JP2017219929A - Generation device, generation method and generation program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To generate classification information for properly classifying nodes contained in a graph.SOLUTION: A generation device includes an acquisition unit, a first generation unit and a second generation unit. The acquisition unit is configured to acquire graph information containing: a plurality of nodes respectively corresponding to subjects on a network; and edges connecting between nodes having predetermined correspondence. The first generation unit is configured to generate first classification information for classifying nodes based on connection of the edges between plural nodes in the graph information acquired by the acquisition unit. The second unit is configured to generate second classification information for classifying nodes, based on the first classification information classified by the first generation unit and classification on the predetermined correspondence.SELECTED DRAWING: Figure 4

Description

  The present invention relates to a generation device, a generation method, and a generation program.

  Currently, many real-time information sharing services are used. Also, conventionally, a technology has been provided that performs classification (hereinafter also referred to as “cluster”) based on the presence or absence of edges (connections) between nodes in a graph (social graph) that shows the correlation of subjects (users) on the network. Has been. For example, a technique for extracting a cluster structure based on a node in a graph and a set of nodes connected to the node is provided. In addition, for example, there is provided a technique for classifying based on the number of nodes common to a set of nodes connected to a certain node in the graph and a set of nodes connected to other nodes.

JP2015-156163A

Xiaowei Xu, Nurcan Yuruk, Zhidan Feng, Thomas A. J. Schweiger, "SCAN: A Structural Clustering Algorithm for Networks," SIGKDD’07, August 12-15, 2007, San Jose, CA, US. H. Shiokawa, Y. Fujiwara, and M. Onizuka, "Scan ++: effecient algorithm for finding clusters, hubs and outliers on large-scale graphs," Proceedings of the VLDB Endowment, vol. 8, no. 11, pp. 1178- 1189, 2015. M. E. Newman and M. Girvan, "Finding and evaluating community structure in networks," Physical review E, vol. 69, no. 2, p. 026113, 2004. L. Page, S. Brin, R. Motwani, and T. Winograd, "The pagerank citation ranking: bringing order to the web." 1999.

  However, in the above-described conventional technology, the nodes included in the graph cannot be appropriately classified. For example, it is not always possible to appropriately classify the nodes included in the graph based on the structural relationship between the nodes in the graph. Specifically, the nodes included in the graph can be appropriately classified only by classifying them based on the number of nodes common to the set of nodes connected to a node and the set of nodes connected to other nodes. It can be difficult.

  The present application has been made in view of the above, and an object thereof is to provide a generation device, a generation method, and a generation program that generate classification information for appropriately classifying nodes included in a graph.

  The generation apparatus according to the present application acquires an acquisition unit that acquires graph information including a plurality of nodes corresponding to each of the entities on the network, and an edge that connects nodes having a predetermined correspondence relationship, and is acquired by the acquisition unit A first generation unit that generates first classification information for classifying nodes based on connection of edges between the plurality of nodes in the graph information, and first classification information generated by the first generation unit; A second generation unit configured to generate second classification information for classifying the node based on the classification related to the predetermined correspondence relationship.

  According to one aspect of the embodiment, there is an effect that classification information that appropriately classifies nodes included in the graph can be generated.

FIG. 1 is a diagram illustrating an example of a generation process according to the embodiment. FIG. 2 is a diagram illustrating an example of a generation process according to the embodiment. FIG. 3 is a diagram illustrating an example of the generation process according to the embodiment. FIG. 4 is a diagram illustrating a configuration example of the generation apparatus according to the embodiment. FIG. 5 is a diagram illustrating an example of a communication count information storage unit according to the embodiment. FIG. 6 is a diagram illustrating an example of a communication content information storage unit according to the embodiment. FIG. 7 is a diagram illustrating an example of the topic storage unit according to the embodiment. FIG. 8 is a diagram illustrating an example of the score information storage unit according to the embodiment. FIG. 9 is a flowchart illustrating a generation processing procedure according to the embodiment. FIG. 10 is a flowchart illustrating the processing procedure of the first clustering according to the embodiment. FIG. 11 is a flowchart illustrating the processing procedure of the second clustering according to the embodiment. FIG. 12 is a flowchart illustrating a processing procedure of third clustering according to the embodiment. FIG. 13 is a hardware configuration diagram illustrating an example of a computer that realizes the function of the generation apparatus.

  Hereinafter, a generation apparatus, a generation method, and a mode for executing a generation program (hereinafter referred to as “embodiment”) according to the present application will be described in detail with reference to the drawings. Note that the generation device, the generation method, and the generation program according to the present application are not limited by this embodiment. In the following embodiments, the same portions are denoted by the same reference numerals, and redundant description is omitted.

(Embodiment)
[1-1. Generation process (first clustering)]
1 to 3, information on a graph (social graph) based on information communication between users in a social networking service (hereinafter sometimes referred to as “SNS”) (hereinafter also referred to as “graph information”) is targeted. The case of classifying is shown. The SNS from which the graph information is acquired may be any social networking service, for example, any service such as Twitter (registered trademark) or Facebook (registered trademark). In the following, classifying nodes may be referred to as clustering.

  First, an example of the generation process according to the embodiment will be described with reference to FIG. FIG. 1 is a diagram illustrating an example of a generation process according to the embodiment. In FIG. 1, the generation apparatus 100 (see FIG. 4) classifies nodes (users) included in the graph based on the graph structure using graph information in a predetermined SNS. Specifically, FIG. 1 illustrates a case where the generation apparatus 100 generates first classification information by the first clustering process.

  As illustrated in FIG. 1, the generation device 100 acquires graph information G11 in a predetermined SNS. The graph information G11 is graph information including a plurality of nodes corresponding to the respective subjects (users) on the network and edges that connect nodes having a correspondence relationship regarding information communication on the network. That is, the node may be read as a subject (user) on the network. The correspondence relationship here may be a post from one node to another node in the SNS. Further, the communication count referred to below may be the posting count from one node to another node in the SNS or the posting count between a certain node and another node. That is, the communication count may be read as the posting count.

  For example, in the graph information G11, among the information regarding the number of times of communication as shown in the communication list IN11, nodes to which a numerical value of 1 or more is assigned to an area where each node intersects are connected by an edge. Hereinafter, information related to communication such as information related to the number of times of communication shown in the communication list IN11 may be used as communication information. For example, in the communication list IN11, “5” is assigned to an area where the row of the node N1 and the column of the node N2 intersect. Therefore, in the graph information G11, it is assumed that the user of the node N1 and the user of the node N2 are users who performed communication on the network, and the node N1 and the node N2 are connected by an edge. Note that the generation apparatus 100 may generate the graph information G11 from the communication information shown in the communication list IN11.

  Here, the communication list IN11 will be described. For example, the communication list IN11 corresponds to the communication count information storage unit 121 illustrated in FIG. For example, the node shown in the row in the communication list IN11 corresponds to the node that is the information transmission source. Further, for example, the nodes shown in the columns in the communication list IN11 correspond to the nodes that are the transmission destinations of information. In the example shown in FIG. 1 to FIG. 3, only 25 nodes N1 to N25 are shown for simplicity of explanation, but the number of nodes is large (like the number of SNS users widely used) For example, it may be 1 million users or 10 million users).

