JP2000172707A - Device and method for analyzing hypertext and storage medium with hypertext analysis program recorded therein - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a hypertext analyzing device which supports obtaining knowledge for deciding the relative merits of the configuration of a hypertext system. SOLUTION: A content distribution trend analyzing part 2 performs clustering processing of a node group in hypertext 1 by using the details of the contents of respective nodes. A hyperlink configuration analyzing part 3 calculates the hyperlink binding degree between nodes constituting each cluster about the cluster generated by the part 2. Because a node group constituting each cluster is a set of similar contents, when the hyperlink bonding degree is high, it is indicated that many paths for shifting to the similar content are offered. It becomes possible to decide the relative merits of the configuration of a hypertext system by such a manner that the hyperlink bonding degree is shown by a hyperlink bonding degree displaying part 4.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a hypertext analysis apparatus and method for analyzing a content distribution and a hyperlink structure in a hypertext system configured on a network, and to realize the hypertext analysis apparatus and method by a computer. The present invention relates to a storage medium storing a hypertext analysis program for performing the above.

[0002]

2. Description of the Related Art In recent years, for example, World Wide
Like the Web (hereinafter abbreviated as Web), a large-scale hypertext system has been configured on a network. In addition, a search service for searching for those hypertexts (for example, Excite or AltaV)
ista) are also provided.

[0003] A technique of performing content-based clustering on a general document set has been established, and is mainly used for browsing search results. For example,
P. Willett, “RECENT TRENDS
IN HIERARCHICDOCUMENT CLU
STARING: A CRITICAL REVIE
W ", Information Processing
and Management, Vol. 24, N
o. 5,1988. And so on.

A technique for efficiently processing a large number of documents by using short snippets displayed in a search result list returned from a search engine as input and performing clustering while focusing on words shared between documents. There is. For example, O. Zamir a
nd O.D. Etzioni, “Web Documen
t Clustering ", 1998 ACM SI
GMOD Works on Research
Issues in Data Mining an
d Knowledge Discovery, 199
8. And so on. However, while this technique may be effective for browsing search results, it is not for improving the configuration of a hypertext system.

As another clustering technique, for example, M.I. Perkowitz and O.M. Etzio
ni, "Adaptive Web Sites: A
automatically synthesizing
Web Pages ", In Proc. Of A
AAI, 1998. There is a technique proposed in. This technology performs clustering on a group of nodes in a Web site based on co-occurrence of access, generates an index for a node not connected by a hyperlink, and generates a W
This is to improve the configuration of the website.

However, in this technique, information relating to the contents of a document is not handled, so that history information of browsing without purpose or browsing by a person having a plurality of purposes becomes noise, and accuracy is reduced in terms of improving a site configuration. Is not always obtained. Also, the browsing activity itself is a trial and error activity, and not all the accessed nodes are significant. In particular, this tendency becomes remarkable in a site determined to be inferior in the present invention.

As described above, in the prior art, it was not possible to obtain knowledge for judging the superiority or inferiority of the content distribution on the hypertext system.

[0008]

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned circumstances, and provides a hypertext analysis device and a hypertext for assisting in obtaining knowledge for judging the configuration of a hypertext system. It is intended to provide an analysis method. Also,
It is an object of the present invention to provide a recording medium recording a hypertext analysis program for realizing the hypertext analysis device or method by a computer.

[0009]

According to the present invention, clustering is performed on nodes constituting a hypertext system based on the similarity of contents, and for each cluster obtained by clustering, a node constituting the cluster is obtained. It is characterized by calculating and displaying the degree of unity of hyperlinks between them. The displayed hyperlink cohesion is a value based on the similarity of the content, and thus indicates the distribution of the content. for that reason,
By obtaining the degree of hyperlink cohesion, it is possible to determine, for example, the hyperlink configuration of a hypertext system (for example, a Web site) and the superiority or inferiority of the configuration of the hypertext system.

Further, according to the present invention, the nodes constituting the hypertext system are clustered on the basis of the similarity of the contents, and for each cluster obtained by the clustering, a node constituting the cluster and a specific node The number of hyperlink transitions (the shortest number of transitions) between (originating nodes) is calculated, and each cluster is displayed based on the number of hyperlink transitions. Thus, based on the number of hyperlink transitions that must be followed from a specific node, it is possible to know the variation of similar contents, and to determine whether or not the hypertext system has the intended configuration.

