JP2010061187A - Quotation graph output device and quotation graph output program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide technology allowing output of a graph representing quotation relation between documents, allowing comprehension of each document by classification, and having excellent visibility of the quotation relation. <P>SOLUTION: A cluster-classified classification part 12 classifies the plurality of documents into the plurality of clusters according to a predetermined criterion, and a cluster order determination part performs processing for connecting two clusters wherein the number of external links (links between the documents belonging to different clusters from each other) becomes maximum to form one cluster with the plurality of clusters of a classification result as targets, and repeats the processing with the connected cluster of a processing result and the residual clusters as the next targets until all the clusters are connected. Thereafter, a quotation graph creation part 15 creates the quotation graph with an axis representing information of preparation time of the document as a first axis and with an axis representing information about a connection position of each cluster constituting the connected cluster as a second axis, and an output part 16 outputs it. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a citation graph output device and a citation graph output program.

In documents such as scientific and technical papers and patent documents, other documents for reference are cited, and various techniques have been proposed so that citation relationships between documents can be easily grasped.
For example, a citation relationship between documents is expressed by placing a graphic representing each document in a two-dimensional space and outputting a graph in which citation documents are connected by line segments. .
Such graph output is generally performed using STT. In STT, for example, as shown in FIG. 15, the creation time of a document is represented by one axis constituting a two-dimensional space, and each document is arranged so that the number of intersections of line segments representing citation relationships between documents is minimized. It arrange | positions (refer nonpatent literature 1-3).

K. Sugiyama, S. Tagawa, M. Toda, “Methods for Visual Understanding of Hierarchical System Structure”, IEEE, 1981 Sugiyama Kozo, “Drawing a Hierarchical Graph Focusing on Regeneration—Improved SKETCH Method”, Japan Society for Simulation Technology, 1983 Sugiyama Kozo, “Automatic Graph Drawing Method and Its Applications”, Corona, 1993

  It is an object of the present invention to propose a technique capable of grasping each document by classification and outputting a graph with good citation relationship visibility in outputting a graph representing a citation relationship between documents.

  According to a first aspect of the present invention, there is provided a classification unit that classifies a plurality of documents into a plurality of document sets according to a predetermined criterion, and a plurality of document sets that are classified by the classification unit as a target. The two linked document sets having the maximum number of citations between sets, which are citation relationships in, are linked along a predetermined virtual direction to form one linked document set, and the resulting linked document The above-mentioned processing is performed with the set and the rest of the document set as the next target until all of the target document sets are connected, and the axis representing the document creation time information is the first axis, and the connection Output means for outputting a graph representing a citation relationship between documents, with the axis representing the information on the link position of each document set constituting the linked document set processed by the means as the second axis. To quote An output device.

  According to a second aspect of the present invention, in the first aspect of the present invention, for each of the plurality of document sets of the classification result by the classification unit, the classification is performed for each division line of the first axis divided in a predetermined unit. The number of citation relationships across the lines is counted to obtain the maximum value of the count values in all the division lines, and each document set in the second axis is determined according to the ratio of the maximum values obtained for each document set. A citation graph output device comprising a width determining means for determining a width of a region to be represented.

  According to a third aspect of the present invention, in the first or second aspect of the present invention, the concatenating unit determines the number of citations between sets between the connected document set and another document set as one end document set in the connected document set. The number of citations between sets between the document set and the other document set, or the number of citations between sets between the other document set and the other document set in the linked document set, A citation graph output device characterized by the above.

  According to a fourth aspect of the present invention, in the first or second aspect of the present invention, the linking means uses each document constituting the linked document set as the number of citations between sets between the linked document set and another document set. An citation graph output device using an average value of the number of citations between sets between each set and the other document set.