  In the communication list IN11, the numerical values of the areas where the nodes N1 to N4 and the like shown in the vertical row intersect with the nodes N1 to N4 shown in the horizontal column are shown in the horizontal column from the node shown in the vertical row. Indicates the number of times information communication was performed to the node. For example, the numerical value “10” in the region where the node N2 in the vertical row and the node N1 in the horizontal column intersect indicates that the user of the node N2 has transmitted information to the user of the node N1 ten times. Further, for example, the numerical value “5” in the area where the node N1 in the vertical row and the node N2 in the horizontal column intersect indicates that the user of the node N1 has transmitted information to the user of the node N2 five times. . As described above, in the example illustrated in FIG. 1, communication information between the nodes is assigned to the communication list IN11 by distinguishing the information transmission source and the transmission destination. The information communication from the transmission source to the transmission destination may have any content depending on the SNS. For example, the information communication from the transmission source to the transmission destination may be transmission of mail from the transmission source to the transmission destination, information posting from the transmission source to the transmission destination timeline, or transmission. Various contents such as retweet from the source to the destination may be used. When the direction of information communication is not considered, the communication list IN11 may be assigned the same numerical value to the intersecting area, or may be assigned only to one area.

  Further, in the example illustrated in FIG. 1, the graph information G11 indicates a case where nodes having a communication count of two or more between two nodes are connected by an edge. The example of FIG. 1 shows a case where two nodes having at least one communication are connected by an edge regardless of which node is the transmission source. Note that the graph information G11 may be graph information in which two nodes are connected by an edge when the number of communications between the two nodes is equal to or greater than a predetermined threshold.

  Here, the generation device 100 generates first classification information G12 that classifies nodes based on the connection of edges between a plurality of nodes in the graph information G11 (step S11). Specifically, the generation device 100 generates the first classification information G12 based on the structural similarity between the nodes. In the example shown in FIG. 1, the graph information G11 is included in the first classification information G12. However, the first classification information G12 is any information as long as it is information indicating the result of the first clustering. There may be. For example, the first classification information G12 may be any information as long as the information indicating the nodes included in each of the first classifications D11 to D15 is included.

  Here, in the first clustering, the generation apparatus 100 may use, for example, a clustering method SCAN (see Non-Patent Document 1). The generation apparatus 100 is not limited to the clustering method SCAN, and any method may be used as long as it is a method for clustering nodes in a graph based on structural similarity. For example, various methods such as a clustering method SCAN ++ (see Non-Patent Document 2) and other clustering methods (see Non-Patent Document 3) may be used as appropriate.

  In the following, the graph information is indicated by “G = (V, E)”. Here, “G” corresponds to the graph information, “V” indicates a node included in the graph information, and “E” indicates an edge included in the graph information. For example, “| V |” indicates the number of nodes included in the graph information, and “| E |” indicates the number of edges included in the graph information. Further, for example, when the clustering method SCAN is used, the generation device 100 calculates the structural similarity between the nodes using the following formula (1).

  In the above equation (1), “u” and “v” indicate nodes for which similarity is calculated, and σ (u, v) indicates the structural similarity between the nodes u and v. | Γ (v) | indicates the number of nodes adjacent to the node v, and is calculated by the following equation (2).

  In the above equation (2), when “v∈V”, the node v adjacent node set includes a node connected to the node v by an edge and the node v itself. That is, “Γ (v)” on the left side of the equation (2) indicates a set of adjacent nodes of the node v.

  The denominator on the right side of the equation (1) corresponds to a value obtained by multiplying the number of adjacent nodes of the node u by the number of adjacent nodes of the node v and taking the root (square root). Further, the numerator on the right side of the above formula (1) corresponds to the number (value) of nodes commonly included between the adjacent node set of the node u and the adjacent node set of the node v.

  Further, according to the above formulas (1) and (2), σ (u, v) indicating the structural similarity between the nodes u and v is “0” when there is no common adjacent node between the nodes u and v. It becomes. Also, σ (u, v) is “1” when the adjacent node of the node u and the adjacent node of the node v are all shared. That is, σ (u, v) takes a value from 0 to 1.

  The generation apparatus 100 extracts the core node in the node in the graph information G11 based on the structural similarity σ between the nodes calculated by the above formula (1). For example, the generation apparatus 100 extracts a core node using the following expressions (3) and (4). Further, the generation device 100 uses the following two parameters “ε” and “μ” when extracting the core node.

(Parameter 1) “ε”: Structural similarity threshold for configuring a cluster (Parameter 2) “μ”: Minimum number of nodes included in the cluster

The left side “N _ε [u]” of the above equation (3) corresponds to a set of nodes whose structural similarity σ between the nodes u is equal to or greater than the threshold “ε”. Hereinafter, “N _ε [u]” may be referred to as an ε neighboring node of the node u. In addition, the left side “| N _ε [u] |” of the above equation (4) indicates the number of ε neighboring nodes of the node u. The right side “μ” of the above equation (4) corresponds to a structural similarity threshold for forming a cluster. That is, the generation device 100 performs the first clustering with a node satisfying the above formula (4) as a core node. For example, the generating apparatus 100 structurally classifies the nodes N1 to N25 in the graph information G11 by a clustering method SCAN (see Non-Patent Document 1).

  Through the processing described above, the generation device 100 includes the first classification D11 that includes nodes N1, N2, and N3, the first classification D12 includes nodes N6, N7, N8, and N9, and the like. Information G12 is generated. For example, the generation apparatus 100 extracts an appropriate core node by appropriately setting the two parameters “ε” and “μ”, and performs first clustering. In FIG. 1, the generation apparatus 100 extracts nodes N4, N12, N17, and the like as hubs (hereinafter also referred to as “hub nodes”). For example, a hub node is a node that connects a plurality of clusters (classifications) in a graph structure, and may be a node that has an influence on surrounding clusters. For example, the node N12 is a hub node that connects the first classification D11 (node N3) and the first classification D13 (node N13). Further, the generation apparatus 100 extracts the nodes N5, N10, N11, N25, and the like as outliers (hereinafter also referred to as “excluded nodes”). For example, an outlier node may be treated as noise.

  The generation apparatus 100 is not limited to the above, and may generate first classification information that structurally clusters (classifies) nodes in the graph by using various methods as appropriate. For example, the generating apparatus 100 may calculate the structural similarity between the nodes using a weight based on the number of communication between the nodes. For example, the generation device 100 may calculate the structural similarity between the nodes using the following equation (5).

In the above equation (5), “u” and “v” indicate nodes for which similarity is calculated, and σ _w (u, v) uses a weight based on the number of communication between the nodes u and v. Shows the structural similarity. ω (u, v) indicates the number of communications between the nodes u and v. For example, in the example shown in FIG. 1, when the node u is the node N1 and the node v is the node N2, ω (N1, N2) is the number of times of communication “5” from the node N1 to the node N2, and from the node N2. The total number of communications “15” may be the total number of communications “10” to the node N1. In addition, by setting the denominator to “ω (u, v) +1” in the above equation (5), the greater the number of communications, the greater the weight. The greater the number of communications, the less the number of communications. The two values are almost approximate. For example, when comparing the number of communication “1000” and “1001”, the two values are almost similar to the case of comparing the number of communication “1” and “2”.

[1-2. Generation process (second clustering)]
Next, an example of the generation process according to the embodiment will be described with reference to FIG. FIG. 2 is a diagram illustrating an example of a generation process according to the embodiment. In FIG. 2, the generation apparatus 100 classifies the first classifications D11 to D15 and the like based on the communication contents using the first classification information G12. Specifically, FIG. 2 illustrates a case where the generation apparatus 100 generates the second classification information by the second clustering process.