[0011]

FIG. 1 is a configuration diagram showing a first embodiment of the present invention. In the figure, 1 is a hypertext system, 2 is a content distribution tendency analysis unit, 3 is a hyperlink configuration analysis unit, and 4 is a hyperlink cohesion degree display unit. The hypertext system 1 stores a plurality of hypertexts (hereinafter, referred to as nodes) in a hyperstructure of nodes and hyperlinks. The content can be obtained by accessing the hypertext system 1. A typical example of the hypertext system 1 is Web.

The content distribution tendency analysis unit 2 performs a clustering process on the nodes in the hypertext 1 using the contents of the content of each node, for example, text information. This clustering process includes:
Existing technology can be used. For example, Aggl
omerative Hierarchical Cl
In the clustering method by using the following, the following steps 1, 2, and 3 are performed. 1. Each node is a cluster. 2. The similarity between each cluster is calculated, and the clusters having the maximum similarity are merged into one cluster. Here, if the maximum similarity is smaller than a preset threshold value, the process ends. 3. Step 2 is repeated until the number of clusters becomes one. Each cluster sequentially generated in the middle of this process is obtained as a processing result of clustering. For example, 1. The clustering of only the nodes generated in the processing of the above, each cluster obtained by merging the clusters, and the like are the clustering results. The above-described clustering method is described in, for example, E. M.
Voorhees, “Implementing Ag
glomerative Hierarchical C
lustering Algorithms for
Use in Document Retrieva
l ", Information Processing
& Management, Vol. 22, no. 6,1
986 and the like.

Here, 2. In the similarity calculation of, the degree of duplication of appearing words in each cluster can be used. Alternatively, for example, using a method described in Japanese Patent Application No. 9-153387, a vector in which each cluster is defined as an important word appearing in the cluster and the appearance frequency of the important word is defined as the value of the term May be generated, and the magnitude of the inner product value between the vectors may be used as the similarity.

[0014] The hyperlink configuration analysis unit 3 performs, for each cluster generated by the content distribution tendency analysis unit 2,
The degree of hyperlink cohesion between the nodes constituting the cluster (hereinafter, simply referred to as cohesion degree) is calculated. The cohesion is
For example, if there is one or more hyperlinks between nodes, it is defined that there is a connection between the nodes. When it is defined that there is a connection between the nodes, the total number of connections between nodes forming the cluster is calculated by subtracting the total number of connections from all nodes forming the cluster. The value can be divided by the number of combinations for selecting one. That is, assuming that the total number of node connections is L and the number of nodes is N, the cohesion degree = L / _N C
₂ can be calculated. The value of the cohesion degree of each cluster obtained by the hyperlink configuration analysis unit 3 is passed to the hyperlink cohesion degree display unit 4.

The hyperlink cohesion degree display section 4 displays the cohesion degree obtained by the hyperlink configuration analysis section 3. For example, the relationship between the cluster size and the value of the cohesion degree can be displayed. As the size of the cluster, for example, the number of nodes constituting the cluster, the total number of words of each node constituting the cluster, the total file size of each node constituting the cluster, and the like can be used.

FIG. 2 is an explanatory diagram of a display example in the hyperlink cohesion degree display section in the first embodiment of the present invention. The display example shown in FIG. 2 shows an example in which a hyperlink cohesion degree of a certain hypertext system (here, a certain Web site) is displayed. Here, the number of nodes constituting the cluster is used as the size of the cluster, and the value obtained by dividing the total number of connections between the nodes constituting the cluster by the number of combinations of selecting two from all the nodes constituting the cluster is used as the cohesion degree. Used.

With the display as shown in FIG. 2, it is possible to look down on the superiority of the configuration of each cluster while paying attention to the cluster size, and to determine the superiority of the configuration of the entire hypertext system. As described above, since each cluster is configured based on the text information included in the node, it can be said that the node group configuring each cluster is a set of similar contents. If the cohesion of the node group is high, the user is provided with a large number of paths for transition between nodes, so that similar content can be efficiently browsed. On the other hand, if the cohesion degree of the node group is low, it indicates that similar contents are scattered, and the browsing efficiency is deteriorated. In FIG. 2, if the number of nodes is the same, the more the clusters with a higher cohesion degree, the better the configuration of the entire hypertext system 1 can be determined. Conversely, if there are many clusters with low cohesion, it can be determined that there is room for improving the configuration.