  According to a fifth aspect of the present invention, in a computer, a classification function for classifying a plurality of documents into a plurality of document sets according to a predetermined criterion, and a plurality of document sets obtained as a result of classification by the classification function are classified into different document sets. A processing result is obtained by connecting two document sets having the maximum number of citations between sets, which is a citation relationship between the documents to which the documents belong, along a predetermined virtual direction to form one linked document set. The first axis is a link function that repeats the above processing for the next set of linked document sets and the remaining set of documents until all of the target document sets are linked, and an axis that represents information on the creation time of the document. And an output function for outputting a graph representing a citation relationship between documents, with the axis representing the information on the link position of each document set constituting the linked document set as a result of processing by the link function as the second axis. A citation graph output program.

  According to the first aspect of the present invention, in outputting a graph representing a citation relationship between documents, it is possible to output a graph in which each document can be grasped by classification and the visibility of the citation relationship is good.

  According to the second aspect of the present invention, the width of each classification area in the graph can be made appropriate.

  According to the third aspect of the present invention, based on the number of citations between sets with the document set at one end or the other end to which the other document set is connected among the document sets constituting the connected document set, the document set to be connected Can be determined.

  According to the fourth aspect of the present invention, the document set to be linked can be determined based on the number of inter-set citations with all the document sets constituting the linked document set.

  According to the fifth aspect of the present invention, in outputting a graph representing a citation relationship between documents, it is possible to output a graph in which each document can be grasped by classification and the citation relationship is highly visible.

The present invention will be specifically described based on an example of an embodiment.
The citation graph output apparatus according to an example of the present invention outputs a citation graph as illustrated in FIG. In the citation graph of this example, as outlined in FIG. 3, the document creation time is defined by one axis (vertical axis in this example), and the other axis (horizontal axis in this example) orthogonal to this is defined. The classification (document collection) to which the document belongs is defined, and the creation time and classification of each document can be grasped by arranging the graphic representing each document at the corresponding position on the graph area. In addition, by linking documents in a citation relationship with a directed line segment (hereinafter referred to as a link) with the citation source document as the start node and the citation destination document as the end node, the citation relationship between documents ( Link relationship).

In the chronological display of citations, the greater the number of crossing links representing citation relationships, the lower the visibility (see Non-Patent Document 3). It is required to reduce the number of link crossings. In the cited graph of this example, the graph area is divided into areas by classification (by document set), and there is a prior art for minimizing the number of link intersections in the same area (non-patent literature). 1 and 2).
Here, as a link in the graph area divided into areas, in addition to the links in the area as described above (quotation within a set that is a citation relationship between documents belonging to the same document set), links between different areas (different) There is also a citation between sets that is a citation relationship between documents belonging to a document set. And among the links between such areas, the link between the areas straddling one or more other areas (thick line link in FIG. 4) is the link between adjacent areas (thin line link in FIG. 4). Since the influence on the visibility of the citation relationship is larger than that of, in order to improve the visibility of the citation relationship, it is required to reduce the number of links between areas straddling other areas.

  Therefore, in the citation graph output device of this example, in order to output a citation graph satisfying the above-described requirements, as illustrated in a functional block diagram in FIG. 1, a document database 11, a cluster-by-cluster classification unit 12, a cluster order determination unit 13, a cluster width determination unit 14, a quote graph creation unit 15, and an output unit 16.

  The document database 11 stores and manages various information related to documents (for example, information related to document citation, information related to document creation time, and information related to document classification). In this example, the document database 11 is provided in the citation graph output device, but the document database 11 may be provided in another device accessible by the citation graph output device.

  The cluster classification unit 12 classifies a plurality of documents into a plurality of document sets (hereinafter referred to as clusters) according to a predetermined standard. Various criteria can be adopted as classification criteria. For example, when a patent application is a target document, classification based on the applicant, classification based on F-term, classification based on similarity of contents of the text And so on. Specifically, the cluster classification unit 12 obtains information related to document classification (information such as applicant, F-term, and body text) by accessing the document database 11, and based on the information obtained for each document. Perform classification. The classification may be performed in consideration of a plurality of criteria comprehensively.

