JP3705331B2 - Hypertext analysis apparatus and method, and storage medium storing hypertext analysis program - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a hypertext analysis apparatus and a hypertext analysis method for analyzing a hyperlink structure in order to discover knowledge for determining the superiority or inferiority of the configuration in a hypertext system configured on a network, and the hypertext The present invention relates to a recording medium on which a hypertext analysis program for realizing an analysis apparatus or method by a computer is recorded.
[0002]
[Prior art]
In a hypertext system (for example, World Wide Web: hereinafter abbreviated as “Web”) configured on a network, a user (visitor) access history can be recorded in a server storing hypertext. This access history generally includes the identifier of the computer used by the accessing user (IP address if the Internet is used), the time of access, the identifier on the server of the accessed node (Web URL).
[0003]
In order to analyze the access history and identify the path of the hyperlink to which each user has transitioned, basically, the accessed nodes may be arranged in time order for each computer. However, it has been difficult to specify a complete path by a computer cache function or the like.
[0004]
In contrast, for example, C.I. Shahabi, A .; M.M. Zarkesh, J. et al. Adibi, and V.D. Shah, “Knowledge Discovery from Users Web-Page Navigation”, Proc. of IEEE RIDE, 1997. The remote agent is used to recognize access to the cache. As a result, a more accurate route can be obtained. However, there is a problem that the cost (time and space) for loading the remote agent program on the user's computer side is sacrificed.
[0005]
As a technique for finding an important route using an access history, for example, J.A. Borges and M.M. Levene, “Minning Association Rules in Hypertext Databases”, Proc. of KDD, 1998. There are techniques described in. This technique first collects access histories and expresses the hyper structure with a directed graph weighted by the amount of hyperlink traffic. In this directed graph, the transition from node A to node B is denoted as A → B, and this is called a combination rule. The join rule is: a confidence value (= total number of transitions A → B with respect to the total number of transitions starting from A) and a support value (= total number of transitions A → B with respect to an average value of the number of transitions of all arcs in the directed graph). ). Further, a combined coupling rule (A → B) & (B → C) & (C → D) &. . . Is defined, and transitions of three or more nodes are evaluated.
[0006]
In this method, since information of a computer identifier (IP address) is not used, the combined link rule merely finds a hyperlink combination route with a large traffic volume. For example, even if it is found that the composite combination rule (A → B) & (B → C) has a high confidence value and support value, the user who actually followed the path of A → B → C It was not always the case.
[0007]
As described above, it is possible to specify the access route of the user or discover the important route by the conventional technique. However, it has not been possible to obtain knowledge for judging the superiority or inferiority of the configuration of a hypertext system (for example, a website).
[0008]
[Problems to be solved by the invention]
The present invention has been made in view of the above-described circumstances, and is capable of recognizing a user's access tendency from an access history, and a hypertext analysis apparatus that assists in obtaining knowledge for determining the superiority or inferiority of the configuration of a hypertext system. It is another object of the present invention to provide a hypertext analysis method. It is another object of the present invention to provide a recording medium on which a hypertext analysis program for realizing the hypertext analysis apparatus or method by a computer is recorded.
[0009]
[Means for Solving the Problems]
The present invention performs clustering on the nodes constituting the hypertext system based on the access history information to the hypertext system, and calculates the hyperlink cohesion degree between the nodes constituting the cluster for each obtained cluster. Then, the calculated hyperlink cohesion degree is displayed. Since the displayed hyperlink cohesion is a value based on the history accessed by the user, it indicates the user's access tendency. Therefore, by obtaining the hyperlink cohesion degree, for example, it is possible to determine the superiority or inferiority of the configuration of the hypertext system together with the hyperlink configuration of the hypertext system (for example, Web site).
[0010]
DETAILED DESCRIPTION OF THE INVENTION
FIG. 1 is a configuration diagram showing a first embodiment of the present invention, and FIG. 2 is an explanatory diagram of an example of the Web. In the figure, 1 is a Web server, 2 is a hypertext analysis device, 11 is access history information, 21 is an access trend analysis unit, 22 is a hyperlink configuration analysis unit, and 23 is a hyperlink cohesion degree display unit. A typical hypertext system configured on a network is the Web. Hereinafter, the Web will be described as an example.
