JP5569443B2 - Directed graph creation device, directed graph creation method, and directed graph creation program - Google Patents

Description

  The present invention relates to a directed graph creation device, a directed graph creation method, and a directed graph creation program.

  In analyzing trouble information related to accidents, it is important to analyze the process or process until trouble occurs. Information indicating the progress or process is generally described as text information in a report or the like. Therefore, by using the text mining technology, it is possible to extract information indicating a process or process until a trouble occurs from text information as event order information (hereinafter referred to as “event chain”).

  FIG. 1 is a diagram showing an example of event chain extraction from text information. In the figure, an example of an event chain extracted from text information indicating a traffic accident report is shown.

  By obtaining a directed graph (hereinafter referred to as “event flow”) that shows the result of merging event chain information across multiple cases, it is possible to extract trouble scenarios that are common to multiple cases, identify the root cause, etc. Can be done (Figure 1).

  FIG. 2 is a diagram illustrating an example of an event flow. This figure shows an event flow created from 500 reports on traffic accidents. From the event flow shown in the figure, for example, a scenario can be read in which the road surface is in a bad state due to rain, and the vehicle has become seriously injured as a result of slipping and mishandling the steering wheel.

JP 2002-236692 A JP 2004-178270 A

  As a method of creating an event flow by merging event chains, there is a method of reconstructing the event chain by expanding the event chain into an ordered binary relation of events.

  For example, FIG. 3 is a diagram illustrating an example of an event chain relating to three cases. In the same figure, an event chain A related to case A, an event chain B related to case B, and an event chain C related to case C are shown.

  When each event chain in FIG. 3 is expanded into a binary relation, it is as shown in FIG. FIG. 4 is a diagram illustrating an example in which three event chains are expanded into a binary relationship. The figure shows an example in which the event chain A, event chain B, and event chain C are expanded into a binary relationship.

  When the binary relationships shown in FIG. 4 are reorganized based on the coincidence of one event of the binary relationship and the other event of the other binary relationship, an event flow eF1 as shown in FIG. can get.

  FIG. 5 is a diagram illustrating an example of an event flow obtained by reorganizing binary relations related to three event chains. In the figure, event chains that were initially separate are merged to form one event flow eF1 in which all nodes are directly or indirectly connected to each other.

  Although the event flow creation method as described above is simple and efficient, there is a problem that a path (route) that did not exist in the original event chain is created in the event flow. As used herein, a path refers to a direct or indirect connection relationship between nodes.

  FIG. 6 is a diagram showing paths that did not exist in the original event chain. The paths p1 to p4 shown in the figure are included in the event flow eF1 shown in FIG. 5, but are not included in any of the event chains A to C shown in FIG.

  It is also conceivable that a serious trouble occurs when events that occur independently are combined. Therefore, for the purpose of trouble analysis, it may be important to identify “paths that do not exist but are considered as combinations”. However, when the purpose is to analyze what actually happened, it is necessary to be able to separate and analyze the “path that originally existed” and the “path that did not originally exist”.

  Therefore, an object of the present invention is to provide a directed graph creation device, a directed graph creation method, and a directed graph creation program that can remove a path that does not exist in the original event chain from a directed graph in which a plurality of event chains are integrated.

  In order to solve the above problem, the directed graph creation device refers to an event chain storage unit that stores a plurality of event chains, and a directed graph storage unit that stores a directed graph created by integrating the plurality of event chains, Among the routes included in the directed graph, a specifying unit that specifies a route that does not exist in any of the plurality of event chains, and an end point of the route among nodes to which a plurality of edges are connected downstream in the specified route The most downstream node excluding the node and a duplication unit that creates a duplication of the edge connected to the node, and any of the paths included in the plurality of event chains is included in the duplicated edge even if the edge is deleted A deletion unit that deletes edges that are not removed from the directed graph.

  A route that does not exist in the original event chain can be removed from the directed graph in which a plurality of event chains are integrated.

It is a figure which shows the example of extraction of the event chain | link from text information. It is a figure which shows the example of an event flow. It is a figure which shows the event chain regarding three examples. It is a figure which shows the example which expand | deployed three event chains in binary relation. It is a figure which shows the example of the event flow obtained by reorganizing the binary relation regarding three event chains. It is a figure which shows the path | pass which did not exist in the original event chain. It is a figure which shows the hardware structural example of the event flow creation apparatus in embodiment of this invention. It is a figure which shows the function structural example of the event flow creation apparatus in embodiment of this invention. It is a flowchart for demonstrating an example of the process sequence of the creation process of an event flow. It is a figure which shows the 1st example of replication of a node. It is a figure which shows the 1st example of deletion of the redundant edge of the downstream connected to the node which concerns on replication. It is a figure which shows the 1st example of deletion of the redundant edge of the upstream connected to the node which concerns on replication. It is a figure which shows the 2nd example of duplication of a node. It is a figure which shows the 2nd example of a deletion of the downstream redundant edge connected to the node which concerns on replication. It is a figure which shows the 2nd example of deletion of the redundant edge of the upstream connected to the node which concerns on replication. It is a flowchart for demonstrating an example of the process sequence of the specific process of a nonexistence path | pass. It is a figure which shows the example of the adjacency matrix for every event chain | chain, and a reachable matrix. It is a figure which shows the example of the logical sum of all the reachability matrices. It is a figure which shows the example of the logical sum of the adjacency matrix for every event chain | chain. It is a figure which shows the example of the difference with the reachable matrix which shows the event flow created by integration of the reachable matrix which shows an initially existing path | pass, and binary relation. It is a flowchart for demonstrating an example of the process sequence of the selection process of a replication target. It is an example of the data figure for demonstrating the selection process of replication object. It is a flowchart for demonstrating an example of the process sequence of the duplication process of an event. It is a figure which shows the example of duplication of the row | line | column and column corresponding to the event of duplication object. It is a flowchart for demonstrating an example of the process sequence of the deletion process of a redundant edge. It is a figure which shows the example of exclusive deletion of the relationship with the direct downstream event regarding each replicated event. It is a figure which shows the example of the reachable matrix based on the adjacency matrix after deletion of a direct downstream event. It is a figure for demonstrating deletion of the relationship of an upstream side. It is a figure for demonstrating deletion of the relationship of an upstream side.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 7 is a diagram illustrating a hardware configuration example of the event flow creation device according to the embodiment of the present invention. The event flow creation device 10 in FIG. 7 includes a drive device 100, an auxiliary storage device 102, a memory device 103, a CPU 104, a display device 105, an input device 106, and the like that are mutually connected by a bus B.