The cluster order determination unit 13 determines the arrangement of the citation graph on the horizontal axis for each cluster of the classification result by the cluster classification unit 12 according to the following procedure. In the following description, links between documents belonging to the same cluster are referred to as internal links, and links between documents belonging to different clusters are referred to as external links.
First, for each cluster of the classification result, the number of external links is obtained for each combination of two clusters, the two clusters having the maximum number of external links are identified, and these clusters are preliminarily virtualized. Are processed along a predetermined direction to form one cluster. Thereafter, the same processing as the above processing is performed with the cluster of the connection result and the remaining cluster as the next target, and is repeated until all the clusters of the classification result are connected. Then, the connection position of each cluster is determined as the arrangement of each cluster on the horizontal axis of the cited graph.

The above process will be described more specifically.
A cluster consisting of only one cluster is a single cluster, and a cluster in which a plurality of clusters are connected is a connected cluster.
First, how to determine the number of external links will be described. When both clusters in the combination are single clusters, the number of links between documents belonging to one cluster and documents belonging to the other cluster is the number of external links in this combination. When at least one of the combinations is a connected cluster, the number of external links can be obtained by various methods, and two calculation methods will be described below as an example.
The first calculation method considers only clusters at both ends of the connected cluster, and does not consider internal clusters sandwiched between these clusters. That is, as shown in FIG. 5, the number of external links between a connected cluster j in which a plurality of clusters p,..., Q are connected in order and another cluster (single cluster or connected cluster) i is Of the number of external links between the cluster p located at one end of the connected cluster j and the other cluster i, or the number of external links between the cluster q located at the other end of the connected cluster j and the other cluster i Many people.
The second calculation method considers all the clusters constituting the connected cluster. That is, as shown in FIG. 6, the number of external links between a connected cluster j in which a plurality of clusters p,..., Q are connected in order and another cluster (single cluster or connected cluster) i is Let the average value of the number of external links between each of the clusters p,..., Q constituting the connected cluster j and the other cluster i.

When the number of external links for each combination of two clusters is obtained as described above, the two clusters of the combination having the maximum number of external links are identified as the connection targets and are connected to each other.
For convenience of explanation, a direction similar to the classification axis in the citation graph (ie, the horizontal direction) is used as the cluster connection direction, but this connection direction is only a virtual one. In short, each cluster is connected along a certain direction, and the arrangement of the clusters in the cited graph may be determined according to the connection position.

  Here, the connection when one of the two clusters to be connected is a single cluster and the other is a connected cluster will be described with reference to FIG. That is, when connecting a connected cluster j and a single cluster i in which a plurality of clusters p,..., Q are connected in order, between the cluster p located at one end of the connected cluster j and the single cluster i. Is compared with the number of external links between the cluster q located at the other end of the connected cluster j and the single cluster i, and the single cluster i is connected to the one with the larger number of external links. Specifically, for example, when the number of external links with the cluster p located at the left end is 2 and the number of external links with the cluster q located at the right end is 3, a single cluster i is placed on the right side of the cluster q. Link.

  In addition, when the two clusters to be connected are both connected clusters, the connected clusters are connected so that the sides with the larger number of external links face each other. That is, when the number of external links in each connected cluster is different (one connected cluster has more right external links and the other connected cluster has more left external links), each connected cluster If the connection direction is maintained and the side with the larger number of links matches (if both of the connected clusters have a large number of external links on the right side or a large number of external links on the left side), one connected cluster Is reversed and the connection direction is reversed.

