JP2024001395A - Logic creation aiding system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a logic creation aiding system that reduces burden on a user who is involved in creation of logics related table reading.

SOLUTION: A logic creation aiding system for recognition of logical structure of a document is connected to an input/output device. A logic flow management unit manages logics and logic flows of structure element recognition and logical structure recognition. The structure element recognition receives a structure element recognition logic designated by a user at the logic flow management unit from an input device and outputs it. The logical structure recognition receives the output and a logical structure recognition logic designated by the user at the logic flow management unit and outputs them. The logic flow management unit reflects a logic visualization and logic application result in an interface.

SELECTED DRAWING: Figure 1

COPYRIGHT: (C)2024,JPO&INPIT

Description

The present invention relates to a logic creation support system for an information acquisition device.

Tables are included in documents such as forms and reports, and the elements (components) that make up the table, such as items and numbers, are visually arranged and arranged, and the logical structure of the table elements (logical structure) is clearly visible to the reader. It is a means of expressing information. Tables exist in a variety of media, such as documents, images, and web pages, and are universally used. By mechanically reading the logical structure of tables from such tables, storing the read results logically, and utilizing them in information extraction systems and information retrieval systems, it is possible to improve the efficiency of business systems and information acquisition. is expected. Therefore, automatic recognition of the logical structure of tables is an important technology.

To obtain the information contained in a table from an input document such as a form or image, first extract the table, then recognize its constituent elements through character recognition and layout recognition, and finally identify the logical relationships between the table elements. Recognize logical structures based on gender. Since the acquired logical structure can be stored in a normalized logical database such as a relational database, it can be easily utilized in an information extraction system or an information retrieval system. For example, from a table showing the basic characteristics of a certain industrial part, items related to the basic characteristics of that part and their numerical values are recognized as constituent elements, and a logical relationship is established between the items related to the basic characteristics and the numerical values. By recognizing that there is a logical structure, it is possible to store the correspondence between basic characteristics and their values in a relational database.

In order to automatically read a table, it is usually assumed that the format of the table to be read is determined in advance, and the shape of the table to be read, the position of cells (minimum range surrounded by ruled lines) that make up the components, and Decide on the position of the character lines, etc., and read the table according to a predetermined procedure. For example, if we consider reading table information from an image of a receipt used in internal business, the structure of the table of the receipt is known in advance in internal business, so It is possible to recognize numerical values from the coordinates where a character line is written, and estimate its constituent elements and logical structure. If the format of the table you want to read is fixed, reading the table according to a predetermined procedure like this will provide sufficient accuracy. Even if the table format were to be slightly modified in the future, it would be possible to do so with realistic time and cost by partially modifying the logic involved in reading the table.

On the other hand, there is a need to automatically read various non-standard tables and obtain information. There are various types of non-standard tables, such as balance sheets, statements, and doctor's certificates, and obtaining information from these tables is expected to improve work efficiency. However, the format of such a non-standard table can easily change depending on the system of the institution issuing the document and the vendor. For example, like a balance sheet, the balance of income and expenditure is divided into sections such as ``assets'' and ``liabilities'' and the format of the table is determined, and monetary values are reported for each item. Balance sheets have different formats depending on the company.For example, assets may be written on the left half of the table, liabilities may be written on the right half of the table, or assets may be written on the top of the table. The liability section may be listed at the bottom of the table. In addition, for example, regarding the asset section, there are cases where the character lines that make up the asset section are visually represented by indentation. When it comes to doctors' medical certificates, the format of the table that is desired to be read differs depending on the hospital, making it difficult to handle the process based on a predetermined table format.

Japanese Patent Application Publication No. 2009-122722

To build an information acquisition system that automatically reads tables and acquires the logical structure of the various types of non-standard tables explained above, it is necessary to construct an information acquisition system that is specific to each table format. It is necessary to construct a logic that recognizes the logical structure. When table formats change frequently or when a completely new table format must be supported, the logic used to read the table and the flow of applying logic that combines multiple of the above logics (logic flow) must be constructed step by step. There is a need to. Examples of logic include rules and algorithms for recognizing table elements, and rules and algorithms for recognizing table structures. As a logic flow, for example, there are two logics for recognizing a logical structure, and the application order of the two logics may be determined. Such logic and logic flows require specialized knowledge and know-how regarding table structures, and the logic itself includes complex program processing, so it is difficult for developers (users) to construct logic and logic flows. The burden is heavy.

Patent Document 1 discloses a method for reducing the burden on a correction worker when correcting recognized form data in a system that recognizes the logical structure of a form and stores the logical information in a database. However, Patent Document 1 corrects the database based on a user's correction instruction on an interface (I/F), and cannot support the creation of complex recognition logic.

