JP5534167B2 - Graph creation device, graph creation method, and graph creation program - Google Patents

Description

  The present invention relates to a technique for creating a graph for showing the relationship between items according to a graph theory, and in particular, a graph creating device, a graph creating method, and a graph for clarifying the meaning of a link generated between items. Involved in the creation program.

  As a means to understand the relationship between items such as people and phrases, items such as people and phrases are used as nodes, and if one item and another item appear simultaneously on a web page or text, that is, they co-occur Based on information such as whether or not a link is created, a graph in which links are generated between nodes is created.

  For example, in the method described in Patent Document 1, when two words co-occur in the same sentence unit (analysis target), a graph is created by attaching a link to the two words (nodes).

  However, in the graph generated as described above, not all the links in which items such as persons and phrases are connected suggest the same relationship. For this reason, it is performed to estimate what kind of relationship each link in the graph suggests.

  For example, in the method described in Patent Document 2, when a plurality of documents in the hypermedia format (analysis targets) are connected by hyperlinks, terms (nodes) to be extracted from the document in consideration of the document type are selected. The links between the nodes (terms) selected and extracted are estimated as to whether the relationship is a hierarchical relationship or an associative relationship in consideration of the hyperlink between documents and the type of document.

  Further, in the method described in Non-Patent Document 1, a link is generated between two persons (nodes) appearing on the same web page, and the web page includes a specific word prepared in advance. Label links according to whether or not.

  A method for calculating the centrality index of the nodes constituting the graph used in the present invention is described in Non-Patent Document 2, for example. As the centrality index, types such as “medial centrality”, “distance centrality”, “degree centrality”, “Bonachic centrality” are known.

JP 2009-128949 A JP 2008-192170 A

Matsuo Yutaka et al., “Extraction of human relationship network from information on the web”, Japanese Society for Artificial Intelligence Vol.20 No.1E, pp46-56, 2005 Yasuda Yuki, "Practical Network Analysis: Theory and Techniques for Solving Relationships", Shinsyo, October 25, 2001

  However, in the methods described in Patent Document 2 and Non-Patent Document 1, it is necessary to prepare in advance a document type or a specific word for estimating the relationship suggested by the link, and the cost for that purpose. It can point out the problem that it takes.

  In these methods, the link is estimated to have any one of predetermined relationships (for example, a hierarchical relationship or an associative relationship). This is effective when the relationship to be noted in the graph is clear in advance, but cannot be applied when it is not clear in advance which relationship should be noted. For example, when it is desired to find an unexpected link among a plurality of links in the graph, the methods described in Patent Document 2 and Non-Patent Document 1 use a specific, that is, “unexpectedness” in advance. It is necessary to prepare a document type or word that is related to `` there is a problem '', but if it is not clear in advance which relationship should be noted, those documents and words can not be prepared Can point out issues.

  Therefore, in order to solve the above-described problem, the present invention appropriately estimates the meaning of a link generated between each item with respect to a graph created in order to understand the relationship between items such as a person and a phrase. It is an object to provide a graph creation device, a graph creation method, and a graph creation program.

The graph creation device according to the present invention is a set of co-occurrence range data that co-occurs when the same or related item appears, and each co-occurrence range data having an item that becomes a node of the graph to be generated A data acquisition unit that accepts data and a division that identifies each item included in each co-occurrence range data acquired by the data acquisition unit and expresses information about whether each item is included in each co-occurrence range A table is generated from the co-occurrence range data and the relationship between the items included in the co-occurrence range data. And an item graph creating means for generating an item graph by generating a link between the nodes based on the contingency table, and a co-occurrence range of the contingency table created by the contingency table creating means as nodes, A co-occurrence range graph creating means for generating a co-occurrence range graph by generating a link between the nodes based on a contingency table, and both nodes forming each link co-occurring for each link of the item graph Each of the co-occurrence range data according to the desired centrality index obtained by analyzing the difference based on the desired centrality index appearing for each node included in the co-occurrence range graph The analysis result representing the relationship between the data included in the nodes is associated with each other according to the relationship between each link of the item graph obtained by the identification processing and the node of the co-occurrence range graph. The link of the item graph created by the item graph creation means is attached with link information that gives information on the co-occurrence range based on the desired centrality index inherent in each co-occurrence range data Characterized by comprising a means.

In addition, the graph creation device according to the present invention is a set of co-occurrence range data that co-occurs when the same or related item appears, and each co-occurrence range in which an item that is a node of the generated graph is included. Data and a data acquisition means for receiving the relevance between the co-occurrence range data from the reference information between the co-occurrence ranges given in advance to the data, and the co-occurrence range data acquired by the data acquisition means Identify each included item, and use each co-occurrence range to which each identified item belongs as a node, and link between the nodes based on the relationship between the co-occurrence ranges obtained by identifying the reference information From the co-occurrence range graph acquisition unit that acquires the co-occurrence range graph and the co-occurrence range data acquired by the data acquisition unit, and sets each item included in each co-occurrence range as a node. Origin range Item graph creation means for generating an item graph by generating a link between the nodes based on the data, and a co-occurrence range in which both nodes forming each link co-occur for each link of the item graph Included in each node according to the desired centrality index obtained by analyzing the difference based on the desired centrality index appearing for each node included in the co-occurrence range graph and identifying the data The item graph creation is performed by associating the analysis result representing the relationship between the data that has been stored according to the relationship between each link of the item graph obtained by the identification processing and the node of the co-occurrence range graph. Link information giving means for giving information on the co-occurrence range based on the desired centrality index inherent in each co-occurrence range data to the link of the item graph created by means;
It is characterized by comprising.

  According to the present invention, a graph creation device, a graph creation method, and a graph creation program for estimating link information that cannot be understood only from an item graph by using the analysis result of the graph of the co-occurrence range for the graph Can be provided.

