Node processing device, node processing method and program

Abstract

A node processing device is used to arrange nodes on a global graph based on a user's desired viewpoint. One embodiment of the present invention synthesizes and extracts feature quantities that meet the user's analysis needs from multiple types of feature quantities assigned to each node of a node set, and the node processing device has: a receiving unit that receives from the user a designation related to the arrangement of a node selected from the node set on an analysis axis envisioned by the user; and a node processing unit that synthesizes and extracts feature quantities based on the received designated arrangement.

Description

Technical Field

[0001] The present invention relates to a node processing device, a node processing method and a program. Background Art

[0002] In the past, a global graph (landscape) has been proposed, which defines the similarity between patent documents based on keywords, classification information, etc., and maps each patent document as a node. In such a global graph, the positional relationship between patent documents is determined by pre-defined feature quantities such as word frequency and commonality in technical fields.

[0003] <Prior Art Literature>

[0004] <Patent Documents>

[0005] Patent Document 1: Japanese Patent Application Publication No. 2009-146213 Summary of the invention

[0006] <Problems to be Solved by the Invention>

[0007] In the present invention, a technique is proposed for configuring nodes on a global graph based on a user's desired viewpoint.

[0008] <Methods used to solve the problem>

[0009] One embodiment of the present invention relates to a node processing device that synthesizes and extracts a feature quantity that meets the analysis needs of a user from a plurality of types of feature quantities assigned to each node of a node set, the node processing device comprising:

[0010] a receiving unit that receives, from a user, a designation regarding the arrangement of nodes selected from the node set on an analysis axis envisioned by the user; and

[0011] A node processing unit synthesizes and extracts feature quantities based on the received specified configuration.

[0012] According to this method, nodes can be arranged on the global graph based on the user's desired viewpoint.

[0013] In one embodiment, the node processing unit may cluster the nodes using the approximation based on the feature quantity.

[0014] According to this embodiment, it is possible to cluster technically related node groups.

[0015] In addition, in one embodiment, the node processing unit may determine the order of the nodes by using the extracted feature quantities, and display the positional relationship of the nodes according to the order.

[0016] In addition, in one embodiment, the node processing unit may select, from the feature quantity of each node or the synthesized feature quantity, a feature quantity with the smallest deviation from the position of the designated node.

[0017] In addition, in one embodiment, the node processing unit may construct and extract the feature quantity from the remaining feature quantities using the position of the designated node as a constraint condition.

[0018] In addition, in one embodiment, the receiving unit may display the nodes on a graph based on a specified feature amount, and receive a designation related to the configuration on the displayed graph.

[0019] According to this embodiment, the user can easily reconfigure nodes on a graph such as a global graph.

[0020] In addition, in one embodiment, the receiving unit may receive the designation related to the configuration through the coordinates of the node's moving target.

[0021] In addition, in one embodiment, the receiving unit may receive a designation related to the configuration through a user operation, wherein the user operation uses an operating device to move a node on the graph selected by the user on the graph.

[0022] Another aspect of the present invention relates to a node processing method, which synthesizes and extracts a feature quantity that meets the analysis requirements of a user from a plurality of types of feature quantities assigned to each node of a node set, the node processing method comprising:

[0023] The processor receives from the user a designated step related to the configuration of nodes selected from the above-mentioned node set on the analysis axis envisioned by the user; and

[0024] The processor performs a step of synthesizing and extracting feature quantities based on the received specified configuration.

[0025] According to this method, nodes can be arranged on the global graph based on the user's desired viewpoint.

[0026] Another aspect of the present invention relates to a program for synthesizing and extracting a feature quantity that meets the analysis needs of a user from a plurality of types of feature quantities assigned to each node of a node set, the program causing a computer to execute the following processing:

[0027] receiving, from a user, a designated process related to the configuration of a node selected from the above-mentioned node set on an analysis axis envisioned by the user; and

[0028] Based on the specified configuration received above, the feature quantities are synthesized and extracted.