Next, identification of a cluster to be connected and its connection will be described based on a specific example.
First, a case where the first calculation method is used will be described.
For example, as shown in FIG. 7, there are four clusters a, b, c, and d, and the number of external links between each cluster is 3, between a and b, 11, between b and c, Assuming 5 between d and 7 between c and d, a correlation matrix C = {c _ij } representing the relationship of the number of external links between these clusters (where c _ij = between cluster i and cluster j) The number of external links, c _ii = 0) is expressed as shown in FIG. In this case, since the combination of clusters with the maximum number of external links is the single cluster b and the single cluster c (the number of external links = 11), these are connected to each other to form a connected cluster (b, c). .
Thereafter, when the number of external links is recalculated using the first calculation method, the correlation matrix C is represented as shown in FIG. 8 (2), and a combination of clusters in which the number of external links in the correlation matrix C is maximum is a connected cluster. Since the cluster c and the single cluster d are located at the right end of (b, c) (the number of external links = 7), the single cluster d is connected to the right side of the connected cluster (b, c) and the connected cluster (( b, c), d).
Thereafter, when the number of external links is recalculated using the first calculation method, the correlation matrix C is represented as shown in FIG. 8 (3), and the combination of clusters in which the number of external links in the correlation matrix C is maximum is connected. Since the cluster b located at the left end of the cluster ((b, c), d) and the single cluster a (the number of external links = 3), the single cluster on the left side of the connected cluster ((b, c), d) a is connected to form a connected cluster (a, ((b, c), d)).
As a result, all the clusters are connected, so that the correlation matrix C has only elements 0 as shown in FIG. 8 (4), and the connection process ends. As a result, the arrangement of each cluster on the horizontal axis of the cited graph is determined according to the connection position of the connection cluster (a, ((b, c), d)).

Next, a case where the second calculation method is used will be described.
Since the first connection is the same as in the case of the first calculation method, the description thereof is omitted.
When the single cluster b and the single cluster c are connected to form a connected cluster (b, c), and the number of external links is recalculated using the second calculation method, the correlation matrix C is as shown in FIG. Since the combination of clusters with the maximum number of external links in the correlation matrix C is a connected cluster (b, c) and a single cluster d (number of external links = 6), the connected cluster (b, c ), The single cluster d is connected to the right side, which is the one with the larger number of external links (cluster c), to form a connected cluster ((b, c), d).
Thereafter, when the number of external links is recalculated using the second calculation method, the correlation matrix C is represented as shown in FIG. 9 (3), and the combination of clusters in which the number of external links in the correlation matrix C is maximum is connected. Since the cluster ((b, c), d) and the single cluster a (external link number = 1), the larger number of external links (cluster b) of both ends of the connected cluster ((b, c), d) The single cluster a is connected to the left side, which is the one), and becomes a connected cluster (a, ((b, c), d)).
As a result, all the clusters are connected, so that the correlation matrix C has only elements 0 as shown in FIG. 9 (4), and the connection process is completed. As a result, the arrangement of each cluster on the horizontal axis of the cited graph is determined according to the connection position of the connection cluster (a, ((b, c), d)).

The cluster width determination unit 14 determines the width of the area representing the cluster on the horizontal axis of the cited graph for each cluster of the classification result by the cluster classification unit 12 according to the following procedure. In addition, the vertical axis of the quote graph is divided by a predetermined unit (for example, one year), and a layer is provided for each division line representing the division, and each document is arranged on the layer corresponding to the creation time. Shall be.
First, for each layer, the number of intersections (hereinafter referred to as “dummy nodes”) between the layers and internal links for each layer (that is, the number of internal links across the partition line) is counted. The maximum value of the count value at is obtained. Then, the ratio of the width of each cluster on the horizontal axis of the citation graph is determined according to the ratio of the maximum number of dummy nodes obtained for each cluster.
FIG. 10 schematically shows the determined cluster width, and it can be seen from FIG. 10 that a wider width is set in the order of clusters b, c, d, and a.

  The width of each cluster on the horizontal axis of the citation graph may be determined in consideration of the number of documents arranged on each layer. That is, for each cluster, the number of documents arranged on that layer and the number of dummy nodes are totaled for each layer to obtain the maximum value of the total value in all the layers, and for each cluster. The ratio of the width of each cluster on the horizontal axis of the cited graph is determined according to the ratio of the maximum values obtained.