[0011]
The web server 1 is a means for transmitting information on a network. As shown in FIG. 2, the Web server 1 stores information to be provided to the user in a hyper structure including nodes (indicated by rectangles) and hyperlinks (indicated by arrows). The user can obtain content by accessing the Web server 1. At this time, in the Web server 1, in general, every time a user accesses, an access history includes a computer identifier (IP address) for identifying the user's computer, an access time, and an address (URL) of the node accessed by the user. Information 11 is recorded.
[0012]
The hypertext analysis device 2 includes an access tendency analysis unit 21, a hyperlink configuration analysis unit 22, a hyperlink cohesion degree display unit 23, and the like. The access tendency analysis unit 21 performs a clustering process on the node group on the Web server 1 using the corresponding access history information 11. An existing technique can be used for the clustering process. For example, in the clustering method using Agglomerative Hierarchical Clustering, the following steps 1, 2, and 3 are performed.
1. Each node on the Web server 1 is set as one cluster.
2. The similarity between the clusters is calculated, and the clusters having the maximum similarity are merged into one cluster.
3. Repeat step 2 until there is one cluster.
Each cluster sequentially generated during this process is obtained as a clustering processing result. For example: The cluster of only each node generated in the process of FIG. 5, each cluster obtained by merging the clusters, one cluster generated last, and the like are clustering results. Note that the above-described clustering method is, for example, E.I. M.M. Voorhees, “Implementing Aggregative Hierarchical Clustering Algorithms for Use in Document Retrieval”, Information Processing & Management, Vol. 22, no. 6, 1986, and the like.
[0013]
In the access tendency analysis unit 21, the above-described clustering process 2. In the similarity calculation in this process, the overlapping degree of the access history information 11 is used. Specifically, for example, paying attention to the IP address in the access history information 11, the number of IP addresses shared between the clusters is defined as the similarity. Alternatively, a value obtained by dividing the number of IP addresses shared between clusters by the number of all IP addresses of the two clusters is defined as similarity. Here, the IP address possessed by the cluster is an IP address that appears in the merged history of all the access histories to the nodes constituting the cluster. The number of IP addresses can be a value obtained by counting different IP addresses obtained by eliminating duplication from such IP addresses. Instead of using the IP address, for example, the access time in the access history information 11 may be used. In this case, it is better to handle in units of date or month instead of units of seconds or minutes.
[0014]
Alternatively, as similarity calculation, a vector is generated in which each cluster has the access history information 11 corresponding to the cluster as a term, and the appearance frequency of the access history information is the value of the term, and the inner product value between the vectors May be used as the similarity.
[0015]
The hyperlink configuration analysis unit 22 calculates, for each cluster generated by the access tendency analysis unit 21, a hyperlink cohesion degree (hereinafter simply referred to as cohesion degree) between nodes constituting the cluster. For example, if one or more hyperlinks exist between nodes, and the degree of cohesion is defined as having a connection between the nodes, the total number of connections between nodes that make up the cluster is the total number of connections that make up the cluster. The value can be divided by the number of combinations that select two from the nodes. That is, when the total number of node-to-node connections is L and the number of nodes is N,
Cohesion degree = L / _{NC 2}
Can be calculated with The value of the cohesion degree of each cluster calculated by the hyperlink configuration analysis unit 22 is passed to the hyperlink cohesion degree display unit 23.
[0016]
The hyperlink cohesion degree display unit 23 displays the cohesion degree value calculated by the hyperlink configuration analysis unit 22 in an easy-to-use form. FIG. 3 is an explanatory diagram of an example of display by the hyperlink cohesion degree display unit. For example, the cohesion degree value can be divided into several segments, and the number of clusters entering each segment can be displayed in a bar graph. In FIG. 3, two Web sites (Web site A and Web site B) are displayed side by side with the distribution of the number of clusters on the vertical axis and the cohesion degree on the horizontal axis. As described above, since each cluster is configured based on access history information, the node group configuring each cluster has a strong tendency to be accessed before and after by the same user, for example. When the degree of cohesion of the node group is high, the user is provided with a large number of paths for transition between nodes, and thus browsing can be performed efficiently. On the other hand, if the degree of cohesion of the node group is low, browsing efficiency is deteriorated. In FIG. 3, it can be seen at a glance that the website A has many clusters with a low degree of cohesion, and the website B has many clusters with a high degree of cohesion. By this display, it can be easily determined that the hypertext system is constructed in the website B with a configuration superior to the website A.