[0029] According to this method, nodes can be arranged on the global graph based on the user's desired viewpoint. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a schematic diagram showing node processing according to one embodiment of the present invention.

[0031] Figure 2 This is a block diagram showing the functional structure of a node processing device according to one embodiment of the present invention.

[0032] Figure 3 is a diagram illustrating feature quantization of nodes according to one embodiment of the present invention.

[0033] Figure 4 is a diagram illustrating feature quantization of nodes according to one embodiment of the present invention.

[0034] Figure 5 : is a diagram showing the coordinates of the principal component axes according to one embodiment of the present invention.

[0035] Figure 6 1 is a diagram showing a visualized node arrangement before and after axis selection according to an embodiment of the present invention.

[0036] Figure 7 It is a diagram showing the trend of technology convergence according to one embodiment of the present invention.

[0037] Figure 8 is a flow chart illustrating node processing according to one embodiment of the present invention.

[0038] Fig. 9 is a diagram illustrating a user input operation according to an embodiment of the present invention.

[0039] Fig.10 is a diagram illustrating a generated global graph according to one embodiment of the present invention.

[0040] Fig.11 This is a schematic diagram showing an example of user operation according to one embodiment of the present invention.

[0041] Fig.12 This is a schematic diagram showing an example of user operation according to one embodiment of the present invention.

[0042] Fig.13 This is a block diagram showing the hardware configuration of a node processing device according to one embodiment of the present invention. DETAILED DESCRIPTION

[0043] The following describes the implementation method.

[0044] [Overview of the Invention]

[0045] In the embodiment described below, a node processing device for analyzing the relationship between documents such as patent documents, technical documents, and papers represented as nodes is disclosed.

[0046] like Figure 1 As shown, in the node processing device 100 of the embodiment of the present invention, if a patent document is given as a node, based on the feature quantity such as the International Patent Classification (IPC) representing the technical field and the frequency of occurrence of the keywords appearing in the patent document, each node is arranged in a coordinate space with each feature quantity as an axis and displayed to the user as a global graph. In the embodiment of the attached figure, nodes in each technical field classified into household air conditioners (RA), facility air conditioners (PA), and ventilation are drawn in a three-dimensional space based on the feature quantity.

[0047] If the user operates the relative position relationship of any two or more nodes in the drawn three-dimensional space, the node processing device 100 performs principal component analysis to explore the principal component axis corresponding to the user's operation, and reconfigures the nodes on the selected principal component axis and displays it to the user. For example, as shown in the figure, if the user moves the position of PA in a way that separates PA from RA for the nodes drawn on the two axes, the node processing device 100 performs principal component analysis to explore the principal component axis that conforms to the separation operated by the user, and reconfigures the nodes through the principal component axis that separates PA and RA.

[0048] This allows the user to obtain a global view of the rearrangement based on the selected positional relationship.

[0049] [Node processing device]

[0050] First, refer to Figures 2 to 7 , a node processing device 100 according to an embodiment of the present invention is described. The node processing device 100 according to this embodiment synthesizes and extracts feature quantities that meet the analysis requirements of the user from a plurality of types of feature quantities assigned to each node of the node set. Figure 2 1 is a block diagram showing the functional configuration of a node processing device 100 according to an embodiment of the present invention.

[0051] like Figure 2 As shown, the node processing device 100 includes a receiving unit 110 and a node processing unit 120.

[0052] The receiving unit 110 receives from the user a designation related to the configuration of the node selected from the node set on the analysis axis envisioned by the user. For example, the receiving unit 110 receives a plurality of patent documents as a node set, and sends the received patent documents to the node processing unit 120. Furthermore, the node processing unit 120 plots each patent document on a multidimensional space based on the characteristic amount of each patent document (for example, IPC, the frequency of occurrence of keywords, etc.), and the receiving unit 110 displays the patent document group plotted on the two-dimensional or three-dimensional space selected by the user or on the specified two-dimensional or three-dimensional space as a global graph, and receives the operation on the node configuration performed by the user on the global graph. For example, as described above, for the node set configured on any two axes selected by the user, the receiving unit 110 receives a user operation such as separating any two nodes, and sends the received user operation to the node processing unit 120.