It is a block diagram which shows the function structural example of the graph preparation apparatus by this invention. It is a flowchart which shows the example of the process progress which concerns on embodiment of this invention. It is explanatory drawing which shows an example of the acquired data. It is explanatory drawing which shows an example of the division | segmentation table produced based on data. It is explanatory drawing which shows an example of the item graph created based on the contingency table. It is explanatory drawing which shows an example of the co-occurrence range graph produced based on the contingency table. It is explanatory drawing which shows an example of the graph which plotted the median centrality on the horizontal axis and the degree centrality on the vertical axis about each node of the co-occurrence range graph. It is explanatory drawing which shows an example of the item graph which highlighted and displayed the link judged to be "the coupling | bonding containing the unexpected possibility". It is a block diagram which shows the function structural example of 2nd Embodiment of the graph preparation apparatus by this invention. It is a flowchart which shows the example of the process progress which concerns on the 2nd Embodiment of this invention. It is a block diagram which shows the function structural example of 3rd Embodiment of the graph preparation apparatus by this invention. It is a flowchart which shows the example of process progress which concerns on the 3rd Embodiment of this invention.

  Embodiments of the present invention will be described below with reference to the drawings.

[First Embodiment]
FIG. 1 is a block diagram showing a functional configuration example of a graph creating apparatus according to the present invention. As shown in FIG. 1, the graph creation apparatus according to the present invention includes a data acquisition unit 1, a contingency table creation unit 2, an item graph creation unit 3, a co-occurrence range graph creation unit 4, and a link information addition unit 5. including. Specifically, the graph creation device is realized by an information processing device that operates according to a program such as a personal computer. The co-occurrence range refers to a range having a predetermined data structure in which the same or related items appear. For example, in the background art described above, sentences and web pages correspond to the co-occurrence range.

  The data acquisition unit 1 inputs or receives data to be analyzed for graph creation from a recording medium such as an HDD provided in a computer or from a web server accessed via a communication network such as the Internet. It has a function to do. The data acquisition unit 1 may acquire data from, for example, a file or sentence input via an input device such as a keyboard in accordance with a user operation.

  As data to be analyzed, for example, document data, image data, numerical data, various files, etc. stored in a storage medium such as a computer, or a web page configured via a communication network such as the Internet However, it is not limited to these.

  In the description of the embodiment, various means generate data such as graphs, but it is obvious that the generated data is appropriately recorded in a storage medium such as a RAM or an HDD and is readable. I will.

  Specifically, the contingency table creating means 2 is realized by a CPU that is a control unit of an information processing apparatus that operates in accordance with a graph creation program (hereinafter referred to as a program) read from a storage unit such as a ROM and expanded in a RAM. The The contingency table creating means 2 has a function of creating a contingency table that expresses information indicating whether each item is included in each co-occurrence range based on the data acquired by the data acquisition means 1.

  The contingency table is a table in which the relationship of each combination of values that can be taken by two variables (for example, each category in the case of categorical variables) is recorded. Here, in a matrix with items such as persons and phrases as columns and co-occurrence ranges such as web pages and sentences as rows, the i-th row and j-th column components are arranged in the i-th row co-occurrence range. If an item is included, a 1 or a finite value is assigned. Otherwise, a value of 0 is assigned to create a contingency table that expresses information as to whether or not each item is included in each co-occurrence range. In the contingency table, the information that can be expressed does not change even if the roles of the rows and columns are interchanged. If the column is a co-occurrence range and the row is an item, then if the row and column are read later, there is no problem. It will be obvious that implementation is possible.

  In the above description, a person or a phrase is used as an item, and a web page or a sentence is used as a co-occurrence range. However, there is no limitation on what is an item and what is a co-occurrence range. For example, the role of the above item and the co-occurrence range can be considered in reverse. That is, a web page or a sentence may be used as an item, and a person or a phrase may be used as a co-occurrence range.

  Specifically, the item graph creation means 3 is realized by a CPU of an information processing apparatus that operates according to a program. The item graph creation means 3 uses the items of the contingency table created by the contingency table creation means 2 as nodes, generates links between the nodes based on the relationship between the nodes represented in the contingency table, and creates an item graph. It has a function.

  More specifically, the co-occurrence range graph creating means 4 is realized by a CPU of an information processing apparatus that operates according to a program. The co-occurrence range graph creation means 4 uses the co-occurrence range of the contingency table created by the contingency table creation means 2 as a node, generates a link between each node based on the relationship between the nodes represented in the contingency table, A function to create a range graph is provided.

  Specifically, the link information providing unit 5 is realized by a CPU of an information processing apparatus that operates according to a program. The link information adding means 5 links the item graph created by the item graph creating means 3 based on the co-occurrence range graph created by the co-occurrence range graph creating means 4 and the contingency table created by the contingency table creating means 2. It has a function to give information.

  Needless to say, in addition to the above-described various means, a storage means for storing and holding the created graph and a display means for displaying the created graph on the screen are provided as necessary.

  Next, the operation of the graph creation device will be described. FIG. 2 is a flowchart illustrating a processing example of the graph creation apparatus.

For example, when the user performs an operation of inputting data to the graph creation device, the data acquisition unit 1 acquires data according to the input operation of the user (step S11). Further, the data acquisition unit 1 may automatically collect data from a web page or the like, for example, every predetermined time.
FIG. 3 is an explanatory diagram illustrating an example of data acquired by the data acquisition unit 1. In the example shown in FIG. 3, x1, y1, z1, etc. are co-occurrence ranges, and a1, b1, c1, d1, etc. are items included in each co-occurrence range. The co-occurrence range is, for example, a sentence, and the item is, for example, a phrase included in each sentence. As another example, in a questionnaire or the like, one respondent's free description answer (usually including a plurality of sentences) is included as one co-occurrence range, that is, x1, etc. It is conceivable that the phrase is an item, that is, a1. It is also conceivable that the web page is a co-occurrence range and the name of a person who appears on each web page is an item.