  A program that realizes processing in the event flow creation apparatus 10 is provided by the recording medium 101. When the recording medium 101 on which the program is recorded is set in the drive device 100, the program is installed from the recording medium 101 to the auxiliary storage device 102 via the drive device 100. However, the program need not be installed from the recording medium 101 and may be downloaded from another computer via a network. The auxiliary storage device 102 stores the installed program and also stores necessary files and data.

  The memory device 103 reads the program from the auxiliary storage device 102 and stores it when there is an instruction to start the program. The CPU 104 realizes functions related to the event flow creation device 10 according to a program stored in the memory device 103. The display device 105 displays a GUI (Graphical User Interface) or the like by a program. The input device 106 includes a keyboard and a mouse, and is used for inputting various operation instructions.

  An example of the recording medium 101 is a portable recording medium such as a CD-ROM, a DVD disk, or a USB memory. An example of the auxiliary storage device 102 is an HDD (Hard Disk Drive) or a flash memory. Both the recording medium 101 and the auxiliary storage device 102 correspond to computer-readable recording media.

  FIG. 8 is a diagram illustrating a functional configuration example of the event flow creation device according to the embodiment of the present invention. In the figure, an event flow creation device 10 includes a document information storage unit 11, an event chain extraction unit 12, an event chain storage unit 21, a development unit 13, a binary relation storage unit 14, an integration unit 15, an event flow storage unit 22, The non-existing path specifying unit 16, the replication target selection unit 17, the node replication unit 18, and the redundant edge deletion unit 19 are included. The document information storage unit 11 can be realized by using, for example, a storage device or a computer connected to the auxiliary storage device 102 or the event flow creation device 10 via a network. The event continuous storage unit 21, the binary relation storage unit 14, and the event flow storage unit 22 use a storage device or a computer connected to the memory device 103, the auxiliary storage device 102, or the event flow creation device 10 via a network. Is feasible. Etc. can be realized. The event chain extraction unit 12, the expansion unit 13, the integration unit 15, the nonexistent path specification unit 16, the replication target selection unit 17, the node replication unit 18, and the redundant edge deletion unit 19 are installed in the event flow creation device 10. Is realized by processing executed by the CPU 104.

  The document information storage unit 11 stores a plurality of document information groups such as accident reports. Note that the document information does not necessarily have to be text format data. It may be document data created by predetermined word processing software or the like. The document information may be data that can finally extract text information.

  The event chain extraction unit 12 extracts order information (hereinafter referred to as “event chain”) of events (hereinafter referred to as “events”) included in the document information for each document information. The event chain extraction method may follow a known technique. The event chain extraction unit 12 records the extracted event chain in the event chain storage unit 21.

  The expansion unit 13 expands each event chain extracted by the event chain extraction unit 12 into a binary relation with an event order. A binary relationship refers to a combination of two events that have a direct relationship with each other in the event chain. A direct relationship refers to a relationship that is adjacent in order information. Also, since the events are ordered, the order relationship between two events having a direct relationship is maintained in a binary relationship.

  The binary relationship storage unit 14 stores information indicating the binary relationship expanded by the expansion unit 13.

  The integration unit 15 integrates (or connects) each binary relationship stored in the binary relationship storage unit 14 based on a match between one event of the binary relationship and the other event of the other binary relationship. To create a directed graph. Each node (node) of the directed graph represents an event. In addition, the edge (branch) of the directed graph indicates an order relationship (or a front-rear relationship or a causal relationship) of events. Hereinafter, the directed graph is referred to as “event flow”. The integration unit 15 records the created event flow in the event flow storage unit 22.

  The non-existing path specifying unit 16 specifies a path that did not exist in each event chain in the event flow. In the event flow, a route (or connection relationship) between arbitrary events is referred to as a “path”. The arbitrary event may be directly connected or indirectly connected. A path that did not exist in the event chain is referred to as a “non-existing path”.

  The replication target selection unit 17 selects a node (event) to be replicated in the event flow created by the integration unit 15. The node duplicating unit 18 creates a duplicate of the node selected by the duplication target selecting unit 17. In other words, node duplication means creating the same node and edge as the existing node and all edges connected to the node. Therefore, by duplicating the node, the node and all edges connected to the node are duplicated. Node replication may be interpreted as a split of nodes. In this case, all edges connected to the node to be divided are also divided.

  Nodes are duplicated (or duplicated) because they are independent from each other in a plurality of event chains. When nodes related to the same event integrate binary relations, This is because the aggregation is considered to be the cause of the occurrence of nonexistent paths. Therefore, it is considered that a non-existing path can be removed if a single node is divided again by duplication. However, if a node is duplicated in the dark clouds, it may return to the original event chain. Accordingly, the replication target selection unit 17 selects a node to be replicated so that a state in which a scenario common to a plurality of event chains can be grasped is maintained.

  The redundant edge deletion unit 19 deletes an edge that does not affect the path included in the original event chain in the event flow even if the edge is deleted among the edges copied along with the replication of the node. “Not affecting the path included in the original event chain” means that any path included in the original event chain is not lost.