[0053] The node processing unit 120 synthesizes and extracts the feature quantity based on the received designated configuration. Specifically, the node processing unit 120 first obtains patent documents as a node set from the receiving unit 110, obtains patent information of each patent document (for example, the text of the specification and claims, bibliographic items, IPC classification symbols, etc.), and saves the obtained patent information.

[0054] For example, the node processing unit 120 can Figure 3 The IPC of each patent document is stored in a chart format as shown in (a). Figure 3 In the specific example shown in (a), the patent document with patent number "aaa" stored in the graph is classified as "IPC_B". In addition, the patent document with patent number "bbb" stored in the graph is classified as "IPC_A" and "IPC_B", and the patent document with patent number "ccc" is classified as "IPC_A". In this way, the IPC classification can be digitized as a feature value.

[0055] Alternatively, the node processing unit 120 can Figure 3 The IPC is stored in a graph format such as that shown in (b) for each data cluster of the patent document. Specifically, the node processing unit 120 can cluster the nodes using the approximation based on the feature quantity. Figure 3In the specific example shown in (b), patent documents with patent numbers "AAA", "BBB" ... are classified into data cluster 1, and one of the patent documents storing data cluster 1 is classified as "IPC_A", and two are classified as "IPC_B". In addition, patent documents with patent numbers "CCC", "DDD" ... are classified into data cluster 2, and 10 of the patent documents storing data cluster 2 are classified as "IPC_A", 5 are classified as "IPC_B", and 6 are classified as "IPC_C". Similarly, patent documents with patent numbers "XXX", "YYY" ... are classified into data cluster 3, and 5 of the patent documents storing data cluster 3 are classified as "IPC_A", and 5 are classified as "IPC_C". In this way, the total number of IPC classifications can be digitized as a feature value.

[0056] In addition, Figure 4 As shown in (a) of FIG. 1 , the node processing unit 120 can quantify the features of keywords in the description of the manual, etc. according to the importance. The importance of a word can be quantified according to the known method for weighting words, such as TF-IDF (Term Frequency-Inverse Document Frequency). According to TF-IDF, as shown in the figure, the importance of each word such as "room", "data" and "IoT" can be determined according to the description of the manual, etc. For example, when the importance is determined for each patent document, the node processing unit 120 can Figure 4 The feature values ​​of each patent document are stored in a graph format such as that shown in (b) of FIG. It should be noted that, when determining the feature value for each data cluster, the average value of the feature values ​​of the patent documents in the data cluster can be used as the feature value of the data cluster. In addition, the frequency of occurrence (TF) of a word can be used as a feature value instead of or in addition to the importance.

[0057] After the characteristic values ​​of the patent documents are calculated in this way, the node processing unit 120 performs principal component analysis based on the calculated characteristic values ​​to determine a plurality of principal component axes PC. i (i=1, 2, ...). For example, the number of the principal component axes may correspond to the eigenvalues. The node processing unit 120 determines the principal component axes PC i Select two principal component axes and draw the node set in a two-dimensional space based on the selected principal component axes. For example, the node processing unit 120 can perform principal component analysis, such as Figure 5 As shown in FIG. 1 , the coefficients of each patent document with respect to the three principal component axes PC1, PC2, and PC3 are calculated from the feature quantities related to the IPC of each patent document. Furthermore, the node processing unit 120 can select any two of these principal component axes to display Figure 6A set of nodes plotted in two-dimensional space as shown in (a).