[0017]
The hyperlink cohesion degree display unit 23 can display the cohesion degree in various display forms in addition to the display form shown in FIG. FIG. 4 is an explanatory diagram of another example of display by the hyperlink cohesion degree display unit. In the example shown in FIG. 4, the relationship between the cluster size and the cohesion degree value is displayed. Each point in the figure represents a cluster. 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, or the like can be used. In the example shown in FIG. 4, an example is shown in which the number of nodes constituting the cluster is used as the size of the cluster. When such a display is performed, if the cluster size is the same, it can be determined that the higher the unity, the better the configuration. Also, with such a display, it is possible to overlook the superiority or inferiority of the configuration of each cluster while paying attention to the cluster size.
[0018]
FIG. 5 is an explanatory diagram of a display example when a cluster is selected in another example of display by the hyperlink cohesion degree display unit. When the cohesion degree display as shown in FIG. 4 is performed, if a point representing a cluster is selected with a pointing device such as a mouse, as shown in FIG. 5, the URLs of the nodes constituting the selected cluster ( Identifier) can be displayed. Alternatively, the title of the node may be displayed instead of the URL. Further, when a node (displayed by URL) displayed as shown in FIG. 5 is selected with a pointing device such as a mouse, the selected node is accessed through the network, and the content of the node is acquired. May be displayed. With such a display, the website administrator can access the problem part in the website, refer to the content, and further edit it, so that the configuration in the website can be easily improved. can go.
[0019]
FIG. 6 is an explanatory diagram of another display example when a cluster is selected in another example of display by the hyperlink cohesion degree display unit. As in the example shown in FIG. 5, when the cohesion degree display as shown in FIG. 4 is performed, a point representing a cluster is selected by a pointing device such as a mouse, as shown in FIG. It is also possible to display hyperlinks between nodes constituting the selected cluster. Here, the labels of the nodes (N1, N2,..., N10 in FIG. 6) may be URLs or titles. By performing such display, the administrator of the Web site can easily find the problem part in the Web site. For example, in the display example of the hyperlink in the cluster shown in FIG. 6, since there is no hyperlink between N1 to N7 and N8 to N10, the user cannot easily go back and forth between the two groups.
[0020]
FIG. 7 is an explanatory diagram of still another example of display by the hyperlink cohesion degree display unit. In this example, the relationship between the similarity between nodes in the cluster and the value of cohesion is displayed. As the similarity between the nodes in the cluster, for example, the value of the similarity used at the time of cluster generation in the access tendency analysis unit 21 can be used. In such a display, if the similarity within the cluster is the same, it can be determined that the higher the cohesion, the better the configuration. With such a display, it is possible to overlook the superiority or inferiority of the configuration of each cluster while paying attention to the similarity between nodes in the cluster.
[0021]
The hyperlink cohesion degree display unit 23 can display the cohesion degree in an arbitrary form regardless of the display forms shown in the above examples.
[0022]
FIG. 8 is a block diagram showing a second embodiment of the present invention. In the figure, the same parts as those in FIG. Reference numeral 24 denotes an identifier acquisition unit. The identifier acquisition unit 24 acquires from the access history information 11 an identifier (for example, an IP address) of a computer that has accessed for a predetermined period for each node on the Web server 1.
[0023]
As in the first embodiment, the access trend analysis unit 21 performs clustering processing on the node group on the Web server 1 using the access history information 11 corresponding to each node group. When calculating the similarity in this clustering process, the degree of duplication of computer identifiers acquired by the identifier acquisition unit 24 can be used.