  Hereinafter, a processing procedure of the event flow creation apparatus 10 will be described. FIG. 9 is a flowchart for explaining an example of a processing procedure of event flow creation processing.

  For example, in response to an instruction input input via the input device 106, the event chain extraction unit 12 extracts an event chain for each piece of document information stored in the document information storage unit 11, and the extracted event chain is converted into an event chain. It records in the memory | storage part 21 (S10). The extraction of the event chain from the document information may be performed using a known technique such as a text mining technique. For example, an event can be extracted from document information by using a method disclosed in paragraphs 0066 to 0068 of JP-A-2002-236692. Further, the order relationship between events can be specified by using the order in which events are extracted from the document information (appearance position in the document information). At this time, if a specific conjunction is included, the order relationship may be reversed.

  Here, for convenience, it is assumed that event chains A to C shown in FIG. 3 are extracted from three pieces of document information.

  Subsequently, the expansion unit 13 expands each of the event chains A to C recorded in the event chain storage unit 21 into a binary relationship (S20). That is, a binary relation is extracted from each of the event chains A to C. Here, the binary relation shown in FIG. 4 is extracted. The expansion unit 13 records information indicating the extracted binary relationship in the binary relationship storage unit 14. Note that the recording format of the information indicating the binary relation is not limited to a predetermined one. The table format may include a record group including items indicating upstream (cause side) events and items indicating downstream (result side or end side) events. Alternatively, the information may be recorded in a structured document format such as an XML (eXtensible Markup Language) format.

  Subsequently, the integration unit 15 integrates each binary relationship stored in the binary relationship storage unit 14 based on a match between one event of the binary relationship and the other event of the other binary relationship (or By connecting, an event flow is created (S30). The integration unit 15 records the created event flow in the event flow storage unit 22. Here, the event flow eF1 shown in FIG. 5 is created. Note that the creation of the directed graph from the binomial relationship may be performed using, for example, a method disclosed in Japanese Patent Application Laid-Open No. 2004-178270.

  Subsequently, the non-existing path specifying unit 16 refers to the event flow storage unit 22 and the event chain storage unit 21, and in the event flow eF1, does not exist in any of the event chains A, B, or C (non-existence) (Path) is specified (S40). The nonexistent path can be specified by comparing the event flow eF1 with each of the event chains A, B, and C. Here, the paths p1 to p4 shown in FIG. 6 are specified as non-existing paths.

  If there is no nonexistent path (No in S50), the event flow eF1 is output as the final event flow. For example, data indicating the event flow eF1 may be stored in the auxiliary storage device 102 or displayed on the display device 105.

When there is a non-existing path (Yes in S50), the duplication target selecting unit 17 selects any one of the specified non-existing paths and selects the non-existing path (hereinafter, “current non-existing path”). The node (event) to be copied is selected from the above nodes (S60). Here, the selection conditions for the replication target are, for example, as follows.
(1) It is a node sandwiched between the start point node and end point node of the current non-existing path (that is, between the end point nodes) (therefore, the end point nodes such as the start point node and the end point node are selected) Removed from).
(2) In the node group corresponding to (1), the downstream edge is located at the most downstream among the nodes that are branched (a plurality of downstream edges are connected).
(3) When there is no node corresponding to (2), the node group corresponding to (1) should be located most downstream.

  For example, when the path p1 in FIG. 6 is a current non-existing path, the replication target is selected as follows.

  First, in the path p1, the start point node is “straight” and the end point node is “center overline”. That is, the two nodes are not directly or indirectly connected in any of the event chains A to C. Nodes sandwiched between these two nodes are “sudden braking” and “slip” (condition (1)). Two nodes are connected to the downstream side of the “slip”, and the “slip” is located on the most downstream side of the “sudden brake” and the “slip” (condition (2)). Therefore, “slip” is selected as a copy target.

  Subsequently, the node replicating unit 18 creates a replica of the selected node and an edge connected to the node (S70).

  FIG. 10 is a diagram illustrating a first example of replicating a node. In the figure, an example in which the event flow eF2 is created by duplicating the node “slip” is shown. In the event flow eF2, the node indicating the event “slip” is divided into two nodes “slip n1a” and “slip n1b”. In addition, a copy (ed11, ed21, ed31, ed41) of four edges (ed1, ed2, ed3, ed4) connected to the node “slip” is also created. Note that the duplicated edge maintains the same connection relationship as the original edge.

  Subsequently, the redundant edge deletion unit 19 deletes redundant edges from the copied edges from the event flow eF2 (S80). First, with respect to the downstream edge of the node related to replication (including both the replication source and the replication destination), the redundant edge deletion unit 19 classifies the edge into an edge belonging to the current non-existing path and an edge not belonging to the current non-existing path. To do. The redundant edge deletion unit 19 deletes an edge belonging to the current non-existing path from one of the replication source node and the replication destination node, and deletes an edge not belonging to the current non-existing path from the other. This corresponds to deleting two edges that cross each other or two edges that do not cross each other in the figure. That is, one of the overlapping edges is deleted as a redundant edge. The overlapping edge is an edge that connects the same nodes (“slip” and “rear collision”) when the replication source node or the replication destination node is not distinguished, such as the edge ed3 and the edge ed31.

  FIG. 11 is a diagram illustrating a first deletion example of a redundant edge on the downstream side connected to a node related to replication. In the figure, an example is shown in which the event flow eF3 is created by deleting the edges ed31 and ed41 crossing each other in the event flow eF2 (FIG. 10) as redundant edges. As a result, the replication source or replication destination node “slip n1a or n1b” is connected to one edge ed3 or ed4, respectively. However, the edges ed3 and ed4 may be deleted.