Therefore, in order to support the creation of logic and logic flows and reduce the burden on users involved in reading tables, we prepare logic templates in advance on the interface, rather than giving correction instructions, and provide advice on how to combine logic and logic flow. Consider reducing the burden on the user by adjusting the parameters. This method requires the user to construct the logic flow, but templates for each logic are provided in advance, so there is no need to write your own programs or think of your own logic unless necessary. Therefore, it is expected that the burden on the user will be reduced.

On the other hand, as mentioned above, the logic involved in table reading involves complex program processing, so even if a logic template is prepared, it is difficult to understand how each logic is executed. This is difficult for the user to understand intuitively, which causes a burden on the user. For example, in the logic for determining components, the component letters are Rules using row coordinates and dictionaries are implemented. At this time, the user may, for example, enter the parameters necessary to implement the rule, such as how much coordinate shift is allowed, or a dictionary consisting of character strings that may be included in the head, side, and values. Specify a regular expression. Processing using early parameters and dictionaries requires a user to make appropriate decisions only after concretely grasping the content of the logic and its results, which places a high burden on the user. Furthermore, the burden on the user cannot be resolved simply by displaying the logical structure based on the logic recognition result on the interface.

Therefore, when a user constructs logic and logic flows on an interface, there are two things to consider: what kind of processing each logic performs, and how the logical structure is recognized by performing the processing. By presenting information to the user on the interface based on the viewpoint, the burden on the user involved in constructing logic and logic flows can be significantly reduced. Furthermore, the information necessary to create the logic is reflected in the interface sequentially, making it possible to intuitively modify the logic specified by the user and the parameters related to the logic.

In order to solve the above-mentioned problems, a logic creation support system for recognizing the logical structure of documents in an information acquisition device, which is one aspect of the invention disclosed in this application, is connected to an input/output device, and a logic flow management unit is configured to It manages the logic and logic flow of recognition and logical structure recognition, and for component recognition, the logic flow management section accepts and outputs the component recognition logic specified by the user from the input device, and for logical structure recognition, the logic and logic flow are A flow management unit receives and outputs logical structure recognition logic specified by a user from an input device, and the logic flow management unit includes a logic creation support system configured to reflect logic visualization and logic application results on an interface, and an information acquisition program. , is a logic creation support system for logical structure recognition of documents, which is connected to an input/output device, and a logic flow management unit manages the logic and logic flow of component recognition and logical structure recognition. A flow management unit receives and outputs a component recognition logic specified by a user from an input device, and for logical structure recognition, the logic flow management unit receives and outputs the output and a logical structure recognition logic specified by a user from an input device, The logic flow management unit reflects the logic visualization and logic application results on the interface.

According to the present invention, the burden on the user involved in creating logic related to table reading is reduced.

FIG. 2 is a diagram illustrating an example of functional blocks of an information acquisition device in each embodiment. FIG. 2 is an explanatory diagram showing Table A that is input in each example. FIG. 7 is an explanatory diagram showing Table B, which is input in each example. FIG. 2 is an explanatory diagram showing an example of an interface of a logic flow management unit according to the first embodiment. FIG. 7 is an explanatory diagram showing an example of an interface of a logic flow management unit according to the second embodiment. FIG. 7 is an explanatory diagram showing an example of an interface of a logic flow management unit according to the third embodiment. FIG. 7 is an explanatory diagram showing an example of an interface of a logic flow management unit according to the third embodiment. FIG. 12 is an explanatory diagram showing an example of an interface of a logic flow management unit according to a fourth embodiment. FIG. 12 is an explanatory diagram showing an example of an interface of a logic flow management unit according to the fifth embodiment. FIG. 12 is an explanatory diagram showing an example of an interface of a logic flow management unit according to a sixth embodiment. FIG. 12 is an explanatory diagram showing an example of an interface of a logic flow management unit according to a seventh embodiment. 1 is a diagram illustrating an example of a hardware configuration of a computer that implements an information acquisition device according to each embodiment; FIG.

A mode for carrying out the present invention will be explained using diagrams. In the following, each embodiment and each modification can be combined in part or in whole without departing from the spirit of the present invention.

In the following explanation, the information acquisition device is a logic creation support system for recognizing the logical structure of documents, and is connected to the input/output device, and the logic flow management section is responsible for managing the logic and logic flow for component recognition and logical structure recognition. For component recognition, the logic flow management section accepts and outputs the component recognition logic specified by the user from the input device, and for logical structure recognition, the logic flow management section receives and outputs the component recognition logic specified by the user, and for logical structure recognition, the logical structure recognition specified by the user in the logic flow management section The logic flow management unit receives and outputs logic from an input device, and reflects the logic visualization and logic application results on the interface, thereby reducing the burden on the user involved in logic creation.
<Configuration of information acquisition device 100 of Example 1>
FIG. 1 shows an information acquisition device 100 according to a first embodiment. The information acquisition device 100 of the first embodiment includes an input unit 101, an input table 101a, a component recognition logic 101b, a logical structure recognition logic 101c, a logic flow management unit 102, a component recognition logic execution unit 102-1, a logical structure recognition It includes a logic execution unit 102-2, a logic visualization I/F update unit 102-3, a logic application result I/F update unit 102-4, an output unit 103, and a completed logic 103a.