  The generation device 100 generates second classification information G13 for classifying nodes based on the first classification information G12 and the classification related to communication contents (step S12). Specifically, the generating apparatus 100 generates second classification information G13 that classifies the first classifications D11 to D15 and the like based on the similarity of topics related to communication contents. In the example shown in FIG. 2, the graph information G11 is included in the second classification information G13. However, the second classification information G13 is any information as long as it is information indicating the result of the second clustering. There may be. For example, the second classification information G13 may be any information as long as information indicating nodes included in each of the second classifications C11, C12, and the like is included.

  For example, the generation apparatus 100 analyzes the content of information in communication between nodes included in each of the first classifications D11 to D15, thereby determining a topic of communication contents between the nodes included in each of the first classifications D11 to D15. It may be estimated (extracted). For example, the generation apparatus 100 extracts communication topics in the first classification D11 based on character information, image information, and the like communicated with the nodes N1, N2, and N3 included in the first classification D11. Also good. Note that the generation apparatus 100 may extract topics from communication contents between nodes included in each of the first classifications D11 to D15 by appropriately using various conventional methods such as topic analysis (analysis). For example, the generation apparatus 100 analyzes the character information communicated between the nodes included in each of the first classifications D11 to D15 using a natural language processing technique such as morphological analysis as appropriate, and uses important keywords from the character information. May be extracted as a topic in the corresponding first classification.

  The generation apparatus 100 may generate the second classification information G13 that classifies the first classifications D11 to D15 and the like based on the cosine similarity regarding the communication contents in the first classifications D11 to D15. For example, the generating apparatus 100 generates second classification information G13 that classifies the first classifications D11 to D15 and the like based on the cosine similarity between topics in each of the first classifications D11 to D15. For example, when the cosine similarity between a certain first classification and another first classification is greater than or equal to a predetermined threshold, the generation apparatus 100 may use the same cluster (second classification) for the certain first classification and the other first classification. It is good.

  When the predetermined SNS is Twitter (registered trademark), the generating apparatus 100 may generate the second classification information G13 based on the similarity of hash tags. For example, the generation apparatus 100 uses the first classifications D11 to D11 based on the similarity of topics related to the most frequent hash tags among the hash tags used in communication between nodes included in the first classifications D11 to D15. The second classification information G13 may be generated by classifying D15 and the like. For example, the generation apparatus 100 uses the first classifications D11 to D11 based on the similarity of topics related to the most frequent hash tags among the hash tags used in communication between nodes included in the first classifications D11 to D15. The second classification information G13 may be generated by classifying D15 and the like.

  Further, for example, the generation apparatus 100 uses a hash tag used in communication between nodes included in each of the first classifications D11 to D15 as a topic, and determines the first classifications D11 to D15 and the like based on the similarity of the distribution. The second classification information G13 may be generated by classification. For example, the generating apparatus 100 determines the first classifications D11 to D15 and the like based on the similarity of the ratio based on the number of hash tags (topics) used in communication between the nodes included in each of the first classifications D11 to D15. The second classification information G13 may be generated by classification. For example, in the example of FIG. 2, in the communication between the nodes included in the first classification D11, the score of topic A is “0.8” and the score of topic C is “0.5”. This indicates that the number of times topic A has been used as a hash tag is greater than the number of times topic C has been used as a hash tag. That is, in the example of FIG. 2, it is indicated that the topic A is a topic reflecting the communication contents between the nodes included in the first classification D11 rather than the topic C. For example, in the case of the example of FIG. 2, in the communication between the nodes included in the first classification D11, since the score of the topic B is “0”, it indicates that the topic B is not used as a hash tag. . For example, the topic A extracted from the communication content information stored in the communication content information storage unit 122 (see FIG. 5) of the generation apparatus 100 may be “singer A”, and is stored in the communication content information storage unit 122. The topic B extracted from the communication content information may be “group B”.

  Here, in the example illustrated in FIG. 2, as illustrated in the topic list IN12, a case is illustrated in which topics in communication between nodes included in the first classifications D11 to D15 are calculated as scores. For example, the topic list IN12 corresponds to the topic storage unit 123 illustrated in FIG. For example, it is assumed that a topic having a higher score indicates that there is more communication regarding the content of information communicated between nodes included in the corresponding first classification. For example, in the topic list IN12, the score of the topic B is the maximum value “1.1” between the nodes included in the cluster D12 (first classification D12), and communication regarding the content of the topic B is performed in the first classification D12. Indicates many. Further, for example, in the topic list IN12, the score of the topic B is the maximum value “0.8” between the nodes included in the cluster D14 (first classification D14), and the first classification D14 relates to the content of the topic B. Indicates that there is a lot of communication. Therefore, the generation device 100 classifies the first class D12 and the first class D14 into the same second class C12.

  Through the above-described processing, the generation device 100 includes the second classification C11 that includes the first classifications D11 and D13, the second classification C12 that includes the first classifications D12, D14, and D15, and the like. Information G13 is generated. Note that the generation apparatus 100 is not limited to the above, and may generate second classification information for clustering (classifying) nodes in the graph based on the content of communication using various methods as appropriate.

[1-3. Generation process (third clustering)]
Next, an example of the generation process according to the embodiment will be described with reference to FIG. FIG. 3 is a diagram illustrating an example of the generation process according to the embodiment. In FIG. 2, the generation apparatus 100 classifies the nodes N1 to N25 and the like using the second classification information G13. Specifically, FIG. 3 illustrates a case where the generation apparatus 100 generates the third classification information by the third clustering process.

  The generation device 100 generates the third classification information G14 for classifying the node based on the score calculated from the second classification information G13 and information (for example, the graph information G11) indicating the edge (connection relationship) between the nodes. (Step S13). For example, in the third clustering, the generation device 100 may use, for example, a technology related to page rank (see Non-Patent Document 3). In the third clustering, the generation apparatus 100 is not limited to the page rank (see Non-Patent Document 3), and may use various methods as appropriate. For example, the generation device 100 calculates the score of each node for each second classification using the following formulas (6) to (8).

The left side “P _{i, j} ” in the above equation (6) corresponds to each element in the matrix P. For example, when the number of nodes in the graph information G11 is “m (= | V |)”, the matrix P may be a matrix of m rows and m columns. That is, the above equation (6) is used to calculate the value of each element in the matrix P (for example, the matrix “P” in the following equation (8)). For example, the left side “P _{i, j} ” in the above equation (6) may be an element of i rows and j columns in the matrix “P” in the following equation (8).

The right side “A _{i, j} ” corresponds to each element in the matrix A. Here, the matrix A indicates the number of times of communication between nodes corresponding to each row and column. For example, the matrix A may be a matrix of m rows and m columns when the number of nodes in the graph information G11 is “m (= | V |)”. For example, each row in the matrix A may correspond to a node serving as a transmission source, and each column in the matrix A may correspond to a node serving as a communication destination.

For example, in the example shown in FIG. 1, when the node N1 corresponds to i row of the matrix A and the node N2 corresponds to j column of the matrix A, the value of “A _{i, j} ” is changed from the node N1 to the node N2. The communication count “5” may be used. Further, for example, in the example shown in FIG. 1, when the node N2 corresponds to i row of the matrix A and the node N1 corresponds to j column of the matrix A, the value of “A _{i, j} ” is changed from the node N2 to the node N2. The number of times of communication with N1 may be “10”.

