JP6612680B2 - Logical relationship recognition apparatus, logical relationship recognition method, and logical relationship recognition program - Google Patents

Description

  The present invention relates to a logical relationship recognition apparatus, a logical relationship recognition method, and a logical relationship recognition program.

  OpS (Operation Support System: Operation Support System) has been introduced for business processing in various companies. On the other hand, for operations that cannot be supported by OpS, each organization has its own operations, and forms are used for information distribution. Since the form is semi-structured data, the meaning (logical relationship) of each value can be understood by a person but difficult for a machine, and a large amount of manual work is required when utilizing the data on the form. There is a burden. On the other hand, a technique for automatically recognizing a logical relationship between item names of a form and a logical relationship between item names and item values based on the arrangement of ruled line frames is known (for example, Patent Documents). 1 and non-patent documents 1 to 3).

JP, 2006-004417, A

Takagi et al., “Examination of Extraction Techniques for Cross-sectional Data Manipulation Techniques for Electronic Forms”, IEICE Technical Report, vol.114, No.150, pp.1-6 (2014) Takagi et al., “Investigation of visual representation of forms for linking electronic form data”, IEICE Communication Society Conference Proceedings (2015) Takagi et al., "Examination of data structure conversion method for electronic forms using visual expression", IEICE Technical Report, vol.115, No.409, pp.25-30 (2016)

  However, in the conventional technology, when a form includes vertical enumeration, horizontal enumeration, a composite type of vertical enumeration and horizontal enumeration, or enumerated type nested structure, the logical relationship between the item names of the form, and the item names and item values There is a problem in that the logical relationship between and cannot be recognized correctly.

  For example, in the case of vertical enumeration, there may be a plurality of ruled line frames of the same width corresponding to the item names in the vertical direction. Further, each item name has an item value, and the width of the ruled line frame corresponding to the item value is the same as the width of the ruled line frame corresponding to the item name. In such a case, the conventional technology cannot identify which ruled line frame is the item name, so it accurately recognizes the logical relationship between the item names and the logical relationship between the item names and the item values. It is difficult to do.

  The logical relationship recognition apparatus of the present invention represents information relating to an area representing an item name or item value of a form as a node, and is set in advance among the nodes based on a graph representing an adjacency relationship between the nodes as an edge. An extraction unit that extracts a node satisfying the condition as an item name node that is a node in an area representing an item name, and when a plurality of nodes are adjacent in a predetermined direction of one node, An item in which an item value node, which is a node in an area representing an item value in a predetermined direction, is adjacent to the first deletion unit that deletes an edge representing an adjacent relationship with the plurality of nodes. A second deletion unit that deletes an edge representing an adjacency relationship between the name node and the item value node, and the deletion of the edge is performed by the first deletion unit and the second deletion unit Based on the rough, based on the first acquisition unit that acquires the logical relationship between the item name node and the item value node, and the graph in which the edge deletion is performed by the first deletion unit, A second acquisition unit that acquires an inclusion relationship between the item name nodes, a logical relationship acquired by the first acquisition unit, and a tree structure that combines the inclusion relationship acquired by the second acquisition unit And a synthesis unit for creating data.

  The logical relationship recognition method of the present invention is a logical relationship recognition method executed by a logical relationship recognition device, wherein information relating to an area representing an item name or item value of a form is represented as a node, and the adjacent relationship between the nodes And extracting a node satisfying a preset condition from among the nodes as an item name node that is a node in a region representing an item name, and a predetermined value of one node When a plurality of nodes are adjacent to each other in the direction, a first deletion step of deleting an edge representing an adjacency relationship between the one node and the plurality of nodes, and deletion of an edge by the first deletion step A second acquisition step of acquiring an inclusion relationship between the item name nodes based on the performed graph; and a node of an area representing an item value in a predetermined direction among the item name nodes An item name node adjacent to an item value node, a second deletion step of deleting an edge representing an adjacency relationship between the item value node, an edge formed by the first deletion step and the second deletion step A first acquisition step of acquiring a logical relationship between the item name node and the item value node based on the graph in which the deletion is performed, and a logical relationship acquired by the first acquisition step, And a synthesizing step of creating data of a tree structure obtained by synthesizing the inclusion relation acquired in the second acquiring step.

  The logical relationship recognition program of the present invention represents information relating to an area representing an item name or item value of a form as a node on a computer, and based on a graph representing an adjacency relationship between the nodes as an edge, An extraction step of extracting a node that satisfies a preset condition as an item name node that is a node in an area representing an item name, and when a plurality of nodes are adjacent in a predetermined direction of one node, the 1 A first deletion step of deleting an edge representing an adjacent relationship between one node and the plurality of nodes, and the item name nodes between the item name nodes based on the graph in which the deletion of the edge is performed by the first deletion step. A second acquisition step of acquiring an inclusion relationship, and an item value node that is a node of an area representing an item value in a predetermined direction is adjacent to the item name node The edge is deleted by the second deletion step of deleting the edge representing the adjacent relationship between the item name node and the item value node, and the first deletion step and the second deletion step. Based on the graph, a first acquisition step of acquiring a logical relationship between the item name node and the item value node, a logical relationship acquired by the first acquisition step, and the second acquisition And a synthesizing step for creating tree-structured data obtained by synthesizing the inclusion relation acquired in the step.

  According to the present invention, even when the form includes a vertical enumeration, a horizontal enumeration, a composite type of vertical enumeration and horizontal enumeration, or an enumerated type nested structure, the logical relationship between the item names of the form, and the item names and The logical relationship between the item values can be recognized accurately.

FIG. 1 is a diagram for explaining the outline of the logical relationship recognition process. FIG. 2 is a diagram illustrating an example of vertical listing. FIG. 3 is a diagram illustrating an example of horizontal enumeration. FIG. 4 is a diagram illustrating an example of a combined type of vertical enumeration and horizontal enumeration. FIG. 5 is a diagram illustrating an example of an enumerated type nested structure. FIG. 6 is a diagram illustrating an example of an enumerated type nested structure. FIG. 7 is a diagram illustrating an example of an enumerated type nested structure. FIG. 8 is a diagram illustrating an example of an enumerated type nested structure. FIG. 9 is a diagram illustrating an example of the configuration of the logical relationship recognition apparatus according to the first embodiment. FIG. 10 is an example of an enumeration list. FIG. 11 is an example of an enumeration list. FIG. 12 is a diagram for explaining the inclusion graph. FIG. 13 is a diagram for explaining the inclusion graph. FIG. 14 is an example of tree structure data. FIG. 15 is an example of tree-structured data. FIG. 16 is a flowchart showing the flow of processing of the extraction unit. FIG. 17 is a flowchart showing the flow of processing of the analysis unit. FIG. 18 is a flowchart showing the flow of processing of the first deletion unit. FIG. 19 is a flowchart showing the flow of processing of the second deletion unit. FIG. 20 is a flowchart showing the flow of processing of the classification unit. FIG. 21 is a flowchart illustrating a process flow of the vertical enumeration acquisition unit. FIG. 22 is a flowchart showing the flow of processing of the horizontal enumeration acquisition unit. FIG. 23 is a flowchart showing the flow of processing of the inclusion relationship acquisition unit. FIG. 24 is a flowchart showing the flow of processing for acquiring the right inclusion relationship. FIG. 25 is a flowchart showing the flow of processing for acquiring the lower inclusion relationship. FIG. 26 is a flowchart illustrating a process flow of the inclusion graph generation unit. FIG. 27 is a flowchart showing the flow of processing of the item name synthesizing unit. FIG. 28 is a flowchart showing the flow of processing of the enumeration synthesis unit. FIG. 29 is a flowchart showing the flow of processing of the adding unit. FIG. 30 is a diagram for explaining another embodiment. FIG. 31 is a diagram illustrating an example of a computer in which a logical relationship recognition apparatus is realized by executing a program.