In this embodiment, an example of the interface of the logic flow management unit in the information acquisition device 100 will be shown, and a mechanism for reducing the load of logic creation related to table reading by the user will be explained. According to this example, when the user instructs the logic on the interface and constructs the logic flow, it is possible to understand what kind of processing each logic performs and how the logical structure of the table is recognized by each logic. It is possible to show the user whether the This embodiment can also be applied to various applications such as developing logic related to table reading, modifying existing logic, and incorporating it into a table reading system.

Note that in the following embodiment, Table A in FIG. 2 and Table B in FIG. 3 are used as input examples for explanation. Table A is an example of a balance sheet, including "Current assets," "1. Cash and deposits," "2. Prepaid expenses," "Fixed assets," "1. Tangible fixed assets," and "2. Intangible fixed assets." It consists of components such as "asset", "amount", "3,000", "1,000", and "500". Components are granular table elements that are generally used to recognize logical structures, and can be thought of as the arrangement of character strings (character lines) that are the components of a table, or rectangles (cells) surrounded by table borders. It will be done. In Table A, there is a text line containing, for example, "current assets", and a rectangle containing a text line containing, for example, "amount" as a cell. Table B is an example of a table regarding basic characteristics of industrial parts. Table B similarly includes character rows and cells. In the following embodiment, the aim is to read Table A and Table B, recognize logical relationships using a logic flow, and convert them into Logical Structure A and Logical Structure B. In order to obtain logical structure A and logical structure B, it is necessary to design component recognition logic and logical structure recognition logic specialized for the formats of table A and table B. The design of the recognition logic of the components and the recognition logic of the logical structure, and part of the mechanism for reducing the burden on the user will be clarified in the following examples.

Note that Tables A and B are merely table formats used for explanation, and the tables may be in any format. In this case, various types of forms such as medical certificates and detailed statements, and various table formats such as tables without ruled lines, tables within tables, and illogical tables may be considered.

Furthermore, although the following embodiments will mainly be explained using a GUI (graphical interface) as an example, the present invention is not limited to the GUI. Furthermore, the configuration, arrangement, and functions of the GUI are merely examples for explaining the embodiments below, and parts or all of them can be added, modified, deleted, and combined without departing from the spirit of the present invention. shall be taken as a thing.
<Processing of information acquisition device 100 of Example 1>
The input unit 101 receives the input table 101a, the component recognition logic 101b, and the logical structure recognition logic 101c as input, and outputs them to the logic flow management unit 102.

In this embodiment, the input table 101a will be explained assuming that it is a table included in a document image, but the input table 101a can accept various types of input formats such as databases, documents, drawings, etc. At this time, for example, the document may be an electronic document or a physically recorded document such as paper or board. At this time, the drawing may be an electronic drawing or a physically recorded drawing such as paper or board. Furthermore, documents and drawings may contain various information such as character strings, figures, and photographs. In this example, for the sake of explanation, the input table 101a is a table in which the borders of the table are recognized as figures, the coordinates and shapes of the figures are recognized, and the coordinates of character lines and information on character strings included in the table are recognized. It is assumed that the coordinates and shape of the figure, the coordinates of the character line, and the information on the character string are input to the input unit 101. In order to recognize the coordinates of ruled line figures and character lines, for example, software and algorithms that perform ruled line extraction and layout analysis through image processing can be used. In order to recognize a character string included in a character line, for example, an algorithm or software such as handwritten character recognition can be used. Although this embodiment will be explained assuming that the character string is Japanese, it may be in any language such as Chinese or English. Furthermore, the character string may include not only language but also figures and symbols.

The component recognition logic 101b is composed of algorithms and rules for recognizing components such as character lines and cells included in a table. At this time, there may be multiple logics for recognizing the constituent elements. For example, two logics, one for recognizing character lines and the other for recognizing cells, can constitute the component recognition logic 101b. Furthermore, you can also specify the order and combination of logic applications.

For Table A, the components recognized by the component recognition logic 101b include, for example, character lines such as "current assets," "1. Cash and deposits," "2. Prepaid expenses," and "fixed assets." , "1. Tangible fixed assets", "2. Intangible fixed assets", "3,000", "1,000", and "500" are considered, and there is a rectangle containing the character line of "amount" as a cell. . At this time, the reason why the component is a cell with "amount" is that the text line contained in the cell is only "amount", and the width and height of the cell are determined based on the visual sense that "amount" is at the top. This is because it shows specific information. On the other hand, text lines are appropriate for constituent elements such as "current assets" and "3,000." This is because the cell containing "current assets" contains many character lines, and if all of these lines were combined into constituent elements, it would not be possible to obtain the correct logical structure. Whether a component is a character line or a cell depends on the format of the input document and table, and the component is not limited to character lines and cells, but is suitable for estimating logical structure. Any grain size is acceptable as long as it is suitable.