Also, the denominator “Σ _j A _{i, j} ” in the right side may be a value obtained by summing the number of times of communication from the node corresponding to the i row of the matrix A to another node. In this case, the left side “P _{i, j} ” in the above equation (6) has the node corresponding to the i row of the matrix A as the transmission source with respect to the total number of communication times where the node corresponding to the i row of the matrix A is the transmission source. , J indicates the ratio of the number of communications with the node corresponding to column j as the transmission destination.

“S” in the above equation (7) is a subset of the node set V and indicates a set of seed nodes (users). Further, the left side “s _i ” in the above formula (7) may be an element in the i-th column of an m-dimensional vector when the number of nodes is “m (= | V |)”. In addition, the left side “s _i ” in the equation (7) may be an i-th element of the matrix “s” in the following equation (8). For example, when the node “V _i ” corresponding to “s _i ” is included in the subset S, the value of “s _i ” is “1” divided by the number of subsets S / | S | ". For example, when the node “V _i ” corresponding to “s _i ” is not included in the subset S, the value of “s _i ” is “0”.

For example, the left side “m _(t) ” in the equation (8) corresponds to the score (probability value) of each node at time t. For example, the left side “m _(t) ” in the above formula (8) corresponds to the score (probability value) of each node at time t when a node included in a certain second classification is a seed node.

“Α” corresponds to a probability value of 0 to 1. For example, the generation device 100 repeats the calculation until the above equation (8) converges. For example, the generation device 100 repeats the calculation until “m _(t) = m _(t−1) ” for all times t exceeding a predetermined value.

  In the example of FIG. 3, the generation apparatus 100 calculates the score of each node in each second classification C11, C12, and the like using the above formulas (6) to (8). The score of each node corresponding to each second classification calculated in this way indicates the probability of being located at each node when the nodes in the graph are randomly walked. That is, the higher the score of each node corresponding to each second classification, the higher the probability that the node in the graph is located when the node in the graph is randomly walked. Therefore, for example, the score of each node corresponding to each second category is an index indicating the importance of each user in the second category. For example, the higher the score of each node corresponding to each second category, the more important the user in the second category.

  Here, in the example shown in FIG. 3, as shown in the score list IN13, a case is shown in which the scores of the nodes N1 to N25 and the like in each of the second classifications C11 and C12 are calculated. For example, the score list IN13 corresponds to the score information storage unit 124 illustrated in FIG. The score list IN13 indicates a state in which each node is ranked in order from the highest score in each of the second classifications C11, C12, and the like. Then, the generation apparatus 100 sets a predetermined number of nodes from the higher ranking in each of the second classifications C11 and C12 as nodes that classify into the second classifications C11 and C12. For example, in each of the second classifications C11 and C12, the generation apparatus 100 may be a node that classifies 100 nodes from the highest ranking into the second classifications C11 and C12.

  In the example illustrated in FIG. 3, the generation apparatus 100 sets nodes N13, N2, N12, N1, and the like having higher ranks corresponding to the second classification C11 as nodes that are classified into the second classification C11. Thus, in the third clustering, the node N12 (hub node N12) that was not included in the second classification C11 in the second clustering is classified into the second classification C11.

  In the example illustrated in FIG. 3, the generation device 100 sets nodes N19, N7, N17, N11 and the like having higher ranks corresponding to the second classification C12 as nodes that are classified into the second classification C12. In this way, in the third clustering, the node N17 (hub node N17) and the node N11 (excluded node N11) that were not included in the second classification C12 in the second clustering are classified into the second classification C12. Each node may be included in a plurality of second classifications C11, C12, etc. after the third clustering. For example, in the example illustrated in FIG. 3, the node N4 may be classified into both of the two second classifications C11 and C12.

  As described above, the generation apparatus 100 indicates that the second classification C11 includes the nodes N13, N2, N12, N1, and the like, the second classification C12 includes the nodes N19, N7, N17, and N11, and the like. Three classification information G14 is generated. Accordingly, the generation device 100 can generate classification information that appropriately classifies the nodes included in the graph based on both the connection structure between the nodes in the graph and the content of information communicated between the nodes.

[2. Configuration of the generator
Next, the configuration of the generation apparatus 100 according to the embodiment will be described with reference to FIG. FIG. 4 is a diagram illustrating a configuration example of the generation apparatus according to the embodiment. The generation device 100 is an information processing device that generates first classification information, second classification information, and third classification information. As illustrated in FIG. 4, the generation apparatus 100 includes a communication unit 110, a storage unit 120, and a control unit 130. Note that the generation apparatus 100 may include a display unit that displays various types of information and an input unit that inputs various types of information.

(Communication unit 110)
The communication unit 110 is realized by a NIC or the like, for example. The communication unit 110 is connected to a predetermined network by wire or wireless, and transmits / receives information to / from an external information processing apparatus.

(Storage unit 120)
The storage unit 120 is realized by, for example, a semiconductor memory device such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk. As illustrated in FIG. 4, the storage unit 120 according to the embodiment includes a communication content information storage unit 122, a communication count information storage unit 121, a topic storage unit 123, and a score information storage unit 124.

(Communication frequency information storage unit 121)
The communication frequency information storage unit 121 according to the embodiment stores information related to the communication frequency in a predetermined SNS (also referred to as “communication frequency information”). FIG. 5 is a diagram illustrating an example of a communication count information storage unit according to the embodiment. For example, the communication frequency information storage unit 121 stores the communication frequency between nodes in a predetermined SNS. For example, the communication count information storage unit 121 stores information used to generate the graph information G11 and the like. As illustrated in FIG. 5, the communication count information storage unit 121 stores the communication count between nodes identified by the node ID as the communication count information.

  For example, in the example illustrated in FIG. 5, the numerical value “5” of the area where the row of the node ID “N1” and the column of the node ID “N2” intersect indicates that the user of the node N1 provides information to the user of the node N2. Indicates that it has been sent 5 times. The numerical value “10” in the area where the row of the node ID “N2” and the column of the node ID “N1” intersect indicates that the user of the node N2 has transmitted information to the user of the node N1 10 times. Moreover, node ID respond | corresponds to the code | symbol of each node shown in FIGS. For example, the user identified by the node ID “N1” corresponds to the node N1 illustrated in FIGS. That is, the node identified by the node ID “N1” and the node N1 in FIGS. 1 to 3 indicate the same node, and the same applies to the other nodes illustrated in FIGS. The same applies to the examples shown in FIGS.

  In addition, when the communication frequency information storage unit 121 stores the total communication frequency regardless of the communication direction, the node ID “N2” is the region where the row of the node ID “N1” and the column of the node ID “N2” intersect. Only one of the regions where the row and the column of the node ID “N1” intersect may be stored. Further, in the example illustrated in FIG. 5, the case of storing in the form of a matrix of node IDs and node IDs is illustrated as an example. However, as long as the communication count information can be stored, it may be stored in any manner. . For example, the communication count information storage unit 121 may store, in a list format, node IDs with which the user of each node ID has communicated and the communication count. For example, the number-of-communications information stored in the number-of-communications information storage unit 121 may be stored in the form of a dictionary that stores a list of node IDs of users who have communicated with users of each node ID and the number of times of communication. .