[0024]
FIG. 9 is a flowchart illustrating an example of processing in the identifier acquisition unit. First, in S31, only access related information within a predetermined period is extracted from the access history information 11 and the rest are discarded. In S32, the number (different number) N of different identifiers among all computer identifiers (IP addresses) existing in the access history information 11 extracted in S31 is obtained. Next, in S33, a vector having a dimension number N is generated for each node for all nodes existing in the access history information 11. Here, each term of the vector corresponds to different computer identifiers (IP addresses) arranged in ascending order (or descending order) of the alphabet. The initial value of each term is set to 0. In step S34, the computer identifier (IP address) of the computer that has accessed each node is extracted from the access history information 11, and 1 is added to the corresponding vector term. This process is sequentially performed for all access history information 11.
[0025]
Using the N-dimensional vector thus obtained, the access tendency analysis unit 21 can perform clustering using the magnitude of the inner product value between the vectors as the similarity. Here, when the clusters are merged, if the sum of the respective vectors is taken, this becomes the vector of the merged cluster.
[0026]
The subsequent processing is the same as that in the first embodiment. The hyperlink configuration analysis unit 22 calculates the cohesion degree, and the hyperlink cohesion degree display unit 23 displays the cohesion degree. At this time, the cohesion degree can be displayed in various display forms as described above or in various other display forms.
[0027]
FIG. 10 is a block diagram showing a third embodiment of the present invention. In the figure, the same parts as those in FIG. Reference numeral 25 denotes a computer identifier shaping unit. The computer identifier shaping unit 25 collects information automatically by comprehensively accessing the Web, not from a computer operated by a user (= human) among accesses to the Web server 1. Eliminate access by computers such as robots. For example, whether or not the access is from the information collection robot depending on whether or not there is access to a special node (in the Web, for example, a file named robots.txt that is placed directly under the root) accessed by the information collection robot according to a certain convention Can be judged. Alternatively, the information collecting robot can be identified based on the presence or absence of a characteristic behavior of the information collecting robot that comprehensively accesses a large number of nodes in a short period of time, and whether or not the computer identifier is a known information collecting robot. Can do. For the computer determined to be an access from the information collecting robot, the access history information 11 related to the computer identifier can be prevented from being acquired by the identifier acquiring unit 24.
[0028]
As a result, it is possible to eliminate access by a computer that collects information automatically by comprehensively accessing the Web, such as an information collecting robot, and to remove these influences on analysis results. It is possible to obtain an analysis result that correctly reflects.
[0029]
Note that other configurations and operations in the third embodiment are the same as those in the second embodiment described above.
[0030]
FIG. 11 is a block diagram showing a fourth embodiment of the present invention. In the figure, the same parts as those in FIG. Reference numeral 26 denotes a hyperlink unity degree evaluation unit. The hyperlink cohesion degree evaluation unit 26 compares the cohesion degree value of each cluster obtained by the hyperlink configuration analysis unit 22 with a predetermined threshold value, and displays a cluster whose cohesion degree is smaller than the threshold value. Pass to part 23. Thereby, in the hyperlink cohesion degree display unit 23, it is possible to easily obtain a cluster having a low cohesion degree, that is, an inferior configuration and nodes constituting the cluster.
[0031]
FIG. 12 is an explanatory diagram of still another example of display by the hyperlink cohesion degree display unit. In the display example shown in FIG. 12, the cluster and the identifiers (URLs) of the nodes constituting the cluster are displayed in ascending order of cohesion value. The cluster is shown in the “ID” column. Here, “ID” is a number uniquely assigned to refer to the cluster. For example, when the cluster is generated in the access tendency analysis unit 21, numbers may be assigned in the order of generation. Since such a display is displayed from a part that is inferior in structure, it is possible to easily know a part that becomes a bottleneck when the user uses it. Of course, also in the fourth embodiment, the cohesion degree can be displayed in various display forms as shown in the first embodiment or other display forms. Further, the identifier acquisition unit 24 and the computer identifier shaping unit 25 described in the second and third embodiments may be provided.
[0032]
The above-described embodiment can also be realized by a computer program. In that case, the program, data used by the program, and the like can be recorded in a computer-readable storage medium. A storage medium is a signal format that causes a state of change in energy such as magnetism, light, electricity, etc. according to the description of a program to a reader provided in the hardware resources of a computer. Thus, the description content of the program 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.