  Here, in the event flow eF2 (FIG. 10), the edges ed4 and ed41 belong to the current non-existing path (path p1), and the edges ed3 and ed31 do not belong to the current non-existing path. The result of deleting the edge ed41 belonging to the current non-existing path from the copy source “slip n1a” and deleting the edge ed31 not belonging to the current non-existing path from the copy destination “slip n1b” is the event flow. eF3 (FIG. 11).

  Even when the replication target node is connected to three or more nodes on the downstream side before replication (when the node branches to three or more on the downstream side), it is connected to the downstream side after replication. The edge deletion method is basically as described above. That is, the redundant edge deletion unit 19 classifies the edge into an edge belonging to the current non-existing path and an edge not belonging to the current non-existing path. The redundant edge deletion unit 19 deletes an edge belonging to the current non-existing path from one of the replication source node and the replication destination node, and deletes an edge not belonging to the current non-existing path from the other.

  After deleting the redundant edge on the downstream side of the replication node, the redundant edge on the upstream side is deleted. With respect to the edge connected to the upstream side of the node related to replication, the path included in any of the event chains A to C (hereinafter referred to as “initially existing path”) is deleted from the event flow eF3 by deleting the edge. Only edges that are not lost (not removed) are deleted.

  For example, regarding “overspeed → slip n1a”, even if the edge ed21 constituting the relationship is deleted, any of the initially existing paths is not removed. This is because a path including “overspeed → slip” exists in the event chain C, but the path is maintained by “overspeed → slip n1b”. Therefore, the edge ed21 related to “overspeed → slip n1a” is deleted as a redundant edge. On the other hand, regarding the “sudden brake → slip n1b”, if the edge ed11 related to the relationship is deleted, the initial existence path “sudden brake → center line” is removed. Therefore, the edge ed11 is not deleted.

  FIG. 12 is a diagram illustrating a first deletion example of redundant edges on the upstream side connected to the nodes related to replication. In the figure, an example in which the edge ed21 is deleted and the event flow eF4 is created is shown. When FIG. 6 is compared with FIG. 6, it can be seen that the nonexistent paths p3 and p4 are removed in the event flow eF4.

  Step S40 and subsequent steps are repeated until no nonexistent path is specified (until no nonexistent path disappears). In the present embodiment, there are non-existing paths p1 and p2 (see FIG. 6) for the event flow eF4. Accordingly, step S60 and subsequent steps are executed for the event flow eF4. Here, it is assumed that the path p1 is selected as the current non-existing path. In this case, in the event flow eF4, “sudden braking” matches the selection condition of step S60 (S60). Accordingly, the “rapid brake” and the duplicate of the edge connected to the “rapid brake” are created (S70).

  FIG. 13 is a diagram illustrating a second example of node duplication. In the figure, an example is shown in which the event flow eF5 is created by duplicating the node “sudden brake”. In the event flow eF5, a node indicating an event of “sudden braking” is divided into two nodes of “sudden braking n2a” and “sudden braking n2b”. A copy (ed51, ed61, ed1c, ed11c) of four edges (ed5, ed6, ed1, ed11) connected to the node “slip” is also created.

  Subsequently, the redundant edge is deleted with respect to the event flow eF5 (S80). First, redundant edges are deleted on the downstream side of a node related to replication.

  FIG. 14 is a diagram illustrating a second example of deleting redundant edges on the downstream side connected to the nodes related to replication. In the figure, an example is shown in which the event flow eF6 is created by deleting the edges ed11 and ed1c crossing each other in the event flow eF5 (FIG. 13) as redundant edges. As a result, the replication source or replication destination node “sudden brake n2a or n2b” is connected to one edge ed1 or ed11c, respectively.

  On the other hand, with respect to the edge connected to the upstream side of the node related to duplication, only the edge whose initial path is not removed from the event flow eF6 is deleted by deleting the edge.

  FIG. 15 is a diagram illustrating a second deletion example of the redundant edge on the upstream side connected to the node related to replication. In the figure, an exception is shown in which the edge ed51 and the edge ed61 are deleted and the event flow eF7 is created.

  Here, comparing FIG. 15 with FIG. 6, it can be seen that the nonexistent paths p1 and p2 are further removed in the event flow eF7. Therefore, the event flow eF7 is output as the final event flow.

  As described above, according to the present embodiment, paths that did not exist in the event chain can be removed or eliminated from the event flow rearranged based on the binary relation. In addition, by appropriately selecting the replication target, it is possible to suppress the degree of event flow division. As a result, a scenario common to a plurality of event chains can be shown, and an event flow indicating an order relationship or a causal relationship included in each document information can be created.

  In addition, for example, it is possible to distinguish and analyze a causal relationship included in the document information and a causal relationship not included in the document information.

  By the way, in the processing procedure described above, the data format of data used for determining nonexistent paths, selecting nodes to be copied, and determining relationships to be deleted (redundant edges) is not limited to a predetermined one. . For example, data representing an event flow in the XML format may be used, or another data format may be used.

  In the following, an example in which step S40 and subsequent steps are executed using a matrix will be described. Note that each matrix created in the following description is recorded in the memory device 103, for example.

  Details of step S40 will be described. FIG. 16 is a flowchart for explaining an example of the processing procedure of the non-existing path specifying process.

  Steps S401 to S404 are loop processes executed for each event chain. Here, for convenience, the event chains A, B, and C will be described in parallel. However, the processing may actually be performed in parallel.

  In step S401, the non-existing path specifying unit 16 creates an adjacency matrix for each of the event chains A, B, and C.

  FIG. 17 is a diagram illustrating an example of an adjacency matrix and a reachability matrix for each event chain. In the figure, adjacency matrices aMA, aMB, and aMC are adjacency matrices of event chains A, B, or C in order. That is, in each adjacency matrix, “1” is recorded for a combination of events having a direct connection relationship in the corresponding event chain. The row direction is an upstream (cause side) event, and the column direction is a downstream (result side) event.