FIG. 3 is an explanatory diagram of another display example in the hyperlink cohesion degree display section in the first embodiment of the present invention. In the hyperlink cohesion degree display section 4, FIG.
In addition to the display example shown in FIG. 3, for example, as shown in FIG. 3, the relationship between the similarity between the nodes in the cluster and the value of the cohesion degree can also be displayed. In the display example shown in FIG. 3, for a certain hypertext system (here, a certain Web site), the similarity between nodes in the cluster is used as the similarity between the nodes in the content distribution trend analysis unit 2 when the cluster is generated. . Further, an example is shown in which a value obtained by dividing the total number of connections between nodes constituting a cluster by the number of combinations of selecting two from all the nodes constituting the cluster is used as the cohesion degree.

In FIG. 3, the cluster group A has a tendency (intuitively natural) that the cohesion increases as the similarity between nodes in the cluster increases, and general contents are included in the cluster. It is speculated that
Here, focusing on a certain point (nearby) on the horizontal axis (= similarity between nodes in the cluster), a cluster having a high cohesion degree has an excellent configuration, and a cluster having a low cohesion degree is inferior. It is presumed. On the other hand, the cluster group B is characterized in that the cohesion is high despite the low similarity between the nodes in the cluster. These include nodes for navigating the user (for example, index nodes), and are presumed to be tightly linked by hyperlinks to nodes containing other general content. In this way, the display as shown in FIG. 3 allows a bird's-eye view of the configuration of each cluster while paying attention to the similarity between the nodes in the cluster.
It is also possible to determine the superiority of the configuration of the entire website.

FIG. 4 is a block diagram showing a modification of the first embodiment of the present invention. In the figure, the same parts as those in FIG. Reference numeral 5 denotes an analysis area designation unit. In this modification, an example is shown in which an analysis area designating unit 5 is added to the configuration shown in FIG.

The analysis area designation unit 5 selects only nodes on a specific server (computer) from the hypertext system 1 as objects to be analyzed, and passes the selected nodes to the content distribution tendency analysis unit 2. Alternatively, the analysis area designation unit 5 may transfer the hyperlink from a specific node (origin node) and reachable to the content distribution tendency analysis unit 2 as an analysis target. At this time, the maximum number of transitions may be set in advance, and a node that can be reached within a range where the number of transitions does not exceed the maximum number of transitions may be an analysis target.

The content distribution tendency analysis unit 2 performs a clustering process on the node selected as an analysis target by the analysis area designating unit 5. Subsequent processing is as described above.

FIG. 5 is a block diagram showing another modification of the first embodiment of the present invention. In the figure, the same parts as those in FIG. Reference numeral 6 denotes a hyperlink cohesion degree evaluation unit. This modification shows an example in which a hyperlink cohesion degree evaluation unit 6 is added to the configuration shown in FIG.

The hyperlink cohesion evaluation section 6 compares the value of the cohesion degree of each cluster obtained by the hyperlink configuration analysis section 3 with a predetermined threshold value, and determines a cluster having a cohesion degree smaller than the threshold value as a hyperlink. It is passed to the cohesion degree display section 4. Thus, the hyperlink cohesion degree display unit 4 can present a cluster having a small cohesion degree, that is, a cluster having a poor configuration.

It should be noted that the configuration shown in FIG. 1 may be added with both the analysis area designation unit 5 and the hyperlink cohesion degree evaluation unit 6 shown in the above two modifications.

FIG. 6 is a block diagram showing a second embodiment of the present invention. In the figure, reference numeral 11 denotes a content distribution tendency analysis unit, 12 denotes a hyperlink path analysis unit, and 13 denotes a cluster display unit. The hypertext system 1 is the same as in the first embodiment.

The content distribution tendency analysis unit 11 performs a clustering process on the nodes in the hypertext system 1 using the text information of each node. As the clustering method, the same method as in the first embodiment can be used.