(Communication content information storage unit 122)
The communication content information storage unit 122 according to the embodiment stores information related to communication content in a predetermined SNS (also referred to as “communication content information”). FIG. 6 is a diagram illustrating an example of a communication content information storage unit according to the embodiment. In the example shown in FIG. 6, communication history information between nodes in a predetermined SNS is stored as communication content information. For example, the communication content information stored in the communication content information storage unit 122 is used for topic extraction used in the second clustering. As shown in FIG. 6, the communication content information storage unit 122 includes “communication ID”, “transmission source ID (node ID)”, “transmission destination ID (node ID)”, “date / time”, “content” as transaction information. And the like.

  “Communication ID” indicates identification information for identifying communication between nodes in a predetermined SNS. The “transmission source ID (node ID)” indicates identification information for identifying a subject (user) that is a transmission source in the communication identified by the corresponding communication ID. The “transmission destination ID (node ID)” indicates identification information for identifying a subject (user) that is a transmission destination in the communication identified by the corresponding communication ID. “Date and time” indicates the date and time when the communication identified by the corresponding communication ID was performed. “Content” indicates character information transmitted / received in the communication identified by the corresponding communication ID.

  For example, in the example shown in FIG. 6, the transaction identified by the communication ID “T11” is identified by the user corresponding to the node identified by the transmission source ID (node ID) “N1” by the transmission destination ID “N2”. It is shown that the character information “Live of singer A ...” has been transmitted to the user to be received on the date “2016/5/1 13:05”.

  The communication content information storage unit 122 is not limited to the above, and may have various items depending on the purpose as long as it is an item related to communication between nodes in a predetermined SNS. For example, when there are a plurality of transmission destinations such as simultaneous transmission, a plurality of “transmission destinations (node IDs)” corresponding to one “communication ID” may be provided. Further, the “content” is not limited to the character information transmitted / received in the communication identified by the corresponding communication ID, and may include any information as long as it is information related to communication. For example, the “content” may include image information, moving image information, position information, information on a transmission source user context, information on a transmission destination user context, and the like.

(Topic storage unit 123)
The topic storage unit 123 according to the embodiment stores information related to topics in communication between nodes included in each first classification (cluster) (also referred to as “topic information”). FIG. 7 is a diagram illustrating an example of the topic storage unit according to the embodiment. In the example illustrated in FIG. 7, the topic storage unit 123 stores a score calculated for each topic in communication between nodes included in each of the first classifications D11 to D15 as topic information. As illustrated in FIG. 7, the topic storage unit 123 includes items such as “cluster”, “topic A”, “topic B”, and “topic C” as topic information.

  For example, in the example illustrated in FIG. 7, in the cluster (first classification D11) identified by the cluster “D11”, the score of “topic A” is “0.8”, and the score of “topic B” is “ “0”, indicating that the score of “topic C” is “0.5”. That is, as the contents in the communication between the nodes included in the first classifications D11 to D15, the topic A having the maximum score among the three topics A to C is appropriate.

  For example, the topic A extracted from the communication content information stored in the communication content information storage unit 122 may be “singer A”, and the topic extracted from the communication content information stored in the communication content information storage unit 122. B may be “Group B”. Note that the topic storage unit 123 is not limited to the items described above, and may have various items depending on the purpose as long as the topic information is related to communication between nodes included in each first category.

(Score information storage unit 124)
The score information storage unit 124 according to the embodiment includes information (also referred to as “score information”) relating to the score of each node calculated for each second classification, and information relating to ranking (ranking) based on the score information (“ranking information”). (Also called). FIG. 8 is a diagram illustrating an example of the score information storage unit according to the embodiment. In the example shown in FIG. 8, information obtained by ranking each of the nodes N1 to N25 based on the score is stored for each of the second classifications C11, C12, and the like. As illustrated in FIG. 8, the score information storage unit 124 includes items of “rank”, “C11”, and “C12”.

  “Rank” indicates the rank of each node for each second classification. The items “C11” and “C12” correspond to the second classifications C11 and C12, respectively, and include items such as “node ID” and “score”. Note that the items are not limited to the above, and the score information storage unit 124 may have various items depending on the purpose.

  For example, in the example illustrated in FIG. 8, for the second classification C11, the score of the node N13 is “0.09”, which is the highest, and the ranking of the node N13 is first, that is, the highest ranking. . For example, in the example illustrated in FIG. 8, for the second classification C12, the score of the node N19 is “0.085” which is the maximum, and the ranking of the node N19 is the first.

(Control unit 130)
Returning to the description of FIG. 4, the control unit 130 is, for example, a controller, and various programs (corresponding to an example of a generation program) stored in a storage device inside the generation apparatus 100 by a CPU, an MPU, or the like. Is implemented by using the RAM as a work area. The control unit 130 is, for example, a controller, and is realized by an integrated circuit such as an ASIC or FPGA.

  As shown in FIG. 4, the control unit 130 includes an acquisition unit 131, a first generation unit 132, a second generation unit 133, a third generation unit 134, and a transmission unit 135, which will be described below. Implement or execute information processing functions and actions.

(Acquisition part 131)
The acquisition unit 131 acquires various types of information. For example, the acquisition unit 131 acquires graph information including a plurality of nodes corresponding to each of the subjects on the network and edges that connect nodes having a predetermined correspondence relationship. For example, the acquisition unit 131 acquires graph information including a plurality of nodes corresponding to each of the users on the network and edges that connect nodes having a correspondence relationship regarding information communication on the network. For example, the acquisition unit 131 acquires communication information in a predetermined SNS. For example, the acquisition unit 131 acquires communication count information and communication content information in a predetermined SNS. For example, the acquisition unit 131 may acquire graph information, first classification information, and second classification information.

(First generation unit 132)
The first generation unit 132 generates first classification information. For example, the first generation unit 132 generates first classification information for classifying nodes based on the connection of edges between a plurality of nodes in the graph information acquired by the acquisition unit 131. For example, the first generation unit 132 generates the first classification information based on the number of information communications between nodes. For example, the first generation unit 132 performs processing related to the first clustering illustrated in FIG.

  In the example of FIG. 1, the first generation unit 132 generates first classification information G12 from the graph information G11. For example, the first generation unit 132 generates the first classification information G12 based on the structural similarity between the nodes. For example, the first generation unit 132 may perform processing related to the first clustering using the clustering method SCAN (see Non-Patent Document 1).

  Further, in the example of FIG. 1, the first generation unit 132 extracts the core node in the node in the graph information G11 based on the structural similarity σ between the nodes calculated by the above formula (1). For example, the 1st production | generation part 132 extracts a core node using the said Formula (3) and Formula (4). The first generation unit 132 performs first clustering with a node satisfying the above formula (4) as a core node. For example, the first generation unit 132 structurally classifies the nodes N1 to N25 in the graph information G11 by the clustering method SCAN (see Non-Patent Document 1).

  In the example of FIG. 1, the first generation unit 132 includes the nodes N1, N2, and N3 in the first classification D11, the nodes N6, N7, N8, and N9 in the first classification D12, and the like. First classification information G12 is generated. For example, the first generation unit 132 extracts appropriate core nodes by appropriately setting the two parameters “ε” and “μ”, and performs first clustering.

  In addition, the 1st production | generation part 132 may produce | generate the 1st classification information which clusters (categorizes) the node in a graph structurally not only using the above but using various methods suitably. For example, the first generation unit 132 may calculate the structural similarity between nodes using a weight based on the number of communication between nodes. For example, the first generation unit 132 may calculate the structural similarity between the nodes using the following equation (5).