[0058] In the display Figure 6 After the global graph as shown in (a), Figure 6 As shown in (b), after receiving the user operation of changing the relative positions of the three specific nodes, the node processing unit 120 reconfigures the nodes in accordance with the received user operation. Specifically, the node processing unit 120 calculates the error with the input value and the overall variance value for all pairs of principal component axes (including rotation, inversion, etc.). For example, when the x-axis is set to PC1 and the y-axis is set to PC2 relative to the input value (coordinate), the node processing unit 120 can calculate the following absolute mean square error as the error.

[0059] (Mathematical formula 1)

[0060]

[0061] Here, node1 and node2 represent the coordinates of node 1 and node 2, respectively. PC1 and PC2 are coefficient vectors derived from principal component analysis. node1 and node1 Respectively represent the x-coordinate and y-coordinate of node 1, x node2 and node2 Respectively represent the x-coordinate and y-coordinate of node 2. Figure 5 In the specific example shown, when mapping the value of node1 to the PC1 axis, the value of each feature value of node1 is multiplied by each coefficient of the PC1 axis. Therefore, the coordinates of node1 on the PC1 axis are determined by node1(IPC_A, IPC_B, IPC_C)*(PC1_A, PC1_B, PC1_C). Mathematical formula 1 is used to derive the coordinates (x node1 ,y node1 ) and the error between the coordinates (node1*PC1,node1*PC2) mapped by principal component analysis.

[0062] The node processing unit 120 defines a cost function based on the error between the input value and the coordinates drawn on the coordinates of the principal component analysis and the overall variance value. Here, the error is calculated only for a part of the nodes specified by the input, but the variance value is calculated for all nodes. Therefore, it is the same as the formula used to derive the variance of general two-dimensional coordinates. Setting the variance value as the cost function in a manner that the overall variance value becomes larger (in a manner that the amount of information drawn becomes larger) is the same as the derivation method of general principal component analysis. For example, the node processing unit 120 can define the cost function in a manner of cost = A*overall variance-B*and the error of the input. Here, A and B are constants. It can be set so that the error between the specified node and the input value is smaller and the overall variance value becomes larger. In addition, the node processing unit 120 determines a pair of principal component axes that maximizes the cost function, and selects the pair of principal component axes as the principal component axes corresponding to the user operation. As Figure 6 As shown in (c), the node processing unit 120 plots the node set in the two-dimensional space defined by the two principal component axes selected in this way, and displays the generated global graph to the user.

[0063] In this way, by generating a global graph corresponding to the user's input operation, Figure 7 It is possible to visualize the integration of the technical field of air conditioning and the technical field of information processing with the development of technology as shown in the figure. Here, the horizontal axis can be considered to be the change of technology, and the vertical axis can be considered to indicate whether it is close to the field of air conditioning or the field of information processing. In this way, when only one axis is used to observe the value, the values ​​are arranged from smaller to larger (in order of years, etc.), and the node processing unit 120 can determine the order of the nodes by extracting the feature quantity, and display the positional relationship of the nodes according to the order.

[0064] [Node processing]

[0065] Next, refer to Figures 8 to 11 , a node processing of an embodiment of the present invention is described. The node processing is implemented by the node processing device 100, for example, it can be implemented by the processor of the node processing device 100 executing a program or command. Figure 8 is a flow chart illustrating node processing according to one embodiment of the present invention.

[0066] like Figure 8 As shown, in step S101, the node processing device 100 obtains patent information of a patent document. For example, the node processing device 100 can obtain the text, bibliographic items, IPC classification symbols, etc. of the specification and claims of the patent document from a database, etc. as patent information for the patent document provided by a user, etc.

[0067] In step S102, the node processing device 100 extracts the feature quantity of the patent document from the patent information. For example, the node processing device 100 can calculate the feature quantity based on the IPC and keywords in the acquired patent information for each data cluster composed of each patent document or multiple patent documents. In addition, the node processing unit 120 can set the position of the specified node as a constraint condition, thereby constructing and extracting the feature quantity from all the feature quantities. That is, the node processing unit 120 can determine the coordinates of all nodes using the determined feature quantity and its coefficient.