Next, the contingency table creating means 2 creates a contingency table expressing information indicating whether each item is included in each co-occurrence range based on the data acquired by the data acquiring means 1 (step S21).
FIG. 4 shows an example of the contingency table created by the contingency table creating means 2 based on the data shown in FIG. In the example shown in FIG. 4, in the matrix in which items such as a1 and b1 are columns and the co-occurrence ranges such as x1 and y1 are rows, the components of the i-th row and j-th column are arranged in the co-occurrence range of the i-th row. By assigning a value of “1” if items in the j column are included, and “0” otherwise, information indicating whether each item is included in each co-occurrence range is expressed.
In the data shown in FIG. 3, items do not overlap in each co-occurrence range, but in general, there may be cases where items overlap. The contingency table created in that case may give the number of items in the j-th column in the co-occurrence range of the i-th row as a component of the i-th row and j-th column, or a plurality of them may be included. Even if it is, a value of 1 may be given.

Next, the item graph creating means 3 uses the items in the contingency table created by the contingency table creating means 2 as nodes, generates links between the nodes based on the contingency table, and creates an item graph (step S31).
As a method of creating an item graph from a contingency table, for example, it is conceivable to simply generate a link for items co-occurring on the same line. When this method is applied to the contingency table shown in FIG. 4, for example, in the row x1, items a1, a2, a3, a5, a6, a7, a8, d1 co-occur, so there is a link between them. Will be generated.
In addition, for each set of items, values such as Jaccard coefficient, Dice coefficient, Overlap coefficient, cosine similarity, mutual information amount, etc. are calculated, and a predetermined number of items are linked in order from the set of items with the largest value. It is also conceivable to create an item graph by the method of generating. Alternatively, if the calculated value is equal to or greater than a predetermined threshold value, a method of generating a link for the set of items is also conceivable. The method of creating the item graph is not limited to these methods, and other methods may be used. FIG. 5 shows an example of an item graph created by calculating a Jaccard coefficient for each set of items based on the contingency table shown in FIG. 4 and generating links in a number determined in descending order. In the example shown in FIG. 5, the number of links is determined so as to have an average of 8 links per node.

Next, the co-occurrence range graph creating unit 4 creates a co-occurrence range graph by generating a link between each node based on the contingency table, using the co-occurrence range of the contingency table created by the contingency table creating unit 2 as a node. (Step S41).
It will be obvious that the method for creating the co-occurrence range graph from the contingency table is exactly the same as the method for creating the item graph from the contingency table described in step S31. The method for creating the item graph and the method for creating the co-occurrence range graph need not be the same. FIG. 6 shows an example of the co-occurrence range graph created by calculating the Jaccard coefficient for each set of co-occurrence ranges based on the contingency table shown in FIG. Show. In the example shown in FIG. 6, the number of links is determined so as to have an average of seven links per node.

  Next, the link information providing unit 5 includes items created by the item graph creating unit 3 based on the co-occurrence range graph created by the co-occurrence range graph creating unit 4 and the contingency table created by the contingency table creating unit 2. Information estimated as the meaning of the link is given to the link of the graph (steps S51 to S53). The details are described below.

The link information providing unit 5 analyzes the co-occurrence range graph (step S51). In the present embodiment, as an analysis of the co-occurrence range graph, the centrality index of each node of the co-occurrence range graph is measured.
Here, the centrality index of a certain node is an index indicating the degree to which the node is central in the graph. There are various types of centrality indices such as “medial centrality”, “distance centrality”, “order centrality”, and “Bonachic centrality”, and the calculation method thereof is known in Non-Patent Document 2 and the like.
“Intermediate centrality” of a certain node is an index that is calculated based on the ratio of the two nodes included in the shortest path of the two nodes when two nodes are selected from the nodes other than the node.
“Distance centrality” of a node is an index calculated by calculating the shortest distance from the node to all other nodes except the node, and calculating the total value. This total value may be normalized by the total value of the shortest distances from all nodes to other nodes, or may be normalized by the number of nodes. However, in the case of an isolated node or a directed graph, it is not suitable for use when the shortest distance cannot be defined.
The “degree centrality” of a certain node is an index calculated by the total number of links possessed by the node in the case of an unweighted graph. This value may be normalized by the number of nodes. In the case of a weighted graph, it is calculated by the sum of the values of each link possessed by the node. This value may be normalized by the number of nodes. Further, in the case of a directed graph, “out-order centrality” that is the sum of links from the node to other nodes and “in-order centrality” that is the sum of links from other nodes to the node. Are calculated respectively.
“Bonatic centrality” of a certain node is an index calculated as a component corresponding to the node of the eigenvector corresponding to the first eigenvalue of the adjacency matrix. This unique vector may be normalized to magnitude 1.
In the present embodiment, an example using “medium centrality” and “order centrality” among the above is shown. FIG. 7 shows a graph in which the median centrality and order centrality of each node in the co-occurrence range graph shown in FIG. 6 are calculated, and the median centrality is plotted on the horizontal axis and the order centrality is plotted on the vertical axis.

  Subsequently, for each link of the item graph, the link information providing unit 5 clarifies in which co-occurrence range the items connected by the links co-occur in the contingency table, and associates them with each link (step S52). For example, regarding the link connecting a1 and a2 in the item graph shown in FIG. 5, when looking at the contingency table shown in FIG. 4, the co-occurrence ranges x1, x2, x3, and x4 co-occur, so these co-occurrence ranges Correspond to the link connecting a1 and a2.