The citation graph creation unit 15 creates a citation graph as described with reference to FIG. 2 based on the processing results of the cluster order determination unit 13 and the cluster width determination unit 14 according to the following procedure.
First, the vertical axis representing the document creation time information is divided by a predetermined unit (for example, one year), and the horizontal axis representing the cluster information to which the document belongs is assigned to the cluster determined by the cluster order determination unit 13. Sort by line. At this time, the width of each cluster is the width determined by the cluster width determination unit 14. Then, using a well-known STT method, each document is placed at a position on the citation graph according to its creation time and the category to which it belongs, and the documents having the internal link relationship are connected to each other. Connect documents with external links without changing the layout.

  The output unit 16 outputs the citation graph created by the citation graph creation unit 15. In this example, display output is performed by the display device. However, for example, printing may be performed on a medium such as paper by the printing device, and output to the storage device for storage as data. May be.

FIG. 11 shows an example of a citation graph creation flow.
A chronological axis representing the creation time of the document is defined in the two-dimensional space as the graph area (step S11). Next, the arrangement order and area width of each cluster i (i = 1 to n) on the cluster axis orthogonal to the chronological axis are determined (steps S12 and S13). Next, for each of the clusters i (i = 1 to n), a display showing the documents and internal links in the cluster i is formed on the graph area using the STT method (steps S14, S15, and S16). Next, a display showing an external link is formed on the graph area using the STT method (step S17).

The process of determining the arrangement order of the regions of each cluster i (i = 1 to n) (step S12) will be specifically described with reference to FIG.
A correlation matrix C = {c _ij } (where c _ij = the number of external links between cluster i and cluster j, c _ii = 0) is generated (step S21). Next, the cluster k and the cluster l having the maximum matrix value (c _ij ) of the correlation matrix C are specified (step S22), and the cluster k and the cluster l are connected to each other to form a connected cluster (step S23). Next, the correlation matrix C is regenerated based on the calculation method (the first calculation method or the second calculation method) (step S24), and after step S22 until the regenerated correlation matrix C becomes empty. The above process is repeated (step S25).

The process of determining the width of the area of each cluster i (i = 1 to n) (step S13) will be specifically described with reference to FIG.
The x-axis area width (span_i) of each cluster i (i = 1 to n) in the two-dimensional space of the graph area represented by the y-axis (age axis) and the x-axis (cluster axis) The determined number of dummy nodes is determined based on the ratio of the maximum values in all layers (step S31). Next, i is incremented by 1 until x ₀ = 0 and i = 1 as initial values until i> n. In the process, the coordinate value (x _i ) of the end point (right end) of the cluster i on the x axis is added. ) Is obtained by adding the area width (span_i) of the cluster i to the coordinate value (x _i-1 ) of the end point on the x-axis of the immediately preceding cluster i-1 (steps S32 to S34).

FIG. 14 illustrates the main hardware of a computer that implements the quote graph output apparatus of this example.
In this example, a CPU (Central Processing Unit) 21 that performs various arithmetic processing, a RAM (Random Access Memory) 22 that is a work area of the CPU 21, a ROM (Read Only Memory) 23 that stores a basic control program, An HDD (Hard Disc Drive) 24 for storing programs and various data for realizing the functions according to the embodiment, a display for displaying information for an operator, a mouse / keyboard for receiving input of information from the operator, and the like It is constituted by a computer having hardware resources such as an input / output I / F 25 that is an interface with a device and a communication I / F 26 that is an interface for communicating with other devices.

Then, a program for realizing the function according to the embodiment of the present invention is read from the HDD 24, loaded into the RAM 22, and executed by the CPU 21, whereby each functional unit according to the embodiment of the present invention is executed by a computer. Realized.
In this example, the classification means 12 according to the present invention is realized by the cluster-specific classification unit 12, the connection means according to the present invention is realized by the cluster order determination unit 13, and the width determination unit according to the present invention is realized by the cluster width determination unit 14. The citation graph creating unit 15 and the output unit 16 realize the output unit according to the present invention.