[0068] In step S103, the node processing device 100 performs principal component analysis on the patent document to be drawn based on the extracted feature quantity to determine a predetermined number of principal component axes, and draws the patent document in a two-dimensional space based on two principal component axes selected from these principal component axes.

[0069] In step S104, the node processing device 100 displays the global graph of the patent document drawn in the two-dimensional space generated in step S103 to the user. In addition, if the user inputs an operation such as a configuration change operation of the patent document on the global graph, the node processing device 100 retrieves the principal component axis corresponding to the node position after the configuration change. For example, Fig. 9 As shown, the user changes the position of one or both of the RA node and the PA node in a manner that separates the RA node from the PA node, thereby performing a configuration change operation.

[0070] In step S105, the node processing device 100 determines the principal component axis corresponding to the node position after the arrangement change, and re-arranges each node on the determined principal component axis. The node processing device 100 searches for a pair of principal component axes corresponding to the node position after the change.

[0071] In step S106, the node processing device 100 calculates the cost of each pair of principal component axes, and determines the pair of principal component axes that minimizes the cost as the most suitable principal component axis. It should be noted that in addition to principal component analysis, there may also be a method of compressing the dimension of high-dimensional feature quantities and displaying them in two dimensions. In order to achieve the specified drawing, for example, it can be considered that the deviation from the input becomes smaller when 500 feature quantities are reduced to important 100 feature quantities and the principal component analysis is performed. In order to achieve the drawing in the most suitable manner, the feature quantity is selected. In addition, as a result of the principal component analysis, the feature quantity with a coefficient of 0 may not be selected. Therefore, the node processing unit 120 can select the feature quantity with the smallest deviation from the position of the specified node from the feature quantity of each node or the synthesized feature quantity.

[0072] In step S107, the node processing device 100 displays the global graph based on the determined most suitable principal component axis to the user. For example, the node processing device 100 can generate Fig.10 The global picture shown.

[0073] According to the above node processing, for example, when the node processing device 100 displays patent information in the global graph, the patent documents are clustered based on keywords and similarities to produce Fig.11 The patent set A is shown in (a) of FIG. Furthermore, the node processing device 100 adds an identifier representing the characteristics of each patent document and arranges each patent document on the global graph. For example, the global graph is defined by an x-axis and a y-axis representing the frequency of occurrence of any two keywords, and each patent document can be arranged according to its frequency of occurrence.

[0074] In this global graph, each patent document is arranged according to the appearance frequency of two set keywords, but there may be a user who wants to set the axis of the global graph according to the content to be analyzed. For example, by setting it to be possible to designate the distribution that should be obtained when analyzing the display positions of a small number of patent documents in patent set A according to the user's analysis content, the user's desired distribution can be communicated to the node processing device 100. For example, Fig.11 As shown in (b), the user uses coordinates to perform analysis of several patent documents known to the user in patent set A. In the specific example shown in the figure, when a is set to be machine technology, b is set to be utilization technology, and c is set to be the latest machine technology, a is specified as coordinates (1,1), b is specified as coordinates (1,9), and c is specified as coordinates (9,1).

[0075] Respond to the user's specified operation, such as Fig.12 As shown in (a) of FIG. 1 , the node processing device 100 selects the principal component axis to be displayed at a position close to the designated coordinates. In addition, the node processing device 100 may synthesize the principal component axes and select the synthesized axis. Fig.12 As shown in (b), the node processing device 100 draws in a two-dimensional space on the principal component axis of each patent document in the selected patent set A, and displays the generated global graph to the user.

[0076] [Hardware configuration of node processing device]

[0077] The node processing device 100 may have, for example, Fig.13That is, the node processing device 100 includes a drive device 101, an auxiliary storage device 102, a memory device 103, a CPU (Central Processing Unit) 104, an interface device 105, and a communication device 106, which are connected to each other via a bus B.