  Subsequently, for each co-occurrence range associated with each link of the item in step S52, the link information providing unit 5 sends information to each link of the item graph based on the result of the analysis of the co-occurrence range graph in step S51. (Step S53). In the present embodiment, a method of giving information to the link of the item graph based on the result of measuring the centrality index of each node of the co-occurrence range graph is shown.

  Here, if a node has a high degree centrality, that is, a node having a large number of links, it is a central node in the sense that it represents the entire graph that is referenced by many other nodes. It can be inferred that the link of the item graph to which such a co-occurrence range corresponds is an obvious connection in a sense.

  On the other hand, if the degree centrality is low, i.e., the number of links is small, but the intermediary centrality is high, i.e. the node often takes the shortest path between other nodes, it is a different concept. It can be inferred that the link of the item graph corresponding to the co-occurrence range is a connection including an unexpected possibility. For example, in the co-occurrence range graph shown in FIG. 6, z1 and z2 are such nodes.

  Also, if a node has a low degree centrality, i.e., a small number of links and a low intermediary centrality, i.e. it is less likely to take the shortest path between other nodes, it is an edge in the graph. It is a node that does not have a significant meaning, and it can be inferred that such a co-occurrence range need not be considered.

In general, since there are a plurality of co-occurrence ranges corresponding to links in the item graph, it is conceivable to give information on whether the link is obvious or unexpected in the following processing.
The link information providing means 5 calculates the maximum value of the degree centrality of the co-occurrence range corresponding to the link in the item graph, and if the maximum value is higher than a predetermined threshold, at least the corresponding co-occurrence range Among them, since there is one co-occurrence range whose degree centrality is higher than a predetermined threshold value, it is determined that the link is “trivial connection”.
On the other hand, when the maximum value of the degree centrality of the co-occurrence range corresponding to the link in the item graph is lower than a predetermined threshold, the link information providing unit 5 sets the maximum value of the median centrality of the co-occurrence range. If the maximum value is higher than a certain threshold value, there is at least one co-occurrence range whose intermediary centrality is higher than a certain threshold value among the corresponding co-occurrence ranges. Is determined to be “a connection that includes an unexpected possibility”.
In addition, the link information giving means 5 is, when both the degree centrality of the co-occurrence range corresponding to the link in the item graph and each maximum value of the mediation centrality are lower than a predetermined threshold, Judge that link as an "unimportant connection".

Here, the predetermined threshold value may be a predetermined threshold value, or may be a value such as the degree centrality of all nodes, or an average value or median value of median centrality. FIG. 8 is an item graph shown in FIG. 5 when the order centrality threshold is set to 5.5 and the median centrality threshold is 20 in the co-occurrence range graph shown in FIG. The example of the item graph which highlighted and displayed the link judged as "the connection which contains" thickly is shown. Among the co-occurrence ranges corresponding to the thick and coordinated links in FIG. 8, in the co-occurrence range graph shown in FIG. 6, a co-occurrence such as z1 or z2 having a low degree centrality and a high intermediary centrality. The range is included. In the item graph shown in FIG. 5, it can be read that the cluster composed of a1 to a8 and the cluster composed of b1 to b2 are linked by links via c1, c2, and d1, respectively. Further, in the graph, the connection via d1 is obvious, but the possibility that the connection via c1 and c2 is unexpected is estimated and suggested on the graph.
In the above, information is given depending on whether or not the maximum value of the corresponding co-occurrence range exceeds a predetermined threshold, but in addition to the maximum value, a minimum value, an average value, a median value, a sum, etc. may be used. Conceivable. For example, if an average value or median value is used, if only one of the co-occurrence ranges has high degree centrality and the other co-occurrence ranges have low degree centrality, the average value or median value is determined. This is likely to be smaller than the threshold value. This is because if the method using the maximum value exceeds a predetermined threshold value having degree centrality in one of the co-occurrence ranges, the link of the corresponding item graph is “trivial connection”. Compared to judgment, it can be considered that there is an effect of relaxing the condition.

Next, the effect of this embodiment will be described.
As described above, in the present embodiment, the link information providing unit 5 is based on the co-occurrence range graph created by the co-occurrence range graph creating unit 4 and the contingency table created by the contingency table creating unit 2. Information based on the analysis of the co-occurrence range is given to the link of the item graph created by the item graph creation means 3.
Therefore, according to the present embodiment, by using the analysis result of the graph of the co-occurrence range, it is possible to estimate link information that cannot be understood from the item graph alone.
Further, in the present embodiment, by using the centrality index of the co-occurrence range graph, it is possible to obtain information such as whether the link of the item graph is an obvious connection or an unexpected connection.

[Second Embodiment]
FIG. 9 is a block diagram showing a functional configuration example of the second embodiment of the graph creating apparatus according to the present invention. As shown in FIG. 9, the second embodiment includes data acquisition means 1, item graph creation means 3 ′, co-occurrence range graph acquisition means 4 ′, and link information provision means 5 ′. Specifically, the graph creation device is realized by an information processing device that operates according to a program such as a personal computer.

Differences from the first embodiment will be described.
Specifically, the item graph creating means 3 ′ is realized by a CPU of an information processing apparatus that operates according to a program. The item graph creation unit 3 ′ uses each item included in each co-occurrence range as a node from the data acquired by the data acquisition unit 1, and generates a link between the nodes based on the data acquired by the data acquisition unit 1. , Has the ability to create item graphs.

  Specifically, the co-occurrence range graph acquisition unit 4 ′ is realized by a CPU of an information processing apparatus that operates according to a program. The co-occurrence range graph acquisition unit 4 ′ uses each co-occurrence range included in the data acquired by the data acquisition unit 1 as a node, acquires the relationship between the co-occurrence ranges, and determines each node based on the acquired relationship. It has a function to generate a link between them and obtain a co-occurrence range graph.