(Second generation unit 133)
The second generation unit 133 generates second classification information. For example, the second generation unit 133 generates second classification information for classifying nodes based on the first classification information generated by the first generation unit 132 and the classification related to the predetermined correspondence relationship. For example, the second generation unit 133 generates second classification information for classifying the nodes based on the first classification information acquired by the acquisition unit 131 and the classification related to the predetermined correspondence relationship. For example, the second generation unit 133 generates the second classification information based on the content of information communication between nodes. For example, the 2nd generation part 133 generates the 2nd classification information based on the classification about the contents contained in the information communication between nodes. For example, the second generation unit 133 generates second classification information by classifying topics estimated (extracted) from the contents of information communication between nodes. For example, the second generation unit 133 performs processing related to the second clustering illustrated in FIG.

  In the example of FIG. 2, the second generation unit 133 generates second classification information G13 for classifying nodes based on the first classification information G12 and the classification related to communication contents. For example, the second generation unit 133 generates second classification information G13 that classifies the first classifications D11 to D15 and the like based on the similarity of topics related to communication contents.

  For example, the second generation unit 133 analyzes the content of information in communication between the nodes included in each of the first classifications D11 to D15, so that the communication content between the nodes included in each of the first classifications D11 to D15 is analyzed. Topics may be extracted. For example, the second generation unit 133 extracts communication topics in the first classification D11 based on character information, image information, and the like communicated with the nodes N1, N2, and N3 included in the first classification D11. May be. Note that the second generation unit 133 may extract a topic from communication contents between nodes included in each of the first classifications D11 to D15 by appropriately using various conventional methods such as topic analysis (analysis). For example, the second generation unit 133 analyzes the character information communicated between the nodes included in each of the first classifications D11 to D15 by appropriately using natural language processing technology such as morphological analysis, so that important information is obtained from the character information. A keyword may be extracted as a topic in the corresponding first category.

  Moreover, the 2nd production | generation part 133 may produce | generate the 2nd classification information G13 which classify | categorizes 1st classification D11-D15 grade | etc., Based on the cosine similarity regarding the communication content in each 1st classification D11-D15. For example, the second generation unit 133 generates second classification information G13 that classifies the first classifications D11 to D15 and the like based on the cosine similarity between topics in each of the first classifications D11 to D15. For example, when the cosine similarity between a certain first classification and another first classification is equal to or greater than a predetermined threshold, the second generation unit 133 sets the certain first classification and the other first classification to the same cluster (second Classification).

  When the predetermined SNS is Twitter (registered trademark), the second generation unit 133 may generate the second classification information G13 based on the similarity of hash tags. For example, the second generation unit 133 determines the first classification based on the similarity of the topics related to the most frequent hash tag among the hash tags used in the communication between the nodes included in each of the first classifications D11 to D15. The second classification information G13 may be generated by classifying D11 to D15 and the like. For example, the second generation unit 133 determines the first classification based on the similarity of the topics related to the most frequent hash tag among the hash tags used in the communication between the nodes included in each of the first classifications D11 to D15. The second classification information G13 may be generated by classifying D11 to D15 and the like.

  Further, for example, the second generation unit 133 uses the hash tag used in the communication between the nodes included in each of the first classifications D11 to D15 as a topic, and based on the distribution similarity, the first classifications D11 to D15 Etc., the second classification information G13 may be generated. For example, the second generation unit 133 determines the first classifications D11 to D15 based on the similarity of the ratios based on the number of hash tags (topics) used in the communication between the nodes included in each of the first classifications D11 to D15. Etc., the second classification information G13 may be generated.

  In the example of FIG. 2, the second generation unit 133 indicates that the second classification C11 includes the first classifications D11 and D13, the second classification C12 includes the first classifications D12, D14, and D15, and the like. Second classification information G13 is generated. In addition, the 2nd production | generation part 133 may produce | generate the 2nd classification information which clusters (classifies) the node in a graph based on the content of communication not only using the above but using various methods suitably.

(Third generation unit 134)
The third generation unit 134 generates third classification information. For example, the third generation unit 134 generates the third classification information G14 that classifies the node based on the score calculated by the second classification information generated by the second generation unit 133 and the edge between the nodes. . For example, the third generation unit 134 may calculate a score. For example, the third generation unit 134 may generate the third classification information based on the second classification information acquired by the acquisition unit 131. For example, based on the score calculated for each cluster included in the second classification information, the third generation unit 134 generates third classification information in which nodes that satisfy the predetermined condition in each cluster are included in each cluster. For example, the third generation unit 134 generates, based on the score calculated for each cluster included in the second classification information, third classification information in which each cluster includes nodes whose rank based on the score is equal to or higher than a predetermined threshold. To do. Note that the third generation unit 134 is not limited to the above, and may generate the third classification information based on various criteria. For example, the third generation unit 134 may generate third classification information in which nodes whose scores are greater than or equal to a predetermined threshold are included in each cluster. Further, for example, the third generation unit 134 may generate third classification information in which nodes whose scores are equal to or less than a predetermined threshold are included in each cluster. In addition, the third generation unit 134 may generate the third classification information using a plurality of threshold values. For example, the third generation unit 134 may generate third classification information in which nodes that are greater than or equal to the first threshold and less than or equal to the second threshold are included in each cluster. The third generation unit 134 may calculate probability values and likelihoods as scores, and generate third classification information in which nodes in which the calculated score satisfies a predetermined condition are included in each cluster. Thus, the 3rd generation part 134 may compute various information as a score. For example, the 3rd generation part 134 generates the 3rd classification information based on the connection of the edge from the seed node selected from the node. For example, the third generation unit 134 performs processing related to the third clustering illustrated in FIG.

  In the example of FIG. 3, the third generation unit 134 classifies the nodes based on the scores calculated based on the second classification information G13 and information (for example, the graph information G11) indicating the edge (connection relationship) between the nodes. The third classification information to be generated is generated. For example, the third generation unit 134 may use a technique related to page rank (see Non-Patent Document 3) in the third clustering. Note that the third generation unit 134 is not limited to the page rank (see Non-Patent Document 3) in the third clustering, and may use various methods as appropriate. For example, the third generation unit 134 calculates the score of each node for each second classification using the above formulas (6) to (8). For example, the third generation unit 134 calculates the score of each node in each of the second classifications C11, C12, and the like using the above formulas (6) to (8). For example, the third generation unit 134 calculates the score of each node for each second classification by repeating the calculation until the formula (8) converges.

  In the example of FIG. 3, the third generation unit 134 sets nodes N13, N2, N12, N1, and the like having higher ranks corresponding to the second classification C11 as nodes that are classified into the second classification C11. In addition, the third generation unit 134 sets nodes N19, N7, N17, N11 and the like having higher ranks corresponding to the second classification C12 as nodes that are classified into the second classification C12. As described above, in the example of FIG. 3, the third generation unit 134 includes the nodes N13, N2, N12, and N1 in the second classification C11, and the nodes N19, N7, N17, and N11 in the second classification C12. 3rd classification information G14 which shows that is included is generated.

(Transmitter 135)
The transmission unit 135 transmits various types of information to an external device. For example, the transmission unit 135 may transmit the third classification information generated by the third generation unit 134 to an external information processing apparatus.

