excluded.
[0051]
Next, a configuration example of the third embodiment of the present invention will be described with reference to FIG. In this embodiment, before the search request is input from the document reader, the inclusion relation of the node group to be searched is analyzed in advance, and the node group included in the other node group is excluded from the search target. By doing so, the operation time (runtime) after inputting the search request is reduced.
[0052]
In this figure, the node group constituting unit 1 generates a node group as a unit of search from the nodes arranged in the hyper document system 10 based on the similarity of the contents and the link.
[0053]
The configuration node storage unit 2 stores path names corresponding to individual nodes included in the node group configured by the node group configuration unit 1.
The inclusion relationship analysis unit 3 analyzes the mutual inclusion relationship of the node groups stored in the configuration node storage unit 2.
[0054]
The similarity calculation means 41 calculates the similarity of the contents of the two sets of nodes in the inclusion relationship and outputs the result to the inclusion relationship analysis means 3.
The search target narrowing means 42 is included in another node group and excludes and outputs a node group whose similarity exceeds a threshold value.
[0055]
Next, the operation of the above embodiment will be described.
The node group configured by the node group configuration unit 1 is stored in the configuration node storage unit 2. The inclusion relation analysis unit 3 and the similarity calculation unit 41 cooperate to execute processing described later with reference to FIG. 9 to exclude nodes that are included in other node groups and have high mutual similarity. An excluded node group list is generated for this purpose.
[0056]
In other words, if only the inclusion relation is analyzed, even if it is a node group that should originally be returned as a search result, a search is performed if it is included in another (not returned as a search result) node group. As a result, the inconvenience of not being output occurs. Therefore, the similarity calculation unit 41 calculates the similarity of contents between nodes in an inclusive relationship, and only determines that the contents are determined to be similar when excluded.
[0057]
When the flowchart shown in FIG. 9 is started, the following processing is executed.
[S91] The inclusion relation analyzing unit 3 stores the node group sequence {G ₁ , G ₂ , G _Three , ..., G _N }.
[S92] The variable i is initialized to the value “1”, and the contents of the excluded node group list are initialized to an empty (= NULL) state.
[S93] i-th node group (composite node) G _i Is included in the excluded node group list (G _i Is included in the excluded node group list), and if included, the process proceeds to step S101, and if not included, the process proceeds to step S94.
[S94] A value (i + 1) is substituted into the variable k.
[S95] i-th synthesis node G _i Are all k nodes (= i + 1) th synthesis node G _k Is included in all the nodes constituting the, and if included, the process proceeds to step S97, where it is not included, or G _i And G _k If they are equal, the process proceeds to step S96.
[S96] i-th synthesis node G _i Are all k nodes (= i + 1) th synthesis node G _k It is determined whether or not all nodes constituting the node are included or are equal to each other, and if included or equal, the process proceeds to step S98, and if not included, the process proceeds to step S102. move on.
[S97] The similarity calculation means 41 uses G _i And G _k The similarity is calculated. As a method of calculating the similarity here, first, G _i And G _k All the nodes constituting the are respectively synthesized. Next, the processing shown in FIG. 3 is executed to create a profile representing the characteristics of the contents of each composite node and calculate their similarity. The inclusion relationship analyzing unit 3 determines that the similarity is similar if the similarity is greater than a predetermined threshold, and proceeds to step S99. If the similarity is small, the inclusion relationship analyzing unit 3 determines that the similarity is not similar and proceeds to step S101. .
[S98] By the same processing as step S97, G _k And G _i If it is determined that they are similar, the process proceeds to step S100. If it is determined that they are not similar, the process proceeds to step 102.
[S99] G _i Is added to the excluded node group list.
[S100] G _k Is added to the excluded node group list.
[S101] The value of variable i is incremented by one.
[S102] The value of the variable k is incremented by 1.
[S103] It is determined whether or not the value of the variable i is equal to (N-1). If they are equal, the process proceeds to step S105, and if they are not equal, the process returns to step S93.
[S104] It is determined whether or not the value of the variable k is larger than the value of N. If it is determined that the value is larger, the process proceeds to step S101. If not, the process returns to step S95.