  Here, the relationship between each co-occurrence range includes, for example, a hyperlink between web pages and a citation relationship between papers that explicitly mentions from one co-occurrence range to another co-occurrence range. However, it is not limited to these examples.

  Specifically, the link information providing unit 5 ′ is realized by a CPU of an information processing apparatus that operates according to a program. The link information providing unit 5 ′ is a method for generating an item graph created by the item graph creating unit 3 ′ based on the co-occurrence range graph acquired by the co-occurrence range graph acquiring unit 4 ′ and the data acquired by the data acquiring unit 1. It has a function to give information to links.

  Next, the operation of the graph creation apparatus in this embodiment will be described. FIG. 10 is a flowchart illustrating a processing example of the graph creating apparatus. The same steps as those in the first embodiment described in FIG. 2 are denoted by the same step numbers, and description thereof is omitted.

The item graph creation means 3 ′ uses each item included in each co-occurrence range as a node from the data obtained by the data acquisition means 1, generates a link between the nodes based on the obtained data, and creates an item graph (Step S31 ′).
As a method for creating an item graph from data, for example, it is conceivable to simply generate a link for items co-occurring in the same co-occurrence range. When this method is applied to the data shown in FIG. 3, for example, items a1, a2, a3, a5, a6, a7, a8, and d1 co-occur in the co-occurrence range of x1, and therefore a link is established between them. Will be generated. Further, it is obvious that a method for calculating a value such as a Jaccard coefficient for each set of items and generating a link to the set of items using the value can be performed in the same manner as in the first embodiment. I will.

  The co-occurrence range graph acquisition unit 4 ′ uses each co-occurrence range included in the data acquired by the data acquisition unit 1 as a node, acquires a relationship between the co-occurrence ranges, and based on the acquired relationship, A link is generated between the nodes, and a co-occurrence range graph is acquired (step S41 ′).

  For example, when a co-occurrence range is a web page, a hyper link between web pages is acquired as a relationship, and if the hyper link is regarded as a link in the graph as it is, a co-occurrence range graph with the web page as a node is acquired. Can do.

The link information providing unit 5 ′ is a method for generating an item graph created by the item graph creating unit 3 ′ based on the co-occurrence range graph acquired by the co-occurrence range graph acquiring unit 4 ′ and the data acquired by the data acquiring unit 1. Information is given to the link (steps S51, S52 ', S53).
After analyzing the co-occurrence range graph in step S51, the link information providing means 5 ′ clarifies in which data the items connected by the links co-occur for each link of the item graph, and each link (Step S52 ′). For example, regarding the link connecting a1 and a2 in the item graph shown in FIG. 5, the data shown in FIG. 3 co-occurs in the co-occurrence ranges x1, x2, x3, x4. Correspond to the link connecting a1 and a2.

  Next, effects of the second embodiment will be described.

  In the second embodiment, the co-occurrence range graph acquisition unit 4 ′ uses each co-occurrence range included in the data acquired by the data acquisition unit 1 as a node, acquires the relationship between the co-occurrence ranges, A link is generated between the nodes based on the sex, and a co-occurrence range graph is obtained.

  Therefore, unlike the first embodiment, the link in the co-occurrence range graph is an explicit reference from one co-occurrence range to another co-occurrence range, for example, a hyperlink between web pages or a citation relationship of a paper. There is based on the relationship between each co-occurrence range. Usually, hyperlinks, citation relationships, etc. ensure that there is a relationship between two co-occurrence ranges. Therefore, it is expected that the accuracy of giving information to the link of the item graph is increased by utilizing the relevance of these co-occurrence ranges.

[Third Embodiment]
FIG. 11 is a block diagram showing a functional configuration example of the third embodiment of the graph creating apparatus according to the present invention. As shown in FIG. 11, in the third embodiment, the data acquisition means 1, the contingency table creation means 2, the item graph creation means 3, the co-occurrence range graph creation means 4, the link information provision means 5, A graph display control means 6 is included. Specifically, the graph creation device is realized by an information processing device that operates according to a program such as a personal computer.

  Differences from the first embodiment will be described.

Specifically, the graph display control unit 6 is realized by a CPU of an information processing apparatus that operates according to a program. The graph display control means 6 controls either the item graph or the co-occurrence range graph by controlling the display form based on the input from the user and the information given to the link by the link information giving means 5, Or it has a function to display both.
Specifically, the graph display control means 6 controls the display form based on the input from the user and the information given to the link by the link information giving means 5 and the item graph and the co-occurrence range graph. Display data for displaying either one or both is generated. Then, the graph display control unit 6 performs control so that the generated display data is displayed on the display unit.

  Hereinafter, in this embodiment, the expression that the graph display control unit 6 displays an item graph or the like is used. However, as described above, the graph display control unit 6 generates display data and generates the generated display. It is assumed that data is controlled to be displayed on the display unit.

  Moreover, the input from a user is acquired from what was input via input devices, such as a keyboard and a mouse, according to a user's operation, for example.

  Next, the operation of the graph creation apparatus in this embodiment will be described. FIG. 12 is a flowchart illustrating a processing example of the graph creating apparatus. The same steps as those in the first embodiment described in FIG. 2 are denoted by the same step numbers, and description thereof is omitted.

  The graph display control means 6 controls either the item graph or the co-occurrence range graph by controlling the display form based on the input from the user and the information given to the link by the link information giving means 5, Alternatively, both are displayed (step S61).

  The graph display control means 6 displays both an item graph and a co-occurrence range graph as an initial display, for example. If the user inputs an instruction to display only one of the graphs, only one graph is displayed accordingly. For example, a spectral method, a multidimensional scaling method, a spring model method, a cross-entropy method, or the like may be used as a method for appropriately displaying the graph structure, but is not limited thereto.