In this example, each functional unit is provided in one device, but each functional unit may be provided in a distributed manner in a plurality of devices.
Each functional unit according to the present invention is not limited to a mode realized by a software configuration as in this example, and may be configured by a dedicated hardware module.
In addition, the program according to the present invention is provided to the user of the program according to the present invention in a format in which an external storage medium such as a CD-ROM storing the program is distributed, a format in which the program is distributed through a communication network, or the like. can do.

It is a figure which illustrates the functional block of the quotation graph output device which concerns on an example of this invention. It is a figure which illustrates the quotation graph which concerns on an example of this invention. It is a figure which shows the outline | summary of the quotation graph which concerns on an example of this invention. It is a figure explaining the quotation graph which concerns on an example of this invention. It is a figure explaining the calculation method of the number of links between clusters concerning an example of the present invention. It is a figure explaining the calculation method of the number of links between clusters concerning an example of the present invention. It is a figure which shows the number of links between the clusters which concern on an example of this invention. It is a figure explaining the connection of the cluster which concerns on an example of this invention. It is a figure explaining the connection of the cluster which concerns on an example of this invention. It is a figure which represents typically the cluster width in the quote graph which concerns on an example of this invention. It is a figure which illustrates the processing flow which produces the quotation graph which concerns on an example of this invention. It is a figure which illustrates the processing flow which determines the cluster order in the quotation graph which concerns on an example of this invention. It is a figure which illustrates the processing flow which determines the cluster width in the quote graph which concerns on an example of this invention. It is a figure which illustrates the hardware of the computer which concerns on an example of this invention. It is a figure which illustrates the quotation graph by a conventional system.

Explanation of symbols

  11: Document database, 12: Classification unit by cluster, 13: Cluster order determination unit, 14: Cluster width determination unit, 15: Citation graph creation unit, 16: Output unit

Claims (5)

A classifying means for classifying a plurality of documents into a plurality of document sets according to a predetermined standard;
A direction in which two document sets having the maximum number of citations between sets, which are citation relationships between documents belonging to different document sets, are virtually determined in advance for a plurality of document sets obtained by the classification means. Are connected to each other and processed as one connected document set, and the above processing is performed with the connected document set as a result and the remaining document set as the next target until all the target document sets are connected. Repeating connection means;
The citation relationship between documents, with the axis representing the document creation time information as the first axis and the axis representing the information on the link position of each document set constituting the linked document set as a result of processing by the linking means as the second axis An output means for outputting a graph representing
A citation graph output device comprising: For each of a plurality of document sets obtained as a result of classification by the classification means, for each division line of the first axis divided in a predetermined unit, the number of citation relations across the division line is counted and all the divisions are counted. A width determining means for determining a maximum value of the count value in the line and determining a width of an area representing each document set in the second axis according to a ratio of the maximum value obtained for each document set;
The citation graph output device according to claim 1. The concatenating means, as the number of inter-set citations between a linked document set and another document set, the number of inter-set citations between a document set at one end of the linked document set and the other document set, or , Using the larger number of inter-set citations between the other document set and the other document set in the linked document set,
The citation graph output device according to claim 1, wherein: The concatenation means is configured to determine the number of citations between sets between a linked document set and another document set as a number of citations between sets of each document set constituting the linked document set and the other document set. Use the average of the numbers,
The citation graph output device according to claim 1, wherein: On the computer,
A classification function for classifying a plurality of documents into a plurality of document sets according to a predetermined standard;
Direction in which two document sets having the maximum number of citations between sets, which are citation relationships between documents belonging to different document sets, are virtually determined in advance for a plurality of document sets of the classification result by the classification function Are connected to each other and processed as one connected document set, and the above processing is performed with the connected document set as a result and the remaining document set as the next target until all the target document sets are connected. Repeat connection function,
The citation relationship between documents, with the axis representing the document creation time information as the first axis and the axis representing the link position information of each document set constituting the linked document set as a result of processing by the link function as the second axis An output function that outputs a graph representing
Quotation graph output program to realize

Images