[S105] An excluded node group list is output.
[0058]
Through the above processing, among the node groups to be searched, nodes that are included in other nodes and whose similarity exceeds a threshold value are output as an excluded node group list.
[0059]
In this way, node groups that are included in other node groups and whose similarity exceeds a predetermined threshold are excluded by the search target narrowing means 42. A node group that is not excluded becomes a new search target.
[0060]
According to the above embodiment, the document reader can perform an efficient search by targeting the node group that has not been excluded.
The above processing functions can be realized by a computer. In this case, the processing contents of the functions that the information retrieval apparatus should have are described in a program recorded on a computer-readable recording medium, and the above processing is realized by the computer by executing the program by the computer. Is done.
[0061]
Examples of the computer-readable recording medium include a magnetic recording device and a semiconductor memory. When distributing to the market, store the program in a portable recording medium such as a CD-ROM (Compact Disk Read Only Memory) or floppy disk, or store it in a computer storage device connected via a network. In addition, it can be transferred to another computer through the network. When executed by a computer, the program is stored in a hard disk device or the like in the computer, loaded into the main memory and executed.
[0062]
【The invention's effect】
As described above, in the information search apparatus according to the present invention, a node group is configured according to the link and the similarity of contents, and when there is a predetermined search request, the corresponding node group is acquired, Since the nodes included in the other node groups are excluded from the acquired node group and output, it is possible to avoid browsing duplicate information.
[0063]
Further, in the information search apparatus according to the present invention, a node group is configured according to the combination of links and the similarity of contents, and a node group having an inclusive relationship is extracted from the configured node group, and each similarity is determined. As a result of calculating and eliminating the included node group for the node group whose similarity is larger than the predetermined threshold, it is possible to shorten the runtime when searching for information. .
[Brief description of the drawings]
FIG. 1 is a configuration example of an embodiment of the present invention.
FIG. 2 is a diagram illustrating a configuration example of a node group configuration unit illustrated in FIG. 1;
FIG. 3 is a flowchart for explaining an example of processing executed in an origin node feature extraction unit and a related node feature extraction unit shown in FIG. 2;
4 is a flowchart illustrating an example of processing executed in a similarity determination unit shown in FIG.
FIG. 5 is a flowchart for explaining an example of processing executed by the inclusion relationship analysis unit shown in FIG. 1;
FIG. 6 is a diagram illustrating a configuration example of a second embodiment of the present invention.
7 is a diagram for explaining an example of processing executed by the pair search request inclusion relation analyzing unit shown in FIG. 6;
FIG. 8 is a diagram illustrating a configuration example of a third embodiment of the present invention.
FIG. 9 is a flowchart for explaining an example of processing executed by the inclusion relationship analyzing unit and the similarity calculating unit shown in FIG. 8;
FIG. 10 is a diagram illustrating a group of nodes that are in an inclusive relationship.
[Explanation of symbols]
1 Node group configuration means
2. Configuration node storage means
3 Information retrieval means
4 Inclusion relationship analysis means
5 Node group exclusion means
6 Search result output means
21 Search request component node storage means
22 Pair search request inclusion relation analysis means
23 Pair search request node group exclusion means
25 Search result output means
41 Similarity calculation means
42 Search target narrowing means

Claims (7)

In an information retrieval apparatus for retrieving information for a hyper document system configured by nodes and links between nodes as a unit of information,
A node group configuration means for configuring a node group based on the link and the similarity of contents;
Configuration node storage means for storing individual nodes constituting the node group;
Information retrieval means for retrieving information from the hyperdocument system in response to a predetermined retrieval request;
Inclusion relation analysis means for analyzing the mutual inclusion relation of a plurality of node groups obtained as a search result of the information search means based on information stored in the component node storage means;
In accordance with the analysis result of the inclusion relation analysis unit, a node group exclusion unit that excludes a node group included in another node group;
Search result output means for outputting a node group remaining without being excluded as a search result;
An information retrieval apparatus comprising: Search request configuration node storage means for storing individual nodes constituting the node group when a group of nodes configured based on the link and content similarity is given as a search request;
Pair search request inclusion relation analyzing means for analyzing an inclusion relation between a node group obtained as a search result candidate and the node group of the search request with reference to information stored in the search request constituent node storage means When,
2. The information search apparatus according to claim 1, further comprising pair search request node group exclusion means for excluding a node group included in the search request node group. In an information search method for searching information for a hyper document system configured by nodes and links between nodes, which are units of information,
A node group configuration step for configuring a node group based on the link and the similarity of contents;
A configuration node storage step of storing individual nodes constituting the node group;
An information search step of searching for information from the hyper document system in response to a predetermined search request;
An inclusion relationship analysis step of analyzing the mutual inclusion relationship of the plurality of node groups obtained as a search result of the information search step based on information stored in the component node storage step;
In accordance with the analysis result of the inclusion relationship analysis step, a node group exclusion step of eliminating a node group included in another node group;
A search result output step for outputting a node group remaining without being excluded as a search result;
A method for retrieving information, comprising: In a computer-readable recording medium in which an information search program for causing a computer to search for information of a hyper document system configured by nodes and links between nodes, which are units of information, is recorded.