  Hereinafter, embodiments of a logical relationship recognition device, a logical relationship recognition method, and a logical relationship recognition program according to the present application will be described in detail with reference to the drawings. In addition, this invention is not limited by this embodiment.

  First, the outline of the logical relationship recognition process by the logical relationship recognition system having the logical relationship recognition device will be described with reference to FIG. FIG. 1 is a diagram for explaining the outline of the logical relationship recognition process. As shown in FIG. 1, first, the logical relationship recognition system reads an enumerated form from a PC or the like (step S1). At this time, the data read by the logical relationship recognition system is not limited to a form, and may be semi-structured data such as a Web GUI, a system GUI, and an enumerated structure on an image. Next, the logical relationship recognition system acquires information on the ruled line frame from the read form (step S2). Further, the logical relationship recognition system may acquire the description content from an image of a paper form read by a scanner or the like by OCR (Optical Character Recognition) (step S3).

  Then, the logical relationship recognition system generates a style graph based on the ruled line frame information and the item name definition information (step S4). The ruled line frame information includes visual information such as the coordinates of the ruled line frame, the character string in the ruled line frame, the fill color of the ruled line frame, the type, thickness, and color of the ruled line. Further, the item name definition information includes a condition for determining a ruled line frame as an item name.

  For example, item name definition information whose item name is a ruled frame with a yellow color is described as “if {node: {color: # FFFF00}} then item”. Also, for example, ruled line frame definition information whose item name is a non-empty ruled line frame of a character string is described as “if {node: {! String: null}} then item”. Also, for example, ruled line frame definition information whose item name is a ruled line frame whose color is not white and whose character string is not empty is “if {node: {! Color: whilte}, {! String: null} then item” Is described. Further, for example, ruled line frame definition information whose item name is a ruled line frame with the character string “y1” is described as “if {node: {string:“ y1 ”}} then item”.

  Further, the style graph is a graph in which a ruled line frame or the like is represented as a node and an adjacent relationship between nodes is represented as an edge for each of a plurality of styles included in the form. In the subsequent processing, the logical relationship recognition system handles the form format as graph format data. The format also includes an enumeration. In the following description, the case where the logical relationship is recognized based on the style graph will be described. However, a form graph in which the entire form is represented by nodes and edges may be used.

  Next, the logical relationship recognition system recognizes the logical relationship from the structure of the enumeration format (step S5), converts the recognized result into a data structure of a predetermined format (xml, yaml, json, etc.) (step S6), Store in DB. At this time, the link path of the data structure may be stored in the DB, or the data may be stored in accordance with a predefined schema.

  Next, the types of enumeration that are targets of logical relationship recognition processing by the logical relationship recognition system will be described with reference to FIGS. 2 to 8 represent nodes. In the following description, for the sake of explanation, the nodes indicated by these symbols are simply referred to as nodes, and are referred to as item name nodes or item value nodes. There is a case.

  FIG. 2 is a diagram illustrating an example of vertical listing. As shown in FIG. 2, the vertical enumeration has a plurality of logical relationships in the vertical direction, which are the relationship between item names and the relationship between item names and item values. For example, the nodes a1 and a2 and the nodes a4 and a5 have a logical relationship between item names and item values. FIG. 3 is a diagram illustrating an example of horizontal enumeration. As shown in FIG. 3, the horizontal enumeration has a plurality of logical relationships in the horizontal direction. For example, the nodes b1 and b2 have a logical relationship between item names and item values. 2 to 8, the shaded portion represents an item name, and the portion not shaded represents an item value.

  FIG. 4 is a diagram illustrating an example of a combined type of vertical enumeration and horizontal enumeration. As shown in FIG. 4, in the case of a combined type of vertical enumeration and horizontal enumeration, vertical enumeration and horizontal enumeration exist in one mode. For example, nodes c1 and c2 and nodes c3 and c4 are listed horizontally. For example, nodes c5 and c7 and nodes c6 and c8 are listed vertically.

  5 to 8 are diagrams illustrating an example of an enumerated type nested structure. As shown in FIG. 5, the enumerated nested structure has a plurality of nested enumerations in the vertical or horizontal direction. For example, nodes d1 and d2 have a logical relationship between item names. Further, the nodes d2 and d3 have a logical relationship between item names and item values. Thus, the logical relationship between the nodes d2 and d3 is a nesting of the logical relationship between the nodes d1 and d2.

  As shown in FIG. 6, for example, nodes e1 and e2 have a logical relationship between item names. Further, the nodes e2 and e3 have a logical relationship between item names and item values. Thus, the logical relationship between the nodes e2 and e3 is a nesting of the logical relationship between the nodes e1 and e2.

  Also, as shown in FIG. 7, for example, the nodes f1 and f2 have a logical relationship between item names. Nodes f2 and f3 have a logical relationship between item names. Further, the nodes f3 and f4 have a logical relationship between item names and item values. Thus, the logical relationship between the nodes f3 and f4 is a nesting of the logical relationship between the nodes f2 and f3. Further, the logical relationship between the nodes f2 and f3 is a nesting of the logical relationship between the nodes f1 and f2.

  As shown in FIG. 8, for example, nodes g2 and g3 have a logical relationship between item names and item values. Nodes g6 and g7 have a logical relationship between item names. Nodes g7 and g8 have a logical relationship between item names and item values. Thus, the logical relationship between the nodes g7 and g8 is a nesting of the logical relationship between the nodes g6 and g7. However, the logical relationship between the nodes g2 and g3 is not nested between the logical relationship between the nodes g7 and g8 and the logical relationship between the nodes g6 and g7.

[Configuration of First Embodiment]
Next, the configuration of the logical relationship recognition apparatus according to the first embodiment will be described with reference to FIG. FIG. 9 is a diagram illustrating an example of the configuration of the logical relationship recognition apparatus according to the first embodiment. As illustrated in FIG. 9, the logical relationship recognition apparatus 10 includes a control unit 20 and a storage unit 30.