[3. Generation process flow)
Next, a generation process procedure performed by the generation apparatus 100 according to the embodiment will be described with reference to FIG. FIG. 9 is a flowchart illustrating a generation processing procedure performed by the generation apparatus 100 according to the embodiment.

  As illustrated in FIG. 9, the acquisition unit 131 of the generation apparatus 100 acquires communication information (step S101). For example, the acquisition unit 131 acquires communication information in a predetermined SNS. Thereafter, the first generation unit 132 of the generation device 100 performs a first clustering process (step S102). For example, the first generation unit 132 generates the first classification information by the first clustering process, and details will be described with reference to FIG.

  Thereafter, the second generation unit 133 of the generation apparatus 100 performs the second clustering process (step S103). For example, the second generation unit 133 generates the second classification information by the second clustering process, and details will be described with reference to FIG. Thereafter, the third generation unit 134 of the generation device 100 performs a third clustering process (step S104). For example, the third generation unit 134 generates the third classification information by the third clustering process, and details will be described with reference to FIG.

[3-1. First clustering]
Next, the procedure of the first clustering process performed by the generation apparatus 100 according to the embodiment will be described with reference to FIG. FIG. 10 is a flowchart illustrating the processing procedure of the first clustering according to the embodiment.

  As illustrated in FIG. 10, the acquisition unit 131 of the generation device 100 acquires information regarding the structure of the graph (step S201). For example, the acquisition unit 131 acquires graph information and communication count information. For example, the acquisition unit 131 acquires the communication count information from the communication count information storage unit 121.

  Then, the first generation unit 132 of the generation apparatus 100 classifies the nodes based on the graph structure (step S202). For example, the first generation unit 132 classifies the nodes based on the graph information G11 and the communication count information acquired by the acquisition unit 131.

  Then, the first generation unit 132 generates first classification information (step S203). For example, the first generation unit 132 generates first classification information G12.

[3-2. Second clustering]
Next, the procedure of the second clustering process performed by the generation apparatus 100 according to the embodiment will be described with reference to FIG. FIG. 11 is a flowchart illustrating the processing procedure of the second clustering according to the embodiment.

  As illustrated in FIG. 11, the acquisition unit 131 of the generation device 100 acquires first classification information (step S301). For example, the acquisition unit 131 acquires the first classification information G12 from the first generation unit 132. In addition, the acquisition unit 131 acquires information related to communication contents between nodes (step S302). For example, the acquisition unit 131 acquires communication content information from the communication content information storage unit 122.

  Then, the second generation unit 133 of the generation device 100 classifies the nodes based on the communication contents between the nodes (step S303). The 2nd generation part 133 classifies a node based on the topic for every 1st classification estimated based on the contents of communication between nodes.

  And the 2nd production | generation part 133 produces | generates 2nd classification information (step S304). For example, the second generation unit 133 generates the second classification information G13.

[3-3. Third clustering]
Next, the procedure of the third clustering process performed by the generation apparatus 100 according to the embodiment will be described with reference to FIG. FIG. 12 is a flowchart illustrating a processing procedure of third clustering according to the embodiment.

  As illustrated in FIG. 12, the acquisition unit 131 of the generation device 100 acquires second classification information (step S401). For example, the acquisition unit 131 acquires the second classification information G13 from the second generation unit 133. Further, the acquisition unit 131 acquires information related to the structure of the graph (step S402). For example, the acquisition unit 131 acquires the graph information G11.

  And the 3rd production | generation part 134 of the production | generation apparatus 100 calculates the score of each node for every 2nd classification (step S403). For example, the third generation unit 134 calculates the scores of the nodes N1 to N25 and the like in each of the second classifications C11 and C12.

  Then, the third generation unit 134 generates third classification information (Step S404). For example, the third generation unit 134 generates the third classification information G14.

[4. About communication information)
Note that the generation apparatus 100 may perform the above processing by classifying information to be communicated based on the type of information. For example, the generation device 100 may perform the above-described processing based on various types of weights assigned according to the type of information and various types of communication times.

  For example, the generating apparatus 100 may classify the communication into first type communication and second type communication based on the type of information, and perform the above processing. For example, when the communication is an e-mail, the generating apparatus 100 sets the first type communication when the number of destinations is one (only one user is a destination), and the case where there are a plurality of destinations (for example, simultaneous transmission or the like). ) As the second type communication, the above processing may be performed. For example, when the target SNS is Twitter (registered trademark), the above processing may be performed by classifying replies and retweets into different types of communication. Further, when there are three or more types, the generation apparatus 100 may perform the above processing by classifying the communication into the first type communication to the Nth type communication according to the communication content.

[5. effect〕
As described above, the generation device 100 according to the embodiment includes the acquisition unit 131, the first generation unit 132, and the second generation unit 133. The acquisition unit 131 obtains graph information (“graph information G11” in FIG. 1; the same applies hereinafter) including a plurality of nodes corresponding to each of the subjects on the network and an edge connecting nodes having a predetermined correspondence relationship. get. The first generation unit 132 classifies nodes based on the connection of edges between a plurality of nodes in the graph information acquired by the acquisition unit 131 (“first classification information G12” in FIG. 1; hereinafter the same). ) Is generated. The second generation unit 133 includes second classification information for classifying nodes based on the first classification information generated by the first generation unit 132 and the classification related to the predetermined correspondence (“second classification information in FIG. 2). G13 ". The same applies hereinafter).

  Thereby, the generating apparatus 100 according to the embodiment generates classification information that appropriately classifies the nodes included in the graph based on both the connection structure between the nodes in the graph and the content of the information communicated between the nodes. be able to.

  In addition, the generation device 100 according to the embodiment includes a third generation unit 134. The third generation unit 134 classifies the node based on the score calculated by the second classification information and the edge between the nodes (“third classification information G14” in FIG. 3; the same applies hereinafter). Is generated.

  Thereby, the generating apparatus 100 according to the embodiment calculates the score calculated by the second classification information based on both the connection structure between the nodes in the graph and the content of the information communicated between the nodes, and the edge between the nodes. In addition, by further classifying the nodes, classification information for appropriately classifying the nodes included in the graph can be generated.

  In the generation device 100 according to the embodiment, the third generation unit 134 includes, in each cluster, a node whose score satisfies a predetermined condition based on the score calculated for each cluster included in the second classification information. Third classification information is generated.

  Thereby, the generation apparatus 100 according to the embodiment sets a node having a higher rank (ranking) based on the score of each node for each second category calculated based on the second category as a node included in the second category. By classifying, it is possible to generate classification information for appropriately classifying nodes included in the graph.

  In the generation device 100 according to the embodiment, the third generation unit 134 generates the third classification information based on the connection of the edges from the seed node selected from the nodes.

  Accordingly, the generation apparatus 100 according to the embodiment generates classification information for appropriately classifying the nodes included in the graph by classifying the nodes based on the connection of the edges from the seed nodes selected from the nodes. be able to.

  In the generation device 100 according to the embodiment, the acquisition unit 131 includes a plurality of nodes corresponding to each of the users on the network, and an edge that connects between nodes having a correspondence relationship regarding information communication on the network. Get information.

  Thereby, the generating apparatus 100 according to the embodiment is based on graph information including a plurality of nodes corresponding to each of users on the network and an edge connecting nodes having a correspondence relationship regarding information communication on the network. Classification information for appropriately classifying nodes included in the graph can be generated.