Node group configuration means for configuring a node group based on the link and the similarity of contents,
Configuration node storage means for storing individual nodes constituting the node group,
Information retrieval means for retrieving information from the hyperdocument system in response to a predetermined retrieval request;
Inclusion relation analysis means for analyzing mutual inclusion relations of a plurality of node groups obtained as a search result of the information search means based on information stored in the component node storage means;
Node group exclusion means for eliminating a node group included in another node group according to the analysis result of the inclusion relation analysis means;
A search result output means for outputting the node group remaining without being excluded as a search result;
As a computer-readable recording medium on which a program for causing a computer to function is recorded. In an information retrieval apparatus for retrieving information for a hyper document system configured by nodes and links between nodes as a unit of information,
A node group configuring means for configuring a node group to be searched based on the link and the similarity of contents;
Configuration node storage means for storing individual nodes belonging to each of the node groups configured by the node group configuration means;
Inclusion relation analysis means for analyzing the mutual inclusion relation of the node group based on information stored in the component node storage means;
Similarity calculating means for calculating the similarity between a set of nodes whose inclusion relation is recognized by the inclusion relation analyzing means;
When the similarity calculated by the similarity calculation means is equal to or greater than a predetermined threshold, the included node group is excluded, and the remaining node group that is not excluded is narrowed down as a search target. Means,
An information retrieval apparatus comprising: In an information search method for searching information for a hyper document system configured by nodes and links between nodes, which are units of information,
A node group configuration step that configures a node group to be searched based on the combination of links and the similarity of contents;
A configuration node storage step of storing individual nodes belonging to each of the node groups configured by the node group configuration step;
An inclusion relationship analysis step of analyzing mutual inclusion relationships of the node groups based on information stored in the configuration node storage step;
A similarity calculation step for calculating a similarity between a set of nodes whose inclusion relationship is recognized by the inclusion relationship analysis step;
When the similarity calculated by the similarity calculation step is equal to or greater than a predetermined threshold, the included node group is excluded, and the remaining node group that is not excluded is narrowed down as a search target. Steps,
A method for retrieving information, comprising: In a computer-readable recording medium in which an information search program for causing a computer to search information of a hyper document system configured by nodes and links between nodes, which are units of information, is recorded.
A node group configuration means for configuring a node group to be searched based on the combination of links and the similarity of contents;
Configuration node storage means for storing individual nodes belonging to each of the node groups configured by the node group configuration means,
Inclusion relation analyzing means for analyzing mutual inclusion relation of the node group based on information stored in the component node storage means;
Similarity calculation means for calculating the similarity between a set of nodes whose inclusion relation is recognized by the inclusion relation analysis means;
When the similarity calculated by the similarity calculation means is equal to or greater than a predetermined threshold, the included node group is excluded, and the remaining node group that is not excluded is narrowed down as a search target. means,
A computer-readable recording medium on which a search program for causing a computer to function is recorded.

Images