  The control unit 20 controls the entire logical relationship recognition apparatus 10. The control unit 20 is, for example, an electronic circuit such as a CPU (Central Processing Unit) or MPU (Micro Processing Unit), or an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array). The control unit 20 has an internal memory for storing programs and control data that define various processing procedures, and executes each process using the internal memory. The control unit 20 functions as various processing units when various programs are operated. For example, the control unit 20 includes an extraction unit 201, an analysis unit 202, a first deletion unit 203, a second deletion unit 204, a classification unit 205, a vertical enumeration acquisition unit 206, a horizontal enumeration acquisition unit 207, and an inclusion relationship acquisition unit 208. , An inclusion graph generation unit 209, an item name synthesis unit 210, an enumeration synthesis unit 211, and an addition unit 212.

  The storage unit 30 is a storage device such as an HDD (Hard Disk Drive), an SSD (Solid State Drive), or an optical disk. Note that the storage unit 30 may be a semiconductor memory that can rewrite data, such as a random access memory (RAM), a flash memory, and a non-volatile static random access memory (NVSRAM). The storage unit 30 stores an OS (Operating System) executed by the logical relationship recognition apparatus 10 and various programs. Furthermore, the storage unit 30 stores various information used in executing the program. In addition, the storage unit 30 stores, for example, an enumeration list 301 and an inclusion graph 302.

  Here, the logical relationship recognition processing by the logical relationship recognition device 10 will be described, and details of each part of the logical relationship recognition device 10 will be described. The extraction unit 201 represents information related to the area representing the item name or item value of the form as a node, and a pre-set condition among the nodes based on a graph representing the adjacent relationship between nodes as an edge, that is, a style graph. Nodes satisfying the condition are extracted as item name nodes that are nodes in the area representing the item name.

  Here, a rule for extracting the item name node is created by the analysis unit 202. The analysis unit 202 reads a pre-analysis visual expression rule created in advance as a text file or the like, analyzes conditions and actions from the read visual expression rule, and creates a visual expression rule that can be recognized by a computer. In the following description, when simply referred to as a visual expression rule, it refers to a visual expression rule after analysis.

  When a plurality of nodes are adjacent to each other in a predetermined direction of one node, the first deletion unit 203 deletes an edge representing an adjacency relationship between the one node and the plurality of nodes. Note that the first deletion unit 203 may delete an edge representing an adjacency relationship between one node and a plurality of nodes when a plurality of nodes are adjacent to the left or upper side of the one node. Good. For example, in the case of the combined type of vertical enumeration and horizontal enumeration in FIG. 4, the first deletion unit 203 does not delete the edge because there are no multiple nodes adjacent to the left or upper side for any node. Further, for example, in the case of the enumerated type nested structure of FIG. 8, since a plurality of nodes are adjacent to the upper side of the node g6, the first deletion unit 203 includes the node g6, the nodes g1, g2, g3, g4, and The edge representing the adjacency relationship with g5 is deleted.

  In addition, the second deletion unit 204 determines the adjacency relationship between the item name node and the item value node adjacent to the item value node that is the node of the area representing the item value in a predetermined direction among the item name nodes. Delete the representing edge. Note that the second deletion unit 204 has an adjacency relationship between an item name node and an item value node that is adjacent to an item value node that is a node of an area representing an item value on the left or upper side of the item name nodes. You may make it delete the edge showing. For example, in the case of the combined type of vertical enumeration and horizontal enumeration in FIG. 4, the item value node c4 is adjacent to the upper side of the item name node c6, so the second deletion unit 204 is adjacent to the node c6 and the node c4 Remove edges that represent relationships. Further, for example, in the case of the enumerated type nested structure of FIG. 6, since the item value nodes e3 and e7 are adjacent to the left side of the item name node e4, the second deletion unit 204 performs the node e4 and the nodes e3 and e7. Delete the edge representing the adjacency relationship with.

  Then, the vertical enumeration acquisition unit 206 and the horizontal enumeration acquisition unit 207, based on the graph in which the edges are deleted by the first deletion unit 203 and the second deletion unit 204, are used for the item name node and the item value node. Get logical relationship between. Specifically, the vertical enumeration acquisition unit 206 and the horizontal enumeration acquisition unit 207 acquire the logical relationship as a list as illustrated in FIG. 10 or 11 and store the acquired list in the storage unit 30 as the enumeration list 301. 10 and 11 are examples of enumerated lists.

  FIG. 10 is an enumeration list showing the logical relationship of the enumeration of FIG. The first line of the list in FIG. 10 indicates that the horizontal child of the item name node c1 is the node c2. Further, the third line of the list of FIG. 10 indicates that the vertical child of the item name node c5 is the node c7.

  FIG. 11 is an enumeration list showing the logical relationship of the enumeration of FIG. The first line of the list of FIG. 11 indicates that the horizontal child of the item name node f3 is the node f4. The second line of the list of FIG. 11 indicates that the horizontal child of the item name node f9 is the node f10.

  Further, for example, in the case of the enumerated type nested structure of FIG. 8, the first deletion unit 203 deletes the edge representing the adjacent relationship between the node g6 and the nodes g1, g2, g3, g4, and g5. The vertical enumeration acquisition unit 206 and the horizontal enumeration acquisition unit 207 do not acquire a logical relationship such that the vertical child of the item name node g2 is g8.

  Further, for example, in the case of the enumerated type nested structure of FIG. 8, since the edge indicating the adjacency relationship between the node g14 and the nodes g13 and g17 is deleted by the second deletion unit 204, the vertical enumeration acquisition unit 206 and The horizontal enumeration acquisition unit 207 does not acquire a logical relationship in which the horizontal child of the item name node g12 is g15.

  The inclusion relationship acquisition unit 208 acquires the inclusion relationship between the item name nodes based on the graph in which the edge is deleted by the first deletion unit 203. Further, the inclusion graph generation unit 209 generates a graph as illustrated in FIG. 12 based on the inclusion relationship acquired by the inclusion relationship acquisition unit 208 and stores the generated graph in the storage unit 30 as the inclusion graph 302. FIG. 12 is a diagram for explaining the inclusion graph. The broken line in FIG. 12 represents the vertical inclusion relationship. Also, the solid line in FIG. 12 represents the horizontal inclusion relationship. Hereinafter, when the target node includes one or more other nodes in the vertical direction (or horizontal direction), the target node is referred to as a vertical (or horizontal) included node. Among nodes having an inclusion relationship, a node that is not included in another node is referred to as a master node, and another node that is included is referred to as a slave node.

  The inclusion graph of FIG. 12 corresponds to the enumerated type nested structure of FIG. As shown in FIGS. 7 and 12, the node f2 includes nodes f3, f5, f7, f9, and f11 in the horizontal direction. The node f13 includes nodes f14, f16, f18, f20, f22, and f24 in the horizontal direction. The node f1 includes nodes f2 and f13 in the vertical direction. Further, f1 is a main node, and f2, f3, f5, f7, f9, f11, f13, f14, f16, f18, f20, f22, and f2 are subordinate nodes.