  It can be said that the adjacency matrices aMA, aMB, and aMC are matrices indicating the respective binary relations of the event chains A, B, or C. Therefore, the adjacency matrices aMA, aMB, and aMC may be generated based on information stored in the binary relation storage unit 14. Of course, the adjacency matrices aMA, aMB, and aMC may be generated from the event chain A, B, or C.

  Subsequently, the non-existing path specifying unit 16 substitutes the adjacency matrices aMA, aMB, and aMC into the new reachable matrices rMA, rMB, or rMC corresponding to each (S402). Note that the reachable matrix rMA, rMB, or rMC shown in FIG. 17 indicates the state after the execution of steps S403 and S404. Here, three reachable matrices are initialized by the adjacency matrices aMA, aMB, or aMC. The reachable matrix is a matrix in which “1” is recorded in a combination of events having a direct or indirect relationship in the event chain.

  Subsequently, the non-existing path specifying unit 16 multiplies each reachable matrix by the adjacency matrix aMA, aMB, or aMC to obtain a new reachable matrix (S403). The reachability matrix obtained here is a reachability matrix indicating a direct relationship and an indirect relationship with one event interposed therebetween. The multiplication of the adjacency matrix aMA, aMB, or aMC for each reachable matrix is performed until each reachable matrix rMA, rMB, or rMC does not change (S404). That is, each adjacency matrix is raised to the Nth power until the calculation result does not change. The state in which the calculation result does not change is the reachable matrices rMA, rMB, and rMC shown in FIG. Therefore, the reachability matrices rMA, rMB, and rMC are reachability matrices that indicate all direct or indirect relationships of each event in the corresponding event chain.

  When the reachable matrices rMA, rMB, and rMC are obtained (Yes in S405), the non-existing path specifying unit 16 calculates the logical sum of all reachable matrices (rMA, rMB, and rMC) (S406).

  FIG. 18 is a diagram illustrating an example of the logical sum of all reachable matrices. The reachable matrix M1 shown in the figure shows the logical sum of the reachable matrices rMA, rMB, and rMC. The reachable matrix M1 is a matrix indicating all the initial existence paths (including direct or indirect) in the event chain A, B, or C. In other words, a path in which “1” is not recorded in the reachable matrix M1 is a nonexistent path. Note that the reachable matrix M1 is used to identify an initially existing path in subsequent processing.

  Subsequently, the non-existing path specifying unit 16 calculates the logical sum of the adjacency matrices (aMA, aMB, and aMC) for each of the event chains A to C (S407).

  FIG. 19 is a diagram illustrating an example of the logical sum of the adjacency matrix for each event chain. In the figure, an adjacency matrix M2 indicates a logical sum of adjacency matrices aMA, aMB, and aMC.

  Subsequently, the non-existing path specifying unit 16 substitutes the adjacency matrix M2 for a new reachable matrix M3 corresponding to the adjacency matrix M2 (S408). Subsequently, the non-existing path specifying unit 16 multiplies the reachable matrix M3 by the adjacency matrix M2 to obtain a new reachable matrix (S409). The multiplication is performed until the reachable matrix M3 does not change (S410). That is, the adjacency matrix M2 is raised to the Nth power until the calculation result does not change. A state where the calculation result does not change is the reachable matrix M3 shown in FIG. The reachability matrix M3 is a reachability matrix that indicates a direct or indirect connection relationship of each event in the event flow eF1 created by integration of the binary relationship.

  Proceeding to step S411, the non-existing path identifying unit 16 extracts a difference between the reachable matrix M1 (FIG. 17) and the reachable matrix M3 (FIG. 19). That is, the difference between the reachable matrix M1 indicating the initial existence path and the reachable matrix indicating the event flow eF1 is extracted.

  FIG. 20 is a diagram illustrating an example of a difference between the reachability matrix indicating the initial existence path and the reachability matrix indicating the event flow created by the integration of the binary relationship.

  In the figure, in the reachable matrix M3, the cell related to the difference from the reachable matrix M1 is surrounded by a thick frame. This difference indicates the paths p1 to p4 shown in FIG. That is, the set of “straight ahead” and “center line over” indicates the path p1. A set of “straight ahead” and “frontal collision” indicates a path p2. A set of “curve” and “rejection” indicates a path p3. A set of “overspeed” and “rear collision” indicates a path p4.

  In this way, the non-existing path specifying unit 16 can specify the non-existing path.

  Next, details of step S60 in FIG. 9 will be described. FIG. 21 is a flowchart for explaining an example of the processing procedure of the copy target selection process.

  In step S <b> 601, the replication target selection unit 17 selects one non-existing path from the non-existing paths specified by the non-existing path specifying unit 16. Here, the selected nonexistent path is the “current nonexistent path”.

  FIG. 22 is an example of a data diagram for explaining a copy target selection process. In the figure, step numbers indicating the correspondence with FIG. 16 are shown. In step S601 of FIG. 22, an example is shown in which a path related to “straight forward → center line over” is selected in the reachable matrix M3.

  Subsequently, the replication target selection unit 17 selects an event downstream of the event at the start point of the current non-existing path (hereinafter referred to as “downstream event”) and the current non-existing path among the events included in the current non-existing path. Event upstream of the event (hereinafter referred to as “upstream event”) is acquired (S602).

  An event corresponding to an event downstream from the event at the start point of the current non-existing path is an event related to a column surrounded by a thick frame in the reachability matrix M3a corresponding to step S602 in FIG. In other words, in the row related to the current non-existing path, among the columns in which “1” is recorded, the event is in the column on the right side of the start point event “straight ahead”.

  Further, an event corresponding to an event upstream from the event at the end point of the current non-existing path is an event related to a line surrounded by a thick frame in the reachability matrix M3b corresponding to step S602 in FIG. That is, in the column related to the current non-existing path, the event is on the upper side of the end event “center line over” in the row in which “1” is recorded.