For each cluster obtained by the content distribution tendency analysis unit 11, the hyperlink path analysis unit 12 calculates the number of hyperlink transitions between nodes constituting the cluster and a given node (starting node) (for example, Minimum value). This can be determined by searching for a hyperlink from the origin node with breadth-first. Here, in order to improve processing efficiency, and
In order to deal with a node that is not connected to the starting point, a maximum value Nmax of the number of hyperlink transitions may be given in advance, and the search range may be limited so as not to search for a hyperlink exceeding the maximum value. In this case, the maximum value Nmax may be given to the nodes that have not been searched as the number of hyperlink transitions.

The cluster display unit 13 displays each cluster obtained by the content distribution tendency analysis unit 11 based on the number of hyperlink transitions obtained by the hyperlink path analysis unit 12. For example, each cluster obtained by the content distribution tendency analysis unit 11 can be displayed on a diagram centered on the number of hyperlink transitions according to the number of hyperlink transitions obtained by the hyperlink path analysis unit 12. .

FIG. 7 is an explanatory diagram of a display example on the cluster display unit in the second embodiment of the present invention. FIG.
In the figure, the radial direction indicates the number of hyperlink transitions, and nodes having the same number of hyperlink transitions are arranged on concentric circles. Here, for two clusters, cluster A and cluster B, nodes included in each cluster are shown as dots. Since the clustering is performed based on the similarity of the content as described above, the nodes included in each cluster have similar content.

The nodes constituting the cluster A are displayed substantially concentrically in FIG. 7, and the number of hyperlink transitions from the originating node is distributed in a narrow range of 2 or 3. On the other hand, the nodes constituting the cluster B are as shown in FIG.
, The number of hyperlink transitions from the origin node is distributed over a wide range of 1 to 5. The administrator of the hypertext system 1 (for example, a website administrator) compares such a display with an initial policy in which the nodes configuring these clusters are arranged, and determines the hypertext system 1 (web site). The superiority of the configuration can be determined. For example, if the initial policy is to dig deeper into a topic each time the hyperlink is followed, cluster B shown in FIG. 7 meets the initial policy, and cluster A has the following policy: It turns out that it is not suitable. Alternatively, if the initial policy is to express a topic within a range close to the origin node (a range in which the number of hyperlink transitions is small), the cluster A shown in FIG. 7 meets the initial policy. It can be seen that cluster B does not meet the policy.

FIG. 8 is a block diagram showing a modification of the second embodiment of the present invention. In the figure, the same parts as those in FIG. 6 are denoted by the same reference numerals, and description thereof will be omitted. Reference numeral 14 denotes an analysis area designating unit. In this modification, the configuration shown in FIG.
This shows an example in which an analysis area designation unit 14 is added.

The analysis area designation unit 14 selects only nodes on a specific server (computer) from the hypertext system 1 as objects to be analyzed, and transfers the selected nodes to the content distribution tendency analysis unit 11. Alternatively, analysis area designating section 1
4 may be such that nodes reachable by transiting hyperlinks from a specific node (origin node) (with a priority on width) may be passed to the content distribution tendency analysis unit 11 as an analysis target. In this case, the maximum number of transitions may be set in advance, and an object that can be reached within a range in which the number of transitions does not exceed the maximum number of transitions may be used as the analysis target.

FIG. 9 is a configuration diagram showing another modification of the second embodiment of the present invention. Reference numerals in the figure are the same as those in FIG. In this example, when the analysis region specifying unit 14 transits a hyperlink and specifies a node to be analyzed, the hyperlink path analysis unit 12
An example is shown in which the number of hyperlink transitions from the origin node is determined. In this configuration, the analysis area designation unit 1
Since the number of hyperlink transitions is calculated by the hyperlink path analysis unit 12 in the process of step 4, it is not necessary to search for a hyperlink again to obtain the number of hyperlink transitions after the processing by the content distribution tendency analysis unit 11. Therefore, processing efficiency can be improved.