  Specifically, the inclusion relationship acquisition unit 208 includes at least one of the first node groups adjacent to the right side of the first item name node as an item name node, and all of the first node groups included in the first node group. If the height of the node is less than or equal to the height of the first item name node, and the vertex of the upper left node of the first node group and the vertex of the first item name node overlap, the first Is determined to include the first node group in the horizontal direction. In addition, the inclusion relationship acquisition unit 208 includes at least one of the second node groups adjacent to the lower side of the second item name node as an item name node, and all nodes included in the second node group The width of the second item name node is less than or equal to the width of the second item name node, and the vertex of the upper left node of the second node group overlaps the vertex of the first item name node, the second item name It is determined that the node includes the second node group in the vertical direction.

  For example, as shown in FIG. 7, at least the node f3 is an item name node in the node group adjacent to the right side of the item name node f2. The heights of all nodes included in the node group adjacent to the right side of the item name node f2 are all equal to or lower than the height of the item name node f2. Also, the upper left node of the node group adjacent to the right side of the item name node f2, that is, the vertex of the node f3 overlaps with the vertex of the item name node f2, and the lower left end of the node group adjacent to the right side of the item name node f2. , That is, the vertex of the node f4 overlaps the vertex of the item name node f2. Accordingly, the inclusion relationship acquisition unit 208 determines that the item name node f2 includes a node group adjacent to the right side of the item name node f2.

  Note that the format graph to be processed by the inclusion relationship acquisition unit 208 has been deleted by the first deletion unit 203 but not by the second deletion unit 204. is there. For this reason, for example, the edge representing the adjacent relationship between the node f4 and the nodes f5 and f7 is not deleted. Therefore, the inclusion relationship acquisition unit 208 determines that the item name node f2 includes the nodes f5 and f7 in the horizontal direction.

  The item name synthesizing unit 210 represents the inclusion graph as tree-structured data as shown in FIG. FIG. 13 is a diagram for explaining the inclusion graph. Then, the item name synthesizing unit 210 and the enumeration synthesizing unit 211 synthesize the logical relationship acquired by the vertical enumeration acquisition unit 206 and the horizontal enumeration acquisition unit 207 and the inclusion relationship acquired by the inclusion relationship acquisition unit 208. Create tree structure data. The adding unit 212 adds a root node that defines the tree structure to the tree structure data created by the item name synthesizing unit 210 and the enumeration synthesizing unit 211.

  For example, the enumeration synthesis unit 211 creates the tree structure data shown in FIG. 14 based on the list shown in FIG. Then, the adding unit 212 adds the root node “form1” to the data of the tree structure. FIG. 14 is an example of tree structure data. In this case, since there is no inclusion relationship in the enumeration in FIG. 4, the inter-name-name combining unit 210 does not create an inclusion graph.

  Further, for example, the enumeration synthesis unit 211 creates the tree structure data shown in FIG. 15 based on the list of FIG. 11 and the inclusion graph of FIG. Then, the adding unit 212 adds the root node “form2” to the data of the tree structure. FIG. 15 is an example of tree-structured data. In addition, since the root node is added to indicate the logical relationship that forms the format, the adding unit 212 is not essential if the information that configures the format is not necessary.

  In addition, the tree node that is each node of the tree structure data includes, for example, information for identifying child and parent nodes in addition to information acquired from form format information such as item name or item value character strings, Information indicating that the adjacent direction to the parent node, the tree node is included in the enumeration, and the like.

[Process of First Embodiment]
Next, the processing flow of the logical relationship recognition apparatus 10 will be described. First, the process of the extraction unit 201 will be described with reference to FIG. FIG. 16 is a flowchart showing the flow of processing of the extraction unit. As shown in FIG. 16, first, the extraction unit 201 reads a style graph and a visual expression rule after analysis (step S11). Next, the extraction unit 201 selects a node group that satisfies the condition of the post-analysis visual expression rule from the style graph (step S12). Next, the extraction unit 201 sets the item name attribute of the selected node group to the item name, that is, sets the value of the flag indicating the item name to true (step S13). Finally, the extraction unit 201 returns it as a style graph and G (step S14).

  Next, processing of the analysis unit 202 will be described with reference to FIG. FIG. 17 is a flowchart showing the flow of processing of the analysis unit. The analysis unit 202 creates a post-analysis visual expression rule based on the visual expression rule. First, the analysis unit 202 reads a visual expression rule group (step S21). Next, the analysis unit 202 generates an Array or Hash type variable rule_list (step S22). Thereafter, the analysis unit 202 processes the read visual expression rules one by one (steps S23 and S27).

  First, the analysis unit 202 analyzes the condition of the visual expression rule (step S24). Next, the analysis unit 202 analyzes the action of the visual expression rule (step S25). Then, the analysis unit 202 stores the analyzed conditions and actions in the rule_list as post-analysis visual expression rules (step S26). The analysis unit 202 processes all the visual expression rules, and then outputs a rule_list storing the post-analysis visual expression rule group (step S28).

  Next, processing of the first deletion unit 203 will be described with reference to FIG. FIG. 18 is a flowchart showing the flow of processing of the first deletion unit. First, the first deletion unit 203 reads a style graph (step S31).

  Next, the 1st deletion part 203 performs the following processes about each node of a style graph (step S32, S37). First, when the number of nodes adjacent in the left direction is two or more (step S33, true), the first deletion unit 203 deletes the adjacent relationship between the node adjacent in the left direction and the target node (step S34). ). When the number of nodes adjacent in the left direction is not two or more (step S33, false), the first deletion unit 203 does not delete the adjacent relationship.

  Next, when the number of nodes adjacent in the upward direction is two or more (step S35, true), the first deletion unit 203 deletes the adjacent relationship between the node adjacent in the upward direction and the target node (step S35). S36). If the number of nodes adjacent in the upward direction is not two or more (step S35, false), the first deletion unit 203 does not delete the adjacent relationship. The 1st deletion part 203 outputs a style graph, after processing about all the object nodes (step S38).

  For example, in the example of FIG. 8, the first deletion unit 203 represents the adjacency relationship between the nodes g1, g2, g3, g4, and g5 and the node g6, and the adjacency relationship between the nodes g13 and g17 and the node g14. Delete the edge.

  Next, the processing of the second deletion unit 204 will be described with reference to FIG. FIG. 19 is a flowchart showing the flow of processing of the second deletion unit. First, the second deletion unit 204 reads a style graph and an item name node list (step S41). The item name node list is a list of item name nodes extracted by the extraction unit 201.

  Next, the second deletion unit 204 performs the following processing for each item name node in the list of item name nodes (steps S42 and S47). First, when the item value node is adjacent in the left direction (step S43, true), the second deletion unit 204 deletes the adjacent relationship between the node adjacent in the left direction and the target node (step S44). When the item value node is not adjacent in the left direction (step S43, false), the second deletion unit 204 does not delete the adjacent relationship.

  Next, when the item value node is adjacent in the upward direction (step S45, true), the second deletion unit 204 deletes the adjacent relationship between the node adjacent in the upward direction and the target node (step S46). If the item value node is not adjacent in the upward direction (step S45, false), the second deletion unit 204 does not delete the adjacent relationship. The 2nd deletion part 204 outputs a style graph, after processing about all the object nodes (step S48).