  Subsequently, the duplication target selection unit 17 determines the logic of downstream events (rapid braking, slip, center line over, frontal collision, rear-end collision) and upstream events (road surface freezing, straight ahead, curve, sudden braking, overspeed, slip). The product is calculated (S603). By the logical product, an event group sandwiched between the start event and the end event in the current non-existing path is specified. Here, “sudden braking” and “slip” are specified.

  Subsequently, the replication target selection unit 17 acquires one of the unprocessed events in order from the downstream in the identified event group (S604). “Unprocessed” means that it is not set as a processing target after step S606. Since the order is from the downstream, first, the event at the most downstream in the event group is acquired. Hereinafter, the acquired event is referred to as “target event”.

  When the target event is acquired (Yes in S605), the replication target selection unit 17 uses the adjacency matrix M2 to directly connect to the downstream side of the target event (hereinafter referred to as “direct downstream event”). ) Is specified (S606).

  In step S606 of FIG. 22, the identification result of each direct downstream event of “sudden braking” and “slip” is shown in the adjacency matrix M2. As described in FIG. 19, the adjacency matrix M2 is a logical sum of the adjacency matrices aMA, aMB, and aMC. Therefore, the adjacency matrix M2 shows only the direct relationship between events. Therefore, in the adjacency matrix M2, the event related to the column in which “1” is recorded in the “slip” or “sudden brake” row is a direct downstream event of “sudden brake” or “slip”. In the figure, one direct downstream event for “rapid braking” and two direct downstream events for “slip” are specified. In step S606 in FIG. 21, a downstream event is directly specified for one target event. However, in FIG. 22, for convenience, all events sandwiched between the start point and end point of the current non-existing path are directly specified. Downstream events are shown.

  Subsequently, the replication target selection unit 17 determines whether or not the number of direct downstream events related to the target event is plural (S607). This corresponds to determination of whether the downstream edge of the target event is branched or whether there are a plurality of such edges. When the number of direct downstream events related to the target event is plural (Yes in S607), the replication target selection unit 17 selects the target event as a replication target (S608). If the target event is “slip”, the number of direct downstream events is plural. Therefore, “slip” is selected as a copy target. Note that step S608 corresponds to the selection condition (2) described in step S60 of FIG.

  On the other hand, when the number of direct downstream events is not plural (No in S607), the duplication target selection unit 17 repeats Step S604 and the subsequent steps. If the replication target is not selected even after processing all of the event groups identified in step S603 (No in S605), the replication target selection unit 17 selects the event at the most downstream in the event group as the replication target. Is selected (S609). Note that step S609 corresponds to the selection condition (3) described in step S60 of FIG.

  In this way, the replication target selection unit 17 can select a replication target.

  Next, details of step S70 in FIG. 9 will be described. FIG. 23 is a flowchart for explaining an example of a processing procedure of event duplication processing.

  In step S701, the node replicating unit 18 generates a row and column replica corresponding to the event to be replicated in the adjacency matrix M2 (FIG. 19). That is, the row and column are duplicated.

  FIG. 24 is a diagram illustrating an example of row and column duplication corresponding to an event to be duplicated. In the drawing, an example is shown in which the adjacency matrix M4 is created by generating the respective duplicates of the rows and columns related to the “slip” selected as the duplication target in the adjacency matrix M2. In the adjacency matrix M4, “slip a” and “slip b” correspond to “slip n1a” and “slip n1b” in FIG. 10 and the like, respectively.

  Next, details of step S80 in FIG. 9 will be described. FIG. 25 is a flowchart for explaining an example of a processing procedure of redundant edge deletion processing.

  In step S801, the redundant edge deletion unit 19 exclusively deletes the relationship between each event related to replication and the downstream event directly in the adjacency matrix M4 (FIG. 24). Each event related to duplication here includes both a duplication source and a duplication destination. Exclusive means that the relationship that is deleted in the relationship between the replication source event and the direct downstream event does not overlap with the relationship that is deleted in the relationship between the replication destination event and the direct downstream event. .

  For example, FIG. 26 is a diagram illustrating an example of exclusive deletion of a relationship with a directly downstream event regarding each replicated event. In the adjacency matrix M4 shown in the figure, slips a and b correspond to each duplicated event.

  As is clear from the adjacency matrix M4, the directly downstream events of the slips a and b are “center line over” and “rear collision”. Here, the form in which the relationship with “center line over” with respect to slip a is deleted and the relationship with “rejection” with respect to slip b is deleted corresponds to the exclusive deletion described above. In the figure, the adjacency matrix M5 is generated as a result of the exclusive deletion. It should be noted that the relationship with “rear collision” with respect to slip a may be deleted, and the relationship with “center line over” with respect to slip b may be deleted.

  That is, exclusive deletion here corresponds to deleting an edge belonging to the current non-existing path from one of the replication source node and the replication destination node, and deleting an edge not belonging to the current non-existing path from the other. To do.

  If there are three or more direct downstream events for slips a and b, the redundant edge deletion unit 19 classifies the direct downstream events into groups of events that are on the current non-existing path and events that are not. . The redundant edge deleting unit 19 sets the value of the cell related to the combination of each event and slip a belonging to one group to “0”, and sets the value of the cell related to the combination of each event belonging to the other group and slip b. “0”.

  Subsequently, the redundant edge deleting unit 19 substitutes the adjacency matrix M5 into a new reachable matrix M6 corresponding to the adjacency matrix M5 (S802). Subsequently, the redundant edge deleting unit 19 multiplies the reachable matrix M6 by the adjacency matrix M5 to obtain a new reachable matrix (S803). The multiplication is performed until the reachable matrix M6 does not change (S804). That is, the adjacency matrix M5 is raised to the Nth power until the calculation result does not change. As a result, the reachable matrix M6 shown in FIG. 27 is obtained.