[0078] Various computer programs including programs for realizing the various functions and processes in the node processing device 100 can be provided by a storage medium 107 such as a CD-ROM (Compact Disk-Read Only Memory). If the storage medium 107 storing the program is placed in the drive device 101, the program is installed from the storage medium 107 to the auxiliary storage device 102 via the drive device 101. However, the program does not necessarily have to be installed from the storage medium 107, and can also be downloaded from any external device via a network or the like. The auxiliary storage device 102 stores the installed program and stores necessary files, data, etc. The memory device 103 reads and stores the program and data from the auxiliary storage device 102 when there is a program startup instruction. The CPU 104, which functions as a processor, executes the various functions and processes of the node processing device 100 according to the program stored in the memory device 103 and various data such as parameters required for executing the program. The interface device 105 is used as a communication interface for connecting to a network or an external device. The communication device 106 executes various communication processes for communicating with an external device.

[0079] However, the node processing device 100 is not limited to the above-mentioned hardware configuration, and may be composed of any other suitable hardware such as circuits.

[0080] Although the embodiments have been described above, it should be understood that various modifications of the embodiments and details can be made without departing from the spirit and scope of the claims.

[0081] This application claims the benefit of priority based on Japanese Patent Application No. 2020-012300 filed on January 29, 2020, and cites the entire contents of No. 2020-012300 in this application.

[0082] Description of Reference Numerals

[0083] 100 Node Processing Device

[0084] 110 Receiving Department

[0085] 120 Node Processing Unit

Claims

1. A node processing device that synthesizes and extracts a feature quantity that meets the analysis needs of a user from a plurality of types of feature quantities assigned to each node of a node set, the node processing device comprising: a receiving unit that receives, from a user, a designation regarding a relative positional relationship of two or more nodes selected from the node set on an analysis axis assumed by the user; and A node processing unit synthesizes and extracts a feature quantity on the analysis axis from a plurality of types of feature quantities assigned to each of the nodes based on the received designation.

2. The node processing device according to claim 1, wherein: The node processing unit clusters the nodes using the approximation based on the feature quantity.

3. The node processing device according to claim 1, wherein: The node processing unit determines the order of the nodes by using the extracted feature quantities, and displays the positional relationship of the nodes according to the order.

4. The node processing device according to any one of claims 1 to 3, wherein: The node processing unit selects a feature quantity having the smallest deviation from a position of a designated node from the feature quantity of each of the nodes or the synthesized feature quantity.

5. The node processing device according to any one of claims 1 to 3, wherein: The node processing unit constructs and extracts the feature quantity using the position of the designated node as a constraint condition.

6. The node processing device according to any one of claims 1 to 3, wherein: The receiving unit displays the node on a graph based on a predetermined feature amount, and receives the designation on the displayed graph.

7. The node processing device according to any one of claims 1 to 3, wherein: The receiving unit receives the designation related to the designation through coordinates of a movement destination of the node.

8. The node processing device according to any one of claims 1 to 3, wherein: The receiving unit receives the designation through a user operation for moving a node on the graph selected by the user using an operating device.

9. A node processing method, which synthesizes and extracts a feature quantity that meets the analysis requirements of a user from a plurality of types of feature quantities assigned to each node of a node set, the node processing method comprising: The processor receives from the user a designation regarding a relative positional relationship of two or more nodes selected from the node set on an analysis axis envisioned by the user; and The processor performs a step of synthesizing and extracting a feature quantity on the analysis axis from a plurality of types of feature quantities assigned to each of the nodes based on the received designation.

10. A program product that synthesizes and extracts a feature quantity that meets the analysis needs of a user from a plurality of types of feature quantities assigned to each node of a node set, the program product causing a computer to execute the following processing: receiving, from the user, designated processing related to a relative positional relationship of two or more nodes selected from the node set on an analysis axis assumed by the user; and Based on the received designation, a process is performed to synthesize and extract the feature quantity on the analysis axis from the plurality of types of feature quantities assigned to each of the nodes.

Images