  For example, in the example of FIG. 8, the second deletion unit 204 includes the adjacency relationship between the node g3 and the node g4, the adjacency relationship between the node g8 and the nodes g12 and g13, the nodes g14 and g17, and the node g15. Delete the edge representing the adjacency relationship.

  Next, the processing of the classification unit 205 will be described with reference to FIG. FIG. 20 is a flowchart showing the flow of processing of the classification unit. From now on, the relationship with the node obtained by tracing the node adjacent to the right side from the target node is “connected to the right side”, and the relationship with the node obtained by tracing the node adjacent to the lower side from the target node Called “connect to the bottom”. As shown in FIG. 20, the classification unit 205 first reads the style graph and the target node (step S51). If the target node is not an item name (step S52, true), the classification unit 205 sets the enumeration flag to “no enumeration” (step S58). The enumeration flag is a variable indicating the classification of each target node as one of “vertical enumeration”, “horizontal enumeration”, and “no enumeration”. In addition, the enumeration includes a composite type of vertical enumeration and horizontal enumeration, and an enumeration nested structure. These enumerations can be expressed by a combination of vertical enumeration and horizontal enumeration.

  If the target node is an item name (step S52, false), the classification unit 205 stores the node group connected to the lower side of the target node in bottoms, and sets the node group connected to the right side of the target node to rights. Store (step S53).

  Here, when the item name is not included in rights and the number of rights is 1 (step S54, true), the classification unit 205 sets the enumeration flag to “horizontal enumeration” (step S55). If the item name is included in rights, or if the number of rights is not 1 (step S54, false), the classification unit 205 performs the following processing.

  When the item name is not included in bottoms and the number of bottoms is 1 (step S56, true), the classification unit 205 sets the enumeration flag to “vertical enumeration” (step S57). When the item name is included in bottoms, or when the number of bottoms is not 1 (step S56, false), the classification unit 205 sets the enumeration flag to “no enumeration” (step S58). Finally, the classification unit 205 outputs an enumeration flag (step S59).

  For example, in the example of FIG. 6, when the target node is the node e1, the classification unit 205 sets the enumeration flag to “no enumeration”. When the target node is the node e4, the classification unit 205 sets the enumeration flag to “horizontal enumeration”.

  Next, processing of the vertical enumeration acquisition unit 206 will be described with reference to FIG. FIG. 21 is a flowchart illustrating a process flow of the vertical enumeration acquisition unit. As shown in FIG. 21, first, the vertical enumeration acquiring unit 206 reads a style graph and a target node (step S61). Next, the vertical enumeration acquisition unit 206 stores a node group connected to the lower side of the target node in bottoms (step S62). Here, when the item name is included in bottoms (step S63, true), the vertical enumeration acquisition unit 206 ends the process. If the item name is not included in bottoms (step S63, false), the vertical enumeration acquisition unit 206 acquires a logical relationship in which the parent is the target node, the child is bottoms, and the direction is vertical (step S64). Then, the vertical enumeration acquiring unit 206 adds the acquired logical relationship to the enumeration list 301 (step S65).

  For example, in the example of FIG. 2, when the target node is the node a1, the vertical enumeration acquisition unit 206 stores the nodes a2 and a3 in bottoms, the parent is the node a1, the child is the nodes a2 and a3, and the direction is the vertical Get the logical relationship. When the target node is the node a4, the vertical enumeration acquisition unit 206 stores the node a5 in bottoms, and acquires a logical relationship in which the parent is the node a4, the child is the node a5, and the direction is vertical.

  Next, the process of the horizontal enumeration acquisition unit 207 will be described with reference to FIG. FIG. 22 is a flowchart showing the flow of processing of the horizontal enumeration acquisition unit. As shown in FIG. 22, first, the horizontal enumeration acquisition unit 207 reads the style graph and the target node (step S71). Next, the horizontal enumeration acquisition unit 207 stores a node group connected to the right side of the target node in rights (step S72). Here, when the item name is included in rights (step S73, true), the horizontal enumeration acquisition unit 207 ends the process. If the item name is not included in rights (step S73, false), the horizontal enumeration acquisition unit 207 acquires a logical relationship in which the parent is the target node, the child is rights, and the direction is horizontal (step S74). Then, the horizontal enumeration acquisition unit 207 adds the acquired logical relationship to the enumeration list 301 (step S75).

  For example, in the example of FIG. 3, when the target node is the node b1, the horizontal enumeration acquisition unit 207 stores the node b2 in rights, sets the parent as the node b1, the child as the node b2, and the direction as the horizontal relationship. get. When the target node is the node b3, the horizontal enumeration acquisition unit 207 stores the node b4 in rights, and acquires a logical relationship in which the parent node is b3, the child is b4, and the direction is horizontal.

  Next, processing of the inclusion relationship acquisition unit 208 will be described with reference to FIG. FIG. 23 is a flowchart showing the flow of processing of the inclusion relationship acquisition unit. The inclusion relationship acquisition unit 208 first reads a style graph (step S81).

  Here, the inclusion relationship acquisition unit 208 performs the following processing for each node included in the style graph (steps S82 and S87). First, when the node itself includes the item name and the node having the item name in the right adjacent node (step S83, true), the inclusion relationship acquisition unit 208 acquires the right inclusion relationship (step S84). Details of the process of acquiring the right inclusion relationship will be described later. Also, if the node itself is not the item name, or if the node adjacent to the right direction does not include the item name node (step S83, false), the inclusion relationship acquisition unit 208 does not acquire the right inclusion relationship.

  Next, when the node itself includes the item name and the node of the item name in the node adjacent in the downward direction (step S85, true), the inclusion relationship acquisition unit 208 acquires the lower inclusion relationship (step S86). . Details of the process of acquiring the lower inclusion relationship will be described later. If the node itself is not an item name, or if the node adjacent to the downward direction does not include the item name node (step S85, false), the inclusion relationship acquisition unit 208 does not acquire the right inclusion relationship. The inclusion relationship acquisition unit 208 processes all the nodes and then outputs the acquired inclusion relationship as an inclusion relationship list (step S88).

  Next, the process of acquiring the right inclusion relationship will be described with reference to FIG. FIG. 24 is a flowchart showing the flow of processing for acquiring the right inclusion relationship. As shown in FIG. 24, first, the inclusion relationship acquisition unit 208 reads the target node and the style graph (step S101). Next, the inclusion relationship acquisition unit 208 stores the minimum y coordinate value of the node group on the right side of the target node in min_y, and stores the maximum y coordinate value of the node group on the right side of the target node in max_y. (Step S102).

  When the y-coordinate range of the target node matches the y-coordinate range of the right node group of the target node (step S103, true), the inclusion relationship acquisition unit 208 sets the list to the right node of the target node. The group is stored (step S107). The inclusion relationship acquisition unit 208 calculates the y-coordinate range of the node group on the right side of the target node using min_y and max_y. Then, the inclusion relationship acquisition unit 208 performs the following processing for each node included in the list (steps S108 and S110). The inclusion relationship acquisition unit 208 returns to step S102 and executes recursion processing every time each node r is set as a target node (step S109).