  FIG. 27 is a diagram illustrating an example of a reachable matrix based on the adjacency matrix after direct downstream event deletion. In the figure, an example in which the adjacency matrix M5 is raised to the Nth power and the reachable matrix M6 is generated is moistened.

  In step S805, the redundant edge deleting unit 19 sets one of the upstream direct relationships (edges) of each event related to replication (including both the replication source and the replication destination) as a processing target. Select (S805).

  Subsequently, the redundant edge deleting unit 19 deletes the selected relationship (hereinafter referred to as “selection relationship”) from the adjacency matrix M5, and recalculates the reachability matrix based on the deleted adjacency matrix (S806). Subsequently, the redundant edge deleting unit 19 compares the recalculated reachable matrix with the reachable matrix M1 (FIG. 18) indicating only the initially existing path, and deletes the selected relationship, thereby deleting the initially existing path. It is determined whether or not has been removed (S807).

  When the initially existing path is not removed (No in S807), the redundant edge deleting unit 19 sets the selection relationship as a deletion target (S808). On the other hand, when the initially existing path is removed (Yes in S807), the selection relationship is not a deletion target, and steps S805 and after are repeated.

  Steps S805 to S807 described above will be described with reference to the drawings. 28 and 29 are diagrams for explaining the deletion of the upstream relationship.

  In FIG. 28, the adjacency matrix M5a is the result of deleting the relationship relating to “sudden braking → slip a” from the adjacency matrix M5. The reachable matrix M6a is a reachable matrix recalculated based on the adjacency matrix M5a. Comparing the reachable matrix M6a and the reachable matrix M1 (FIG. 18), the reachable matrix M6a has a relationship relating to “straight forward → rear collision” or “sudden braking → rear collision” existing in the reachable matrix M1 ( Path) does not exist. That is, the deletion of the relationship (edge) related to “sudden braking → slip a” leads to the removal of the initially existing path. Accordingly, the relationship relating to “sudden braking → slip a” is not a deletion target. The reachable matrix M6a does not have the relationship of “straight forward → slip a” and “sudden brake → slip a”, but there is a problem because “straight forward → slip b” and “sudden brake → slip b” exist. There is no.

  Also, in FIG. 28, the adjacency matrix M5b is a result of deleting the relationship relating to “sudden braking → slip b” from the adjacency matrix M5. The reachable matrix M6b is a reachable matrix recalculated based on the adjacency matrix M5b. Comparing the reachable matrix M6b and the reachable matrix M1 (FIG. 18), the reachable matrix M6b includes “sudden brake → center line over” or “sudden brake → frontal collision” existing in the reachable matrix M1. There is no relationship (path) related to. That is, the deletion of the relationship (edge) related to “sudden braking → slip b” leads to the removal of the initially existing path. Therefore, the relationship relating to “sudden braking → slip b” is not a deletion target. The reachable matrix M6b does not have a relationship related to “straight forward → slip b” and “sudden brake → slip b”, but there is a problem because “straight forward → slip a” and “sudden brake → slip a” exist. There is no.

  In FIG. 29, the adjacency matrix M5c is the result of deleting the relationship relating to “overspeed → slip a” from the adjacency matrix M5. The reachable matrix M6c is a reachable matrix recalculated based on the adjacency matrix M5c. Comparing the reachable matrix M6c and the reachable matrix M1 (FIG. 18), the reachable matrix M6c has a relationship relating to “overspeed → slip a” corresponding to “overspeed → slip” in the reachable matrix M1 ( (Pass) does not exist, but there is no problem because the relationship (pass) relating to “overspeed → slip b” exists. That is, the deletion of the relationship (edge) related to “overspeed → slip a” does not involve removal of the initially existing path. Therefore, “overspeed → slip a” is a deletion target.

Also, in FIG. 29, the adjacency matrix M5d is a result of deleting the relationship relating to “overspeed → slip b” from the adjacency matrix M5. The reachable matrix M6d is a reachable matrix recalculated based on the adjacency matrix M5d. Comparing the reachable matrix M6d and the reachable matrix M1 (FIG. 18), the reachable matrix M6d includes “overspeed → slip”, “overspeed → center line over” existing in the reachable matrix M1, Or, there is no relationship (path) related to “overspeed → frontal collision”. In other words, the deletion of the relationship (edge) related to “overspeed → slip b” leads to the removal of the initially existing path. Therefore, the relationship relating to “overspeed → slip b” is not a deletion target. Note that the determination result of the deletion target described with reference to FIGS. 28 and 29 coincides with FIG.

  As described above, a process based on a matrix indicating a relationship between events may be used to specify a nonexistent path, select a node to be copied, and determine a relationship to be deleted (redundant edge). .

  In the present embodiment, the non-existing path specifying unit 16 is an example of a specifying unit. The redundant edge deletion unit 19 is an example of a deletion unit. The event flow storage unit 22 is an example of a directed graph storage unit.

  As mentioned above, although the Example of this invention was explained in full detail, this invention is not limited to such specific embodiment, In the range of the summary of this invention described in the claim, various deformation | transformation・ Change is possible.