[0033]
【The invention's effect】
As is clear from the above description, according to the present invention, clustering is performed on the nodes constituting the hypertext system based on the user access history information, and the obtained hypertext between the nodes constituting each cluster is obtained. Calculate link cohesion. Based on the degree of hyperlink cohesion, it is possible to easily determine the superiority or inferiority of the configuration of the hypertext system. For example, there is an effect that a Web administrator can change a portion having a configuration problem to construct a hypertext system having a better configuration.
[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 an example of the Web.
FIG. 3 is an explanatory diagram of an example of display by a hyperlink cohesion degree display unit.
FIG. 4 is an explanatory diagram of another example of display by a hyperlink cohesion degree display unit.
FIG. 5 is an explanatory diagram of a display example when a cluster is selected in another example of display by the hyperlink cohesion degree display unit.
FIG. 6 is an explanatory diagram of another display example when a cluster is selected in another example of display by the hyperlink cohesion degree display unit.
FIG. 7 is an explanatory diagram of still another example of display by a hyperlink cohesion degree display unit.
FIG. 8 is a block diagram showing a second embodiment of the present invention.
FIG. 9 is a flowchart illustrating an example of processing in an identifier acquisition unit.
FIG. 10 is a configuration diagram showing a third embodiment of the present invention.
FIG. 11 is a block diagram showing a fourth embodiment of the present invention.
FIG. 12 is an explanatory diagram of still another example of display by the hyperlink cohesion degree display unit.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Web server, 2 ... Hypertext analysis apparatus, 11 ... Access history information, 21 ... Access tendency analysis part, 22 ... Hyperlink structure analysis part, 23 ... Hyperlink cohesion degree display part, 24 ... Identifier acquisition part, 25 ... Computer identifier shaping unit, 26... Hyperlink cohesion degree evaluation unit.

Claims (7)

  Access trend analysis means for clustering the nodes constituting the hypertext system based on access history information to the hypertext system, and between the nodes constituting the cluster for each cluster obtained by the access trend analysis means A hypertext analysis apparatus, comprising: a hyperlink configuration analysis unit that calculates a hyperlink cohesion degree of the hyperlink; and a hyperlink cohesion degree display unit that displays the hyperlink cohesion degree obtained by the hyperlink configuration analysis unit.   Identifier acquisition means for permitting duplication and holding identifiers of computers that have been accessed for a predetermined period for each node constituting the hypertext system, and the access tendency analysis means includes the identifier acquisition means The hypertext analysis apparatus according to claim 1, wherein clustering is performed by determining a degree of duplication of identifiers held.   And an identifier shaping unit that specifies an identifier of a computer that collects information automatically by comprehensively accessing the hypertext system and deletes the identifier from the list held by the identifier acquisition unit. The hypertext analysis apparatus according to claim 2, wherein the apparatus is a hypertext analysis apparatus.   Hyperlink cohesion degree evaluation means for determining whether or not the hyperlink cohesion degree satisfies a predetermined condition, and the hyperlink cohesion degree display means is configured to display a hyperlink according to a determination result in the hyperlink cohesion degree evaluation means. 4. The hypertext analysis apparatus according to claim 1, wherein a link cohesion degree is displayed.   5. The hyperlink cohesion degree display means further displays a node constituting the cluster obtained by the access tendency analysis means and a hyperlink between the nodes. The hypertext analysis apparatus according to item 1.   Based on the access history information to the hypertext system, the access tendency analysis means performs clustering on the nodes constituting the hypertext system, and for each obtained cluster, the degree of hyperlink cohesion between the nodes constituting the cluster is determined. A hypertext analysis method, characterized in that the hyperlink structure analysis means calculates and the hyperlink cohesion display means displays the calculated hyperlink cohesion degree.   Access trend analysis processing for clustering nodes constituting the hypertext system based on access history information to the hypertext system, and between the nodes constituting the cluster for each cluster obtained by the access trend analysis processing Hyperlink cohesion degree calculation processing for calculating the hyperlink cohesion degree, and hyperlink cohesion degree display for displaying the hyperlink cohesion degree obtained by the hyperlink constitution analysis process and indicating the superiority or inferiority of the hypertext system configuration A storage medium storing a hypertext analysis program for causing a computer to execute processing.

Images