  Also, when the y-coordinate range of the target node does not match the y-coordinate range of the right node group of the target node (step S103, false), and the item value is not included in the included node ( In step S104, false), the inclusion relationship acquisition unit 208 initializes the slave node set (list) included on the right side (step S105). In addition, when the item value is included in the included node (step S104, true), the inclusion relationship acquisition unit 208 does not perform initialization. Finally, the inclusion relationship acquisition unit 208 outputs list (step S106).

  Next, processing for acquiring the lower inclusion relationship will be described with reference to FIG. FIG. 25 is a flowchart showing the flow of processing for acquiring the lower inclusion relationship. As shown in FIG. 25, first, the inclusion relationship acquisition unit 208 reads the target node and the style graph (step S151). Next, the inclusion relationship acquisition unit 208 stores the minimum x coordinate value of the node group below the target node in min_x, and the maximum x coordinate value of the node group below the target node in max_x. Is stored (step S152).

  If the x-coordinate range of the target node matches the x-coordinate range of the lower node group of the target node (step S153, true), the inclusion relationship acquisition unit 208 sets the list below the target node. Are stored (step S157). The inclusion relationship acquisition unit 208 calculates the y-coordinate range of the lower node group of the target node using min_x and max_x. Then, the inclusion relationship acquisition unit 208 performs the following processing for each node included in the list (steps S158 and S160). The inclusion relationship acquisition unit 208 returns to step S152 and executes recursion processing every time each node r is the target node (step S159).

  Further, when the x-coordinate range of the target node does not match the x-coordinate range of the lower node group of the target node (step S153, false), and the item value is not included in the included node (Step S154, false), the inclusion relationship acquisition unit 208 initializes a subordinate node set (list) included in the lower side (Step S155). In addition, when the item value is included in the included node (step S154, true), the inclusion relationship acquisition unit 208 does not perform initialization. Finally, the inclusion relationship acquisition unit 208 outputs list (step S156).

  Next, processing of the inclusion graph generation unit 209 will be described with reference to FIG. FIG. 26 is a flowchart illustrating a process flow of the inclusion graph generation unit. As shown in FIG. 26, first, the inclusion graph generation unit 209 reads a style graph and an inclusion relation list (step S201). Next, the inclusion graph generation unit 209 stores the vertical inclusion node acquired from the inclusion relation list in Nv, the horizontal inclusion node acquired from the inclusion relation list in Nh, and the new in IG. The inclusion graph set is stored (step S202).

  Here, the inclusion graph generation unit 209 performs the following processing for each inclusion node i included in Nv and Nh (steps S203 and S213). Hereinafter, when the target node is already assigned to the inclusion graph due to the inclusion relationship of other nodes, the target node is referred to as a “divided node”. First, when the inclusion node i has been divided (step S204, true), the inclusion graph generation unit 209 proceeds to processing of the next inclusion node. If the inclusion node i has not been divided (step S204, false), the inclusion graph generation unit 209 stores the inclusion node i and the subordinate item name node set in inc (step S205).

  Here, when there is a divided node in inc (step S206, true), the inclusion graph generation unit 209 searches the generated inclusion graph for a graph in which the slave node overlaps, and sets the inclusion node to n. Store (step S207). Next, the inclusion graph generation unit 209 adds a node group that has not been added to the nodes in the inclusion graph starting from the inclusion node n (step S208). Then, the inclusion graph generation unit 209 sets the inclusion direction from the inclusion node's inclusion direction or the list of item names (step S209).

  On the other hand, when there is no divided node in inc (step S206, false), the inclusion graph generation unit 209 generates a new inclusion graph (step S210). Then, the inclusion graph generation unit 209 sets the inclusion direction from the direction included in the main node or the list of item names (step S211), and adds a new inclusion graph to the inclusion graph set IG (step S212). The inclusion graph generation unit 209 outputs the inclusion graph set IG after processing all i (Step S214).

  In this way, the inclusion graph generation unit 209 preferentially processes a node having a large number of included nodes among nodes having an inclusion relationship. The inclusion graph generation unit 209 first acquires all inclusion relationships in the style graph and searches for nodes in the inclusion direction. For example, in the example of FIG. 7, an inclusion graph as shown in FIG. 12 is generated. In the example of FIG. 8, since g1 is not included in g6, an inclusion graph starting from g1 is generated separately from the inclusion graph starting from g6.

  Next, processing of the item name synthesizing unit 210 will be described with reference to FIG. FIG. 27 is a flowchart showing the flow of processing of the item name synthesizing unit. As shown in FIG. 27, first, the item name synthesizing unit 210 reads the style graph, the inclusion graph 302, the enumeration list 301, and the start node group (step S301). Next, the item name synthesizing unit 210 generates a new tree structure T (step S302). The starting point node group is a set of nodes that are the starting points of the inclusion graph.

  Here, the inter-item name composition unit 210 performs the following processing for each start point node included in the start point node group (steps S303 and S319). First, the item name synthesizing unit 210 obtains a node group whose node is an item name from the inclusion graph, and stores it in the children (step S304). Next, the item name synthesizing unit 210 generates a new tree node t for the start point node having children as children and no parents (step S305). Then, the item name synthesizing unit 210 sets the target tree node to t (step S306). Then, the item name synthesizing unit 210 adds to the tree structure without setting the type of the target tree node (step S307).

  Here, when the target tree node is in the enumeration list (step S308, false), the inter-name-name synthesis unit 210 proceeds to processing of the next start point node. If the target tree node is not in the enumeration list (step S308, true) and there is a node set that the target tree node includes in the horizontal direction (step S309, true), the inter-item name composition unit 210 Then, the type of the target tree node is set to inclusion (step S310). Then, the item name synthesizing unit 210 sets the next node in the horizontally included node set as the target node, returns to step S307, and performs recursive processing (steps S311, S312, and S313).

  When the target tree node is not in the enumeration list (step S308, true), there is no node set that the target tree node includes in the horizontal direction (step S309, false), and the target tree node is in the vertical direction. If there is an included node set (step S314, true), the item name synthesizing unit 210 sets the type of the target tree node to include (step S315). Then, the item name synthesizing unit 210 sets the next node in the horizontally included node set as the target node, returns to step S307, and performs recursive processing (steps S316, S317, and S318). The item name synthesizing unit 210 performs processing on all the start point nodes, and then outputs tree structure data (step S320).

  When the target tree node is not in the enumeration list (step S308, true), there is no node set that the target tree node includes in the horizontal direction (step S309, false), and the target tree node is vertical. If there is no node set included in the direction (step S314, false), the inter-name-name combining unit 210 proceeds to the processing of the next start point node.

  Next, processing of the enumeration synthesis unit 211 will be described with reference to FIG. As shown in FIG. 28, the enumeration synthesis unit 211 first reads the style graph, the enumeration list 301, and the tree structure data (step S341).