  As a method for controlling and displaying the display form of the item graph and the co-occurrence range graph based on the information given to the link by the link information giving means 5, for example, as shown in FIG. It is possible to highlight a link that seems to be. In addition, although not shown in the co-occurrence range graph of FIG. 6, highlighting is performed such as coloring nodes such as z1 and z2 that have high median centrality and low degree centrality, and which link is highlighted in the item graph? Such co-occurrence range can be displayed in a co-occurrence range graph in an easy-to-understand manner.

  In the description of the first embodiment, the determination that the mediation centrality of the co-occurrence range corresponding to the link of the item graph is high or the degree centrality is high is determined based on a predetermined threshold. . Instead, for example, it is conceivable to change the link color of the item graph so that it is proportional to the median centrality and the degree centrality of the co-occurrence range. Specifically, when the link color is displayed in RGB, the median centrality value is proportional to the R value, the order centrality value is proportional to the B value, and the G value is 0. Therefore, the co-occurrence range corresponding to the link of the item graph is displayed in a color closer to red as the median centrality is lower and the order centrality is lower, and as the mediation centrality is higher and the order centrality is higher, it is closer to magenta The color is displayed. The lower the median centrality and the higher the order centrality, the closer to blue, the lower the median centrality and the lower the order centrality, the closer to black. In this way, the user may be able to select whether to display the color of the link continuously, or to highlight the link only when it is above a certain threshold as described above.

Next, effects of the third embodiment will be described.
In the third embodiment, the graph display control means 6 controls the display form based on the input from the user and the information given to the link by the link information giving means 5 to control the item graph and the common display. Display either one or both of the range graphs.

  Therefore, not only the item graph but also the co-occurrence range graph is displayed at the same time, so that an effect of helping the user to understand the link of the item graph is expected.

  As described above, according to the present invention, it is possible to estimate link information that cannot be understood only from the item graph by using the analysis result of the graph of the co-occurrence range for the graph.

  In addition, the specific configuration of the present invention is not limited to the above-described embodiment, and changes within a range not departing from the gist of the present invention are included in the present invention.

  For example, if the above embodiment is described in another expression, an information processing apparatus that operates as a graph display system includes an input unit that receives a user's input of data, a selection operation of a range of data to be acquired, and an item graph. A graph generation program including a storage unit for storing and holding a graph generation program for estimating the meaning of the link having the link and storing the information estimated for the link; and a control unit (CPU) operating based on the graph generation program The program uses the control unit to start processing from the input unit as a key, and appropriately obtains data to be graphed from the Internet or a storage medium. , A division that expresses information about whether each item that becomes a node of the graph to be created is included in each co-occurrence range To create a. Thereafter, according to the graph creation program, the control unit creates the item graph by generating a link between the nodes based on the partition table, using the items of the partition table created by the partition table creation means as nodes, and the partition table. A co-occurrence range graph is created by generating links between nodes based on the contingency table. Next, using the item graph, co-occurrence range graph, and contingency table created in this way, the control unit assigns information obtained from the analysis of the co-occurrence range to the link of the item graph and records it in the storage unit. The control unit displays the item graph stored in the storage unit in response to the user's requested operation on the display unit, and assists in understanding the relationship of items such as people, phrases, and websites via the graph I do.

  The present invention can be applied to applications for understanding the relationship between items. For example, by applying the present invention to a free-form questionnaire for a certain product (in this case, the item is a phrase and the co-occurrence range is a sentence of one respondent), a phrase having an unexpected relationship is found. This can be useful for discovering hidden needs of products.

(Appendix 1)
The control unit of the information processing device
Data acquisition means for acquiring data to be graphed;
From the data acquired by the data acquisition means, a contingency table creating means for creating a contingency table expressing information on whether each item is included in each co-occurrence range;
Item graph creation means for creating an item graph by creating a link between each node based on the partition table, with the items of the partition table created by the partition table creation means as nodes,
A co-occurrence range graph creating means for creating a co-occurrence range graph by generating a link between the nodes based on the contingency table, with a co-occurrence range of the contingency table created by the contingency table creating means,
Based on the co-occurrence range graph created by the co-occurrence range graph creating unit and the contingency table created by the contingency table creating unit, information is given to the link of the item graph created by the item graph creating unit Link information giving means,
Graph creation program characterized by functioning as

(Appendix 2)
The control unit of the information processing device
A data acquisition unit including a plurality of items to be nodes and acquiring data to be a graph by generating a desired link between the items;
Each co-occurrence range to which each item included in the data acquired by the data acquisition unit belongs is set as a node, the relevance between each co-occurrence range is acquired, and a link is generated between the nodes based on the relevance A co-occurrence range graph acquisition means for acquiring a co-occurrence range graph;
From the data acquired by the data acquisition means, each item included in each co-occurrence range as a node, generating a link between the nodes based on the data, an item graph creation means for creating an item graph,
Based on the co-occurrence range graph acquired by the co-occurrence range graph acquisition unit and the data acquired by the data acquisition unit, a link for giving information to the link of the item graph generated by the item graph generation unit Information giving means,
Graph creation program characterized by functioning as

(Appendix 3)
The said link information provision means measures the centrality index of the node of a graph about the said co-occurrence range graph, and assign | provides information to the link of an item graph based on the said centrality index, The said graph characterized by the above-mentioned. Creation program.

(Appendix 4)
The link information adding unit extracts, for a predetermined link of an item graph, which data causes an item connected by the link to co-occur, and gives the result as information of the link. Graphing program.