  In the generation device 100 according to the embodiment, the first generation unit 132 generates first classification information based on the structural similarity between nodes.

  Thereby, the production | generation apparatus 100 which concerns on embodiment can produce | generate the classification information which classify | categorizes the node contained in a graph appropriately based on the structural similarity between the nodes in a graph.

  Further, in the generation device 100 according to the embodiment, the first generation unit 132 generates first classification information based on the number of information communications between nodes.

  Thereby, the production | generation apparatus 100 which concerns on embodiment can produce | generate the classification information which classify | categorizes the node contained in a graph appropriately based on the structural information of the graph based on the communication frequency between the nodes in a graph.

  In the generation device 100 according to the embodiment, the second generation unit 133 generates the second classification information based on the classification related to the content included in the information communication between the nodes.

  Thereby, the production | generation apparatus 100 which concerns on embodiment can produce | generate the classification information which classify | categorizes the node contained in a graph appropriately based on the classification regarding the content of the information communicated between the nodes in a graph.

  In the generation device 100 according to the embodiment, the second generation unit 133 generates second classification information by classifying topics estimated from the contents of information communication between nodes.

  Thereby, the generation apparatus 100 according to the embodiment generates classification information that appropriately classifies the nodes included in the graph by classifying the nodes according to the topics estimated from the contents of information communicated between the nodes in the graph. can do.

[6. Hardware configuration)
The generation apparatus 100 according to the above-described embodiment is realized by a computer 1000 having a configuration as illustrated in FIG. FIG. 13 is a hardware configuration diagram illustrating an example of a computer that realizes the function of the generation apparatus. The computer 1000 includes a CPU 1100, RAM 1200, ROM 1300, HDD 1400, communication interface (I / F) 1500, input / output interface (I / F) 1600, and media interface (I / F) 1700.

  The CPU 1100 operates based on a program stored in the ROM 1300 or the HDD 1400 and controls each unit. The ROM 1300 stores a boot program executed by the CPU 1100 when the computer 1000 is started up, a program depending on the hardware of the computer 1000, and the like.

  The HDD 1400 stores programs executed by the CPU 1100, data used by the programs, and the like. The communication interface 1500 receives data from other devices via a predetermined network N and sends the data to the CPU 1100, and transmits data generated by the CPU 1100 to other devices via the predetermined network N.

  The CPU 1100 controls an output device such as a display and a printer and an input device such as a keyboard and a mouse via the input / output interface 1600. The CPU 1100 acquires data from the input device via the input / output interface 1600. In addition, the CPU 1100 outputs the generated data to the output device via the input / output interface 1600.

  The media interface 1700 reads a program or data stored in the recording medium 1800 and provides it to the CPU 1100 via the RAM 1200. The CPU 1100 loads such a program from the recording medium 1800 onto the RAM 1200 via the media interface 1700, and executes the loaded program. The recording medium 1800 is, for example, an optical recording medium such as a DVD (Digital Versatile Disc) or PD (Phase change rewritable disk), a magneto-optical recording medium such as an MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory. Etc.

  For example, when the computer 1000 functions as the generation apparatus 100 according to the embodiment, the CPU 1100 of the computer 1000 implements the function of the control unit 130 by executing a program loaded on the RAM 1200. The CPU 1100 of the computer 1000 reads these programs from the recording medium 1800 and executes them. However, as another example, these programs may be acquired from other devices via the predetermined network N.

  As described above, some of the embodiments of the present application have been described in detail with reference to the drawings. It is possible to implement the present invention in other forms with improvements.

[7. Others]
In addition, among the processes described in the above embodiments, all or a part of the processes described as being automatically performed can be manually performed, or the processes described as being manually performed All or a part of the above can be automatically performed by a known method. In addition, the processing procedures, specific names, and information including various data and parameters shown in the document and drawings can be arbitrarily changed unless otherwise specified. For example, the various types of information illustrated in each drawing is not limited to the illustrated information.

  Further, each component of each illustrated apparatus is functionally conceptual, and does not necessarily need to be physically configured as illustrated. In other words, the specific form of distribution / integration of each device is not limited to that shown in the figure, and all or a part thereof may be functionally or physically distributed or arbitrarily distributed in arbitrary units according to various loads or usage conditions. Can be integrated and configured. For example, the device that generates the first classification information, the device that generates the second classification information, and the device that generates the third classification information may be separate.

  Moreover, each embodiment mentioned above can be combined suitably in the range which does not contradict a process content.

  In addition, the “section (module, unit)” described above can be read as “means” or “circuit”. For example, the acquisition unit can be read as acquisition means or an acquisition circuit.

DESCRIPTION OF SYMBOLS 100 Generating device 121 Communication frequency information storage unit 122 Communication content information storage unit 123 Topic storage unit 124 Score information storage unit 130 Control unit 131 Acquisition unit 132 First generation unit 133 Second generation unit 134 Third generation unit 135 Transmission unit

Claims (11)

An acquisition unit that acquires graph information including a plurality of nodes corresponding to each of the subjects on the network and an edge connecting nodes having a predetermined correspondence;
A first generation unit that generates first classification information for classifying nodes based on connection of edges between the plurality of nodes in the graph information acquired by the acquisition unit;
A second generation unit that generates second classification information for classifying nodes based on the first classification information generated by the first generation unit and the classification related to the predetermined correspondence;
A generating apparatus comprising: A third generator for generating third classification information for classifying nodes based on the score calculated by the second classification information and the edge between the nodes;
The generation device according to claim 1, further comprising: The third generator is
The node according to which the score satisfies a predetermined condition based on the score calculated for each cluster included in the second classification information generates the third classification information included in each cluster. 2. The generating apparatus according to 2. The third generator is
The generation apparatus according to claim 2 or 3, wherein the third classification information is generated based on a connection of edges from a seed node selected from the node. The acquisition unit
The graph information including a plurality of nodes corresponding to each of users on the network and an edge connecting between nodes having a correspondence relationship regarding information communication on the network is acquired. The production | generation apparatus of any one. The first generator is
The generation apparatus according to claim 5, wherein the first classification information is generated based on a structural similarity between the nodes. The first generator is
The generation apparatus according to claim 5 or 6, wherein the first classification information is generated based on the number of times of the information communication between the nodes. The second generator is
The generation apparatus according to any one of claims 5 to 7, wherein the second classification information is generated based on a classification regarding contents included in information communication between the nodes. The second generator is
The generation apparatus according to claim 8, wherein the second classification information is generated by classifying topics estimated from information communication contents between the nodes. A generation method executed by a computer,
An acquisition step of acquiring graph information including a plurality of nodes corresponding to each of the subjects on the network and an edge connecting nodes having a predetermined correspondence;
A first generation step of generating first classification information for classifying nodes based on connection of edges between the plurality of nodes in the graph information acquired by the acquisition step;
A second generation step of generating second classification information for classifying nodes based on the first classification information generated by the first generation step and the classification relating to the predetermined correspondence;
A generation method comprising: An acquisition procedure for acquiring graph information including a plurality of nodes corresponding to each of the subjects on the network and an edge connecting nodes having a predetermined correspondence;
A first generation procedure for generating first classification information for classifying nodes based on connection of edges between the plurality of nodes in the graph information acquired by the acquisition procedure;
A second generation procedure for generating second classification information for classifying nodes based on the first classification information generated by the first generation procedure and the classification relating to the predetermined correspondence relationship;
A program for causing a computer to execute.

Images