  The enumeration synthesis unit 211 performs the following processing for each node in the enumeration list 301 (steps S342 and S347). First, the enumeration synthesis unit 211 generates a new tree node (step S343). Next, the enumeration composition unit 211 sets the children of the tree node as children of the enumeration list (step S344). Then, the enumeration synthesis unit 211 synthesizes the tree node with the tree structure data (step S346). The enumeration synthesis unit 211 outputs the tree structure data after performing the above processing for all the nodes (step S348). For example, in the example of FIG. 7, tree structure data as shown in FIG. 13 is output.

  Next, processing of the adding unit 212 will be described with reference to FIG. FIG. 29 is a flowchart showing the flow of processing of the adding unit. As shown in FIG. 29, the adding unit 212 first reads the style graph and the tree structure data, stores the tree structure data in S, and stores the style graph in G (step S351). Next, the adding unit 212 stores a parentless node set in nds, and stores an arbitrary character string in str (step S352). Here, if the character string outside the ruled line frame in the left or upward direction of nds can be acquired, the adding unit 212 stores the acquired character string in str (step S353).

  When a character string outside the ruled line frame is not found (step S354, false), the adding unit 212 has an arbitrary character string stored in str, and uses a tree node whose type is style as tree structure data as a parent of nds. (Step S355). In addition, when a character string outside the ruled line frame is found (step S354, true), the adding unit 212 has an arbitrary character string stored in str, and uses a tree node whose type is a style as a parent of nds. It adds to data (step S356). Finally, the adding unit 212 outputs tree structure data (step S357).

  In addition, as an arbitrary character string, the character string may be changed in accordance with the count as in form # (i). In this case, since the part (i) changes in accordance with the count, the character string of the root node becomes “form1”, “form2”. Further, if there is a description such as “XX orthogonal table” outside the left or upper ruled line frame of the table, the character string of the root node may be set as “XX orthogonal table”.

[Effect of the first embodiment]
The extraction unit 201 represents information related to the area representing the item name or item value of the form as a node, and based on the graph representing the adjacency relationship between the nodes as an edge, the node satisfying a preset condition is selected. , It is extracted as an item name node which is a node of the area representing the item name. In addition, when a plurality of nodes are adjacent to each other in a predetermined direction of one node, the first deletion unit 203 deletes an edge representing an adjacency relationship between the one node and the plurality of nodes. In addition, the second deletion unit 204 determines the adjacency relationship between the item name node and the item value node adjacent to the item value node that is the node of the area representing the item value in a predetermined direction among the item name nodes. Delete the representing edge.

  The vertical enumeration acquisition unit 206 and the horizontal enumeration acquisition unit 207 are arranged between the item name node and the item value node based on the graph in which the edge is deleted by the first deletion unit 203 and the second deletion unit 204. Get logical relationship. The inclusion relationship acquisition unit 208 acquires the inclusion relationship between the item name nodes based on the graph in which the edge is deleted by the first deletion unit 203. The item name synthesizing unit 210 and the enumeration synthesizing unit 211 combine the logical relationship acquired by the vertical enumeration acquisition unit 206 and the horizontal enumeration acquisition unit 207 and the inclusion relationship acquired by the inclusion relationship acquisition unit 208. Create tree structure data.

  Therefore, according to the present embodiment, even if the form includes vertical enumeration, horizontal enumeration, a composite type of vertical enumeration and horizontal enumeration, or an enumerated type nested structure, the logical relationship between the item names of the forms, In addition, the logical relationship between the item name and the item value can be accurately recognized. Further, in the present embodiment, since the logical relationship can be acquired semi-automatically, it is possible to efficiently acquire and utilize the semi-structured data of the form.

  In addition, when a plurality of nodes are adjacent to the left side or the upper side of one node, the first deletion unit 203 may delete an edge representing the adjacent relationship between the one node and the plurality of nodes. In addition, the second deletion unit 204 has an adjacency relationship between an item name node and an item value node adjacent to an item value node that is a node of an area representing an item value on the left or upper side of the item name nodes. You may delete the edge showing. In general, the positional relationship between item names in a form and between item names and item values is often left to right or top to bottom. For this reason, by setting the direction of the adjacent relationship to be deleted on the left side and the upper side, it is possible to deal with many forms.

  The inclusion relationship acquisition unit 208 includes at least one of the first node groups adjacent to the right side of the first item name node as the item name node, and heights of all the nodes included in the first node group. Is less than or equal to the height of the first item name node, and the vertex of the upper left node of the first node group and the vertex of the first item name node overlap, the first item name node Are included in the first node group. In addition, the inclusion relationship acquisition unit 208 includes at least one of the second node groups adjacent to the lower side of the second item name node as an item name node, and all nodes included in the second node group The width of the second item name node is less than or equal to the width of the second item name node, and the vertex of the upper left node of the second node group overlaps the vertex of the first item name node, the second item name It is determined that the node includes the second node group. Thus, the inclusion relation can be accurately recognized by utilizing the adjacent relation, height, and width of the nodes.

[Other Embodiments]
The logical relationship recognition target may be a Web screen or a system GUI as long as it can be formatted into a form format. For example, an item name and an item value are acquired from a Web screen on which a ticket is made on the Web as shown in FIG. 30, and the Web screen is subjected to logical relationship recognition processing by formatting it into a form format. be able to. FIG. 30 is a diagram for explaining another embodiment.

[System configuration, etc.]
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. Furthermore, all or a part of each processing function performed in each device may be realized by a CPU and a program that is analyzed and executed by the CPU, or may be realized as hardware by wired logic.

  Also, among the processes described in this embodiment, all or part of the processes described as being performed automatically can be performed manually, or the processes described as being performed manually can be performed. All or a part can be automatically performed by a known method. In addition, the processing procedure, control procedure, specific name, and information including various data and parameters shown in the above-described document and drawings can be arbitrarily changed unless otherwise specified.

[program]
As an embodiment, the logical relationship recognition apparatus 10 can be implemented by installing a logical relationship recognition program for executing the logical relationship recognition as package software or online software on a desired computer. For example, the information processing apparatus can function as the logical relation recognition apparatus 10 by causing the information processing apparatus to execute the above logical relation recognition program. The information processing apparatus referred to here includes a desktop or notebook personal computer. In addition, the information processing apparatus includes mobile communication terminals such as smartphones, mobile phones and PHS (Personal Handyphone System), and slate terminals such as PDA (Personal Digital Assistant).

  The logical relationship recognition apparatus 10 can also be implemented as a logical relationship recognition server device that uses a terminal device used by a user as a client and provides the client with the above-described service related to logical relationship recognition. For example, the logical relationship recognition server device is implemented as a server device that provides a logical relationship recognition service that takes a form as input and outputs tree structure data. In this case, the logical relationship recognition server device may be implemented as a Web server, or may be implemented as a cloud that provides the above-described service relating to logical relationship recognition by outsourcing.