(Appendix 5)
The link information giving means
Calculate the degree centrality of the co-occurrence range corresponding to a given link in the item graph, and if the value is equal to or higher than a given threshold, the link is determined to be “trivial connection”
If the value of the degree centrality of the co-occurrence range corresponding to the link in the item graph is lower than the threshold, the median centrality of the co-occurrence range is calculated, and if the value is equal to or higher than the predetermined threshold , We determine that the link is a "connection with an unexpected possibility"
The graph creation program described above, wherein information indicating whether the link is an obvious connection or an unexpected connection is added.

(Appendix 6)
The control unit
Based on the input from the user and the information given to the link by the link information giving means, the display form is controlled to display one or both of the item graph and the co-occurrence range graph. The above-described graph creation program which functions as a graph display control means.

(Appendix 7)
Get the data for graph creation,
From the acquired data, create and process a contingency table that represents whether each item is included in each co-occurrence range,
The item of the created contingency table is a node, and an item graph is created by generating a link between the nodes based on the contingency table,
The co-occurrence range of the created contingency table is set as a node, and a link is generated between the nodes based on the contingency table to create a co-occurrence range graph.
A method for creating a graph, characterized in that, based on the created co-occurrence range graph and the contingency table, a process for giving information to a link of the created item graph is performed.

(Appendix 8)
It includes a plurality of items that become nodes, and obtains data that becomes a graph by generating a desired link between the items,
From the acquired data, each item included in each co-occurrence range is set as a node, and an item graph is created by generating a link between the nodes based on the data,
Each co-occurrence range to which each item included in the acquired data belongs is set as a node, a relevance between each co-occurrence range is obtained, and a link is generated between the nodes based on the relevance to generate a co-occurrence range Get the graph,
A method for creating a graph, characterized in that, based on the acquired co-occurrence range graph and the data acquired by the data acquisition means, a process of giving information to the link of the generated item graph is performed.

(Appendix 9)
The processing for assigning information to a link measures the centrality index of a node of the graph for the co-occurrence range graph, and assigns information to the link of the item graph based on the centrality index. Chart creation method.

(Appendix 10)
The process of assigning information to the link is characterized in that, for a predetermined link of the item graph, the data that the item connected by the link co-occurs is extracted and the result is given as the information of the link. The graph creation method described above.

(Appendix 11)
The process of adding information to the link
Calculate the degree centrality of the co-occurrence range corresponding to a given link in the item graph, and if the value is equal to or higher than a given threshold, the link is determined to be “trivial connection”
If the value of the degree centrality of the co-occurrence range corresponding to the link in the item graph is lower than the threshold, the median centrality of the co-occurrence range is calculated, and if the value is equal to or higher than the predetermined threshold , We determine that the link is a "connection with an unexpected possibility"
The graph creating method according to the above, wherein information indicating whether the link is an obvious connection or an unexpected connection is given.

(Appendix 12)
Based on the input from the user and the information given to the link by the link information giving means, the display form is controlled to display one or both of the item graph and the co-occurrence range graph. A graph creation method characterized by controlling display of a graph.

DESCRIPTION OF SYMBOLS 1 Data acquisition means 2 Contingency table creation means 3, 3 'Item graph creation means 4 Co-occurrence range graph creation means 4' Co-occurrence range graph acquisition means 5 Link information provision means 6 Graph display control means

Claims (8)