Regarding the above description, the following items are further disclosed.
(Appendix 1)
With reference to an event chain storage unit that stores a plurality of event chains and a directed graph storage unit that stores a directed graph created by integrating the plurality of event chains, the plurality of event chains are included in a path included in the directed graph. A specific part that identifies a route that does not exist in any of the event chains;
A node that connects a plurality of edges on the downstream side in the identified path, a most downstream node excluding an end point node of the path, and a duplication unit that creates a duplicate of the edge connected to the node;
A directed graph creation device comprising: a deletion unit that deletes an edge in which any of the paths included in the plurality of event chains is not removed from the directed graph even if the edge is deleted among the copied edges.
(Appendix 2)
The deletion unit deletes an edge belonging to the specified path from one of the replication source node and the replication destination node among the edges connected to the downstream side of the replication source node or the replication destination node, and identifies the identified node. The directed graph creation device according to supplementary note 1, wherein an edge that does not belong to the route is deleted from the other of the replication source node and the replication destination node.
(Appendix 3)
The deletion unit deletes the edge connected to the downstream side and then deletes the edge among the edges connected to the upstream side of the replication source node or the replication destination node. The directed graph generation device according to supplementary note 2, wherein an edge that is not removed from any of the directed graphs is deleted.
(Appendix 4)
With reference to an event chain storage unit that stores a plurality of event chains and a directed graph storage unit that stores a directed graph created by integrating the plurality of event chains, the plurality of event chains are included in a path included in the directed graph. Identify routes that do not exist in any of the event chains,
Create a copy of the edge connected to the most downstream node other than the end point node of the route among the nodes connected to the downstream side in the identified route, and the edge connected to the node,
A directed graph creation method in which a computer executes a process of deleting an edge in which any path included in the plurality of event chains is not removed from the directed graph even if the edge is deleted among the copied edges.
(Appendix 5)
In the deletion process, the edge belonging to the specified path is deleted from one of the replication source node and the replication destination node among the edges connected to the downstream side of the replication source node or the replication destination node, and the identification is performed. The directed graph creation method according to supplementary note 4, wherein an edge that does not belong to the route is deleted from the other of the replication source node and the replication destination node.
(Appendix 6)
Even if the edge connected to the upstream side of the replication source node or the replication destination node is deleted after the edge connected to the downstream side is deleted, the plurality of event chains are deleted. 6. The directed graph generation method according to appendix 5, wherein an edge that is not removed from the directed graph is deleted by any of the paths included in the directed graph.
(Appendix 7)
With reference to an event chain storage unit that stores a plurality of event chains and a directed graph storage unit that stores a directed graph created by integrating the plurality of event chains, the plurality of event chains are included in a path included in the directed graph. Identify routes that do not exist in any of the event chains,
Create a copy of the edge connected to the most downstream node other than the end point node of the route among the nodes connected to the downstream side in the identified route, and the edge connected to the node,
A directed graph creation program for causing a computer to execute processing for deleting an edge in which any of the paths included in the plurality of event chains is not removed from the directed graph even if the edge is deleted among the copied edges.
(Appendix 8)
In the deletion process, the edge belonging to the specified path is deleted from one of the replication source node and the replication destination node among the edges connected to the downstream side of the replication source node or the replication destination node, and the identification is performed. The directed graph creation method according to appendix 7, wherein an edge that does not belong to the route is deleted from the other of the replication source node and the replication destination node.
(Appendix 9)
Even if the edge connected to the upstream side of the replication source node or the replication destination node is deleted after the edge connected to the downstream side is deleted, the plurality of event chains are deleted. The directed graph generation method according to appendix 8, wherein an edge that is not removed from the directed graph is deleted by any of the paths included in the directed graph.

DESCRIPTION OF SYMBOLS 10 Event flow creation apparatus 11 Document information storage part 12 Event chain extraction part 13 Expansion part 14 Binary relation storage part 15 Integration part 16 Nonexistent path specification part 17 Duplication target selection part 18 Node duplication part 19 Redundant edge deletion part 21 Event chain Storage unit 22 Event flow storage unit 100 Drive device 101 Recording medium 102 Auxiliary storage device 103 Memory device 104 CPU
105 Display device 106 Input device B Bus

Claims (5)

With reference to an event chain storage unit that stores a plurality of event chains and a directed graph storage unit that stores a directed graph created by integrating the plurality of event chains, the plurality of event chains are included in a path included in the directed graph. A specific part that identifies a route that does not exist in any of the event chains;
A node that connects a plurality of edges on the downstream side in the identified path, a most downstream node excluding an end point node of the path, and a duplication unit that creates a duplicate of the edge connected to the node;
A directed graph creation device comprising: a deletion unit that deletes an edge in which any of the paths included in the plurality of event chains is not removed from the directed graph even if the edge is deleted among the copied edges.   The deletion unit deletes an edge belonging to the specified path from one of the replication source node and the replication destination node among the edges connected to the downstream side of the replication source node or the replication destination node, and identifies the identified node. The directed graph creation device according to claim 1, wherein an edge that does not belong to the route is deleted from the other of the replication source node and the replication destination node.   The deletion unit deletes the edge connected to the downstream side and then deletes the edge among the edges connected to the upstream side of the replication source node or the replication destination node. The directed graph generation device according to claim 2, wherein an edge that is not removed from the directed graph is deleted by any one of the paths. With reference to an event chain storage unit that stores a plurality of event chains and a directed graph storage unit that stores a directed graph created by integrating the plurality of event chains, the plurality of event chains are included in a path included in the directed graph. Identify routes that do not exist in any of the event chains,
Create a copy of the edge connected to the most downstream node other than the end point node of the route among the nodes connected to the downstream side in the identified route, and the edge connected to the node,
A directed graph creation method in which a computer executes a process of deleting an edge in which any path included in the plurality of event chains is not removed from the directed graph even if the edge is deleted among the copied edges. With reference to an event chain storage unit that stores a plurality of event chains and a directed graph storage unit that stores a directed graph created by integrating the plurality of event chains, the plurality of event chains are included in a path included in the directed graph. Identify routes that do not exist in any of the event chains,
Create a copy of the edge connected to the most downstream node other than the end point node of the route among the nodes connected to the downstream side in the identified route, and the edge connected to the node,
A directed graph creation program for causing a computer to execute processing for deleting an edge in which any of the paths included in the plurality of event chains is not removed from the directed graph even if the edge is deleted among the copied edges.

Images