  FIG. 31 is a diagram illustrating an example of a computer in which a logical relationship recognition apparatus is realized by executing a program. The computer 1000 includes a memory 1010 and a CPU 1020, for example. The computer 1000 also includes a hard disk drive interface 1030, a disk drive interface 1040, a serial port interface 1050, a video adapter 1060, and a network interface 1070. These units are connected by a bus 1080.

  The memory 1010 includes a ROM (Read Only Memory) 1011 and a RAM 1012. The ROM 1011 stores a boot program such as BIOS (Basic Input Output System). The hard disk drive interface 1030 is connected to the hard disk drive 1090. The disk drive interface 1040 is connected to the disk drive 1100. For example, a removable storage medium such as a magnetic disk or an optical disk is inserted into the disk drive 1100. The serial port interface 1050 is connected to a mouse 1110 and a keyboard 1120, for example. The video adapter 1060 is connected to the display 1130, for example.

  The hard disk drive 1090 stores, for example, an OS 1091, an application program 1092, a program module 1093, and program data 1094. That is, a program that defines each process of the logical relationship recognition apparatus 10 is implemented as a program module 1093 in which a code executable by a computer is described. The program module 1093 is stored in the hard disk drive 1090, for example. For example, a program module 1093 for executing processing similar to the functional configuration in the logical relationship recognition apparatus 10 is stored in the hard disk drive 1090. Note that the hard disk drive 1090 may be replaced by an SSD.

  The setting data used in the processing of the above-described embodiment is stored as program data 1094 in, for example, the memory 1010 or the hard disk drive 1090. Then, the CPU 1020 reads the program module 1093 and the program data 1094 stored in the memory 1010 and the hard disk drive 1090 to the RAM 1012 as necessary, and executes them.

  The program module 1093 and the program data 1094 are not limited to being stored in the hard disk drive 1090, but may be stored in, for example, a removable storage medium and read out by the CPU 1020 via the disk drive 1100 or the like. Alternatively, the program module 1093 and the program data 1094 may be stored in another computer connected via a network (LAN, WAN (Wide Area Network), etc.). The program module 1093 and the program data 1094 may be read by the CPU 1020 from another computer via the network interface 1070.

DESCRIPTION OF SYMBOLS 10 Logical relationship recognition apparatus 20 Control part 30 Memory | storage part 201 Extraction part 202 Analysis part 203 1st deletion part 204 2nd deletion part 205 Classification part 206 Vertical enumeration acquisition part 207 Horizontal enumeration acquisition part 208 Inclusion relation acquisition part 209 Inclusion graph Generation unit 210 Inter-name-name synthesis unit 211 Enumeration synthesis unit 212 Addition unit 301 Enumeration list 302 Inclusion graph

Claims (5)

Based on a graph that represents information related to the area representing the item name or item value of the form as a node and the adjacent relationship between the nodes as an edge, a node that satisfies a preset condition is selected as the item name. An extraction unit that extracts as an item name node that is a node of an area that represents
A first deletion unit that deletes an edge representing an adjacency relationship between the one node and the plurality of nodes when a plurality of nodes are adjacent to each other in a predetermined direction of the one node;
Among the item name nodes, an item name node that is adjacent to an item value node that is a node of an area that represents an item value in a predetermined direction, and an edge that represents an adjacency relationship between the item value node are deleted. Delete part,
A first acquisition unit that acquires a logical relationship between the item name node and the item value node based on the graph in which an edge is deleted by the first deletion unit and the second deletion unit. When,
A second acquisition unit that acquires an inclusion relationship between the item name nodes based on the graph in which an edge is deleted by the first deletion unit;
A synthesis unit that creates data of a tree structure obtained by synthesizing the logical relationship acquired by the first acquisition unit and the inclusion relationship acquired by the second acquisition unit;
A logical relationship recognition apparatus comprising: The first deletion unit, when a plurality of nodes are adjacent to the left or upper side of one node, deletes an edge representing an adjacency relationship between the one node and the plurality of nodes;
The second deletion unit includes an item name node adjacent to an item value node that is a node of an area representing an item value on the left or upper side of the item name nodes, and an adjacency relationship between the item value nodes. The logical relationship recognition apparatus according to claim 1, wherein an edge representing the data is deleted. In the second acquisition unit, at least one of the first node groups adjacent to the right side of the first item name node is the item name node, and all the nodes included in the first node group The height of the first item name node is equal to or less than the height of the first item name node, and the vertex of the upper left node of the first node group overlaps the vertex of the first item name node, It is determined that the first item name node includes the first node group, and at least one of the second node groups adjacent to the lower side of the second item name node is the item name node. And the width of all the nodes included in the second node group is equal to or less than the width of the second item name node, and the vertex of the upper left node of the second node group, and If the vertex of the second item name nodes overlap, the second term Logical relationship recognition apparatus according to claim 2 in which the name node and determines that encompasses the second node group. A logical relationship recognition method executed by a logical relationship recognition device,
Based on a graph that represents information related to the area representing the item name or item value of the form as a node and the adjacent relationship between the nodes as an edge, a node that satisfies a preset condition is selected as the item name. An extraction step of extracting as an item name node that is a node of an area representing
When a plurality of nodes are adjacent to each other in a predetermined direction of one node, a first deletion step of deleting an edge representing an adjacent relationship between the one node and the plurality of nodes;
A second acquisition step of acquiring an inclusion relationship between the item name nodes based on the graph in which an edge is deleted by the first deletion step;
Among the item name nodes, an item name node that is adjacent to an item value node that is a node of an area that represents an item value in a predetermined direction, and an edge that represents an adjacency relationship between the item value node are deleted. Delete process,
A first acquisition step of acquiring a logical relationship between the item name node and the item value node based on the graph in which an edge is deleted by the first deletion step and the second deletion step. When,
A synthesis step of creating data of a tree structure obtained by synthesizing the logical relationship acquired by the first acquisition step and the inclusion relationship acquired by the second acquisition step;
The logical relationship recognition method characterized by including. On the computer,
Based on a graph that represents information related to the area representing the item name or item value of the form as a node and the adjacent relationship between the nodes as an edge, a node that satisfies a preset condition is selected as the item name. An extraction step of extracting as an item name node that is a node of an area representing
A first deletion step of deleting an edge representing an adjacency relationship between the one node and the plurality of nodes when a plurality of nodes are adjacent to each other in a predetermined direction of the one node;
A second acquisition step of acquiring an inclusion relationship between the item name nodes based on the graph in which an edge is deleted by the first deletion step;
Among the item name nodes, an item name node that is adjacent to an item value node that is a node of an area that represents an item value in a predetermined direction, and an edge that represents an adjacency relationship between the item value node are deleted. Delete step,
A first acquisition step of acquiring a logical relationship between the item name node and the item value node based on the graph in which an edge is deleted by the first deletion step and the second deletion step. When,
A synthesis step of creating data of a tree structure obtained by synthesizing the logical relationship acquired by the first acquisition step and the inclusion relationship acquired by the second acquisition step;
A logical relationship recognition program characterized in that

Images