Each of the above-described embodiments and their modifications can also be realized by a computer program. In that case, the program, data used by the program, and the like can be recorded on a computer-readable storage medium. A storage medium is a type of signal corresponding to a change state of energy such as magnetism, light, electricity, etc., caused to a reading device provided in a hardware resource of a computer in accordance with a description content of a program. Thus, the program description can be transmitted to the reading device. For example, a magnetic disk, an optical disk, a CD-ROM, a memory built in a computer, and the like.

[0036]

As is clear from the above description, according to the present invention, clustering is performed on nodes constituting the hypertext system based on the similarity of contents, and each obtained cluster is formed. Calculate and display hyperlink cohesion between nodes. Alternatively, based on the similarity of the contents, clustering is performed on the nodes constituting the hypertext system, and the obtained number of transitions of the hyperlink between the nodes constituting each cluster is calculated and displayed. The degree of hyperlink cohesion and the number of hyperlink transitions can be an index indicating the superiority or inferiority of the configuration of the hypertext system. By referring to the display based on these values, it is possible to have a bird's-eye view of the superiority of the configuration of the hypertext system, and to make a change to a portion having a problem in the configuration to construct a hypertext system having a better configuration. There is an effect that can be.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing a first embodiment of the present invention.

FIG. 2 is an explanatory diagram of a display example in a hyperlink cohesion degree display unit in the first embodiment of the present invention.

FIG. 3 is an explanatory diagram of another display example in the hyperlink cohesion degree display unit according to the first embodiment of the present invention.

FIG. 4 is a configuration diagram showing a modification of the first embodiment of the present invention.

FIG. 5 is a configuration diagram showing another modified example of the first embodiment of the present invention.

FIG. 6 is a configuration diagram showing a second embodiment of the present invention.

FIG. 7 is an explanatory diagram of a display example on a cluster display unit according to the second embodiment of the present invention.

FIG. 8 is a configuration diagram showing a modification of the second embodiment of the present invention.

FIG. 9 is a configuration diagram showing another modification of the second embodiment of the present invention.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Hypertext system, 2 ... Content distribution tendency analysis part, 3 ... Hyperlink structure analysis part, 4 ... Hyperlink cohesion degree display part, 5 ... Analysis area designation part, 6 ... Hyperlink cohesion degree evaluation part, 11 ... Content Distribution trend analysis unit, 12: hyperlink route analysis unit, 13: cluster display unit, 14: analysis area designation unit.

Claims (5)

[Claims] 1. Content distribution tendency analyzing means for performing clustering on nodes constituting a hypertext system based on similarity of contents, and forming the cluster for each cluster obtained by the content distribution tendency analyzing means. Hypertext analysis, comprising: hyperlink configuration analysis means for calculating a hyperlink cohesion degree between nodes; and hyperlink cohesion degree display means for displaying the hyperlink cohesion degree obtained by the hyperlink configuration analysis means. apparatus. 2. The apparatus according to claim 1, further comprising: a hyperlink cohesion degree evaluation unit configured to determine whether the hyperlink cohesion degree satisfies a predetermined condition. 2. The hypertext analysis apparatus according to claim 1, wherein a hyperlink cohesion degree is displayed according to a determination result of the means. 3. A content distribution tendency analyzing means for performing clustering on nodes constituting a hypertext system based on the similarity of contents, and the cluster is formed for each cluster obtained by the content distribution tendency analyzing means. Hyperlink path analysis means for calculating the number of hyperlink transitions between a node and a specific node, and cluster display means for displaying each cluster obtained by the content distribution trend analysis means based on the number of hyperlink transitions A hypertext analysis device, characterized in that: 4. Clustering is performed on the nodes constituting the hypertext system based on the similarity of the contents, and for each of the obtained clusters, the hyperlink cohesion between the nodes constituting the cluster is calculated and calculated. A hypertext analysis method characterized by displaying a hyperlink cohesion degree. 5. A content distribution trend analysis process for performing clustering on nodes constituting a hypertext system based on content similarity, and forming a cluster for each cluster obtained by the content distribution trend analysis process. Hyperlink configuration analysis processing for calculating the degree of hyperlink cohesion between nodes, and hyperlink cohesion indicating the degree of the configuration of the hypertext system to the user by displaying the hyperlink cohesion degree obtained in the hyperlink configuration analysis processing Storage medium storing a hypertext analysis program for causing a computer to execute a degree display process.