A data acquisition means for accepting each co-occurrence range data, which is a set of co-occurrence range data that co-occurs when the same or related item appears, and that includes an item that becomes a node of a graph to be generated ;
The co-occurrence table for identifying each item included in each co-occurrence range data acquired by the data acquisition means and expressing information indicating whether each item is included in each co-occurrence range A contingency table creating means for generating and processing from the relationship between each item included in the range data and the co-occurrence range data ;
Item graph creation means for generating an item graph by generating a link between each node based on the partition table, with the item of the partition table created by the partition table creation means as a node,
A co-occurrence range graph creating means for generating a co-occurrence range graph by generating a link between the nodes based on the contingency table, with the co-occurrence range of the contingency table created by the contingency table creating means as a node; ,
For each link of the item graph, the co-occurrence range data in which both nodes forming each link co-occur are identified,
The relationship between the data included in each node according to the desired centrality index obtained by analyzing the difference based on the desired centrality index that appears for each node included in the co-occurrence range graph By performing an association process according to the relationship between each link of the item graph obtained by the identification process and the node of the co-occurrence range graph,
Link information giving means for giving information on the co-occurrence range based on the desired centrality index inherent in each co-occurrence range data to the link of the item graph created by the item graph creation means;
A graph creating apparatus comprising: A set of co-occurrence range data that co-occurs when the same or related item appears, and each co-occurrence range data that contains an item that becomes a node of the graph to be generated, a data acquisition means for receiving and relationships between the co-occurrence range data from the reference information between the co-occurrence range are,
With identifying each item of which the data acquisition means is included to the each co-occurrence range data acquired, each cooccurrence scope each item identified belongs to a node, each cooccurrence obtained by identifying the reference information and it generates a link between the nodes based on the relationship between range and co-occurrence range graph acquiring unit that acquires processing co-occurrence range graph,
A process for generating an item graph by generating a link between the nodes based on the co-occurrence range data , using each item included in each co-occurrence range as a node from the co-occurrence range data acquired by the data acquisition unit Item graph creation means to
For each link of the item graph, the co-occurrence range data in which both nodes forming each link co-occur are identified,
The relationship between the data included in each node according to the desired centrality index obtained by analyzing the difference based on the desired centrality index that appears for each node included in the co-occurrence range graph By performing an association process according to the relationship between each link of the item graph obtained by the identification process and the node of the co-occurrence range graph,
Link information giving means for giving information on the co-occurrence range based on the desired centrality index inherent in each co-occurrence range data to the link of the item graph created by the item graph creation means;
A graph creating apparatus comprising: The link information giving means
Calculate the degree centrality of the co-occurrence range corresponding to a given link in the item graph, and if the value is equal to or higher than a given threshold, the link is determined to be “trivial connection”
If the value of the degree centrality of the co-occurrence range corresponding to the link in the item graph is lower than the threshold, the median centrality of the co-occurrence range is calculated, and if the value is equal to or higher than the predetermined threshold , We determine that the link is a "connection with an unexpected possibility"
The graph creating apparatus according to claim 1 , wherein information indicating whether the link is an obvious connection or an unexpected connection is given. Based on the input from the user and the information given to the link by the link information giving means, the display form is controlled to display one or both of the item graph and the co-occurrence range graph. graph generating apparatus according to any one of claims 1 to 3, characterized in that it comprises a graph display control unit. It is a set of co-occurrence range data that co-occurs when the same or related item appears , and acquires each co-occurrence range data that contains the item that becomes the node of the graph to be generated .
A contingency table that identifies each item included in each acquired co-occurrence range data and expresses information about whether or not each item is included in each co-occurrence range is divided into each co-occurrence range data and its co-occurrence range data. Generate and process from the relationship of each item included in the origin data ,
The item of the created contingency table is set as a node, and an item graph is generated by generating a link between the nodes based on the contingency table,
The co-occurrence scope of the contingency table was created as nodes, to generate a link between each node generating process cooccurrence range graph based on the contingency table,
For each link of the item graph, the co-occurrence range data in which both nodes forming each link co-occur are identified, and the desired centrality index that appears for each node included in the co-occurrence range graph is displayed. Each link of the item graph obtained by the identification process is obtained by analyzing the difference based on the analysis result representing the relationship between the data included in each node according to the desired centrality index. And the co-occurrence range based on the desired centrality index inherent in each co-occurrence range data in the link of the created item graph based on the association processing according to the relationship between the node and the node of the co-occurrence range graph A method of creating a graph by an information processing system , which executes a process of recording a result to which information related to information is recorded in a storage unit, and creates a graph reflecting the result . A set of co-occurrence range data that co-occurs when the same or related item appears, and each co-occurrence range data that contains an item that becomes a node of the graph to be generated, And obtain the relationship between each co-occurrence range data from the reference information between the co-occurrence ranges ,
From the acquired co-occurrence range data , each item included in each co-occurrence range is set as a node, and an item graph is generated by generating a link between the nodes based on the co-occurrence range data ,
With identifying each item included the acquired each co-occurrence range data, each cooccurrence scope each item identified belongs to a node, the association between the co-occurrence range obtained by identifying the reference information the generated links between the nodes, and acquisition processing of cooccurrence scope graph based on,
For each link of the item graph, the co-occurrence range data in which both nodes forming each link co-occur are identified, and the desired centrality index that appears for each node included in the co-occurrence range graph is displayed. Each link of the item graph obtained by the identification process is obtained by analyzing the difference based on the analysis result representing the relationship between the data included in each node according to the desired centrality index. And the co-occurrence range based on the desired centrality index inherent in each of the co-occurrence range data in the link of the item graph Give information about
A graph creation method by an information processing system , which executes a process of recording the result in a storage means and creates a graph reflecting the result . Information processing system
A data acquisition means for accepting each co-occurrence range data, which is a set of co-occurrence range data that co-occurs when the same or related item appears, and that includes an item that becomes a node of a graph to be generated ;
The co-occurrence table for identifying each item included in each co-occurrence range data acquired by the data acquisition means and expressing information indicating whether each item is included in each co-occurrence range A contingency table creating means for generating and processing from the relationship between each item included in the range data and the co-occurrence range data ;
Item graph creation means for generating an item graph by generating a link between each node based on the partition table, with the item of the partition table created by the partition table creation means as a node,
A co-occurrence range graph creating means for generating a co-occurrence range graph by generating a link between the nodes based on the contingency table, with the co-occurrence range of the contingency table created by the contingency table creating means as a node; ,
For each link of the item graph, the co-occurrence range data in which both nodes forming each link co-occur are identified,
The relationship between the data included in each node according to the desired centrality index obtained by analyzing the difference based on the desired centrality index that appears for each node included in the co-occurrence range graph By performing an association process according to the relationship between each link of the item graph obtained by the identification process and the node of the co-occurrence range graph,
Link information giving means for giving information on the co-occurrence range based on the desired centrality index inherent in each co-occurrence range data to the link of the item graph created by the item graph creation means,
Graph creation program characterized by functioning as Information processing system
A set of co-occurrence range data that co-occurs when the same or related item appears, and each co-occurrence range data that contains an item that becomes a node of the graph to be generated, a data acquisition means for receiving and relationships between the co-occurrence range data from the reference information between the co-occurrence range are,
With identifying each item of which the data acquisition means is included to the each co-occurrence range data acquired, each cooccurrence scope each item identified belongs to a node, each cooccurrence obtained by identifying the reference information and it generates a link between the nodes based on the relationship between range and co-occurrence range graph acquiring unit that acquires processing co-occurrence range graph,
A process for generating an item graph by generating a link between the nodes based on the co-occurrence range data , using each item included in each co-occurrence range as a node from the co-occurrence range data acquired by the data acquisition unit Item graph creation means to
For each link of the item graph, the co-occurrence range data in which both nodes forming each link co-occur are identified,
The relationship between the data included in each node according to the desired centrality index obtained by analyzing the difference based on the desired centrality index that appears for each node included in the co-occurrence range graph By performing an association process according to the relationship between each link of the item graph obtained by the identification process and the node of the co-occurrence range graph,
Link information giving means for giving information on the co-occurrence range based on the desired centrality index inherent in each co-occurrence range data to the link of the item graph created by the item graph creation means,
Graph creation program characterized by functioning as

Images