The component recognition logic 101b also includes logic for classifying components into front, front side, and value. For example, in Table A, "1. Cash and deposits" is classified as the front side, and "3,000" is classified as the value. Similarly, when applied to all components of Table A, the results are shown in Table A-1. In this way, by classifying each component into head, front side, and value, it becomes easier to estimate the logical structure. This is because when estimating the logical structure, it is only necessary to consider the logical structure between the head of the table and the logical structure of the front side, the logical structure of the front side and the front side, the logical structure of the head and the value, and the logical structure of the front side and the value. This is because the search range necessary for estimating the structure can be limited. For example, in Table A, the logical structure of "1. Cash and deposits" and "3,000" is recognized from the combination of the relationship between the front side and the value, and "Amount" and "3,000" are the relationship between the front side and the value. Since it is sufficient to recognize the logical structure from the combinations of , there is no need to search for all combinations of constituent elements. Furthermore, classifying components into front, front, and value means that, for example, in the process of converting table A to logical structure A, the "item 1" and "item 2" columns of logical structure A are recognized as front. This is equivalent to recognizing the "amount" of logical structure A as the head of the table, and identifying other cells containing only numerical values of logical structure A as values.

The logical structure recognition logic 101c is composed of algorithms and rules for recognizing a logical structure that is a logical relationship between constituent elements. At this time, there may be multiple logics that recognize the logical structure. For example, two pieces of logic can constitute the logical structure recognition logic 101c: a logic that recognizes the logical structure of the front side and the value, and a logic that recognizes the logical structure of the front side and the value. Furthermore, you can also specify the order and combination of logic applications.

The logic flow management unit 102 receives the input table 101a, component recognition logic 101b, and logical structure recognition logic 101c as input from the input unit 101, manages and reflects the logic, and updates an interface that supports the user's logic creation. . The logic flow management unit 102 first checks whether the logic is specified in the component recognition logic 101b, and if the logic is specified, executes the logic in the component recognition logic execution unit 102-1. , the logic visualization I/F update unit 102-3 visualizes the processing content of the logic on an interface and presents it to the user, and the logic application result I/F update unit visualizes the application result of the logic on the interface and presents it to the user. present. If the processing from the component recognition logic execution unit 102-1 to the logic application result I/F update unit 102-4 is completed and the next logic to be executed is specified in the component recognition logic 101b, the next logic is executed again. Processing is performed for the logic to be executed from the component recognition logic execution unit 102-1 to the logic application result I/F update unit 102-4.

The logical structure recognition logic execution unit 102-2 is executed in the logic flow management unit 102 when the user does not designate the component recognition logic or when the processing of the component recognition logic is completed. At this time, the logic flow management unit 102 first checks whether the logic is specified in the logical structure recognition logic 101c, and if the logic is specified, the logic flow management unit 102 executes the logic in the logical structure recognition logic execution unit 102-2. is executed, the logic visualization I/F update unit 102-3 visualizes the processing content of the logic on an interface and presents it to the user, and the logic application result I/F update unit visualizes the application result of the logic on the interface. and present it to the user. If the processing from the logical structure recognition logic execution unit 102-2 to the logic application result I/F update unit 102-4 is completed and the next logic to be executed is specified in the component recognition logic 101b, the next logic is executed again. Processing is performed for the logic to be executed from the component recognition logic execution unit 102-1 to the logic application result I/F update unit 102-4.

In FIG. 1, the logic flow management unit 102 is merely a block diagram showing an example of processing, and the configuration of the blocks can be changed without departing from the spirit of the present invention. At this time, for example, the processing order of the logic visualization I/F update unit 102-3 and the logic application result I/F update unit 102-4 may be changed. Besides, the method of specifying the logic by the user may be in any form. At this time, for example, it is conceivable that the logic specification is received from the user via input to a GUI (graphical user interface). Further, the logic flow management unit 102 may be partially or entirely configured as an interface as long as it can be appropriately presented to the user.

FIG. 4 is an explanatory diagram showing an example of implementing the logic flow management unit 102 using a GUI. In FIG. 4, the logic flow management unit 102 is implemented in a GUI window. Further, the GUI window may have functions necessary for GUI operation, such as a window minimize button C1 and a window close button C2. The logic processing contents visualized by the logic visualization I/F update unit 102-3 are reflected one by one in the panel box C4. The logic application results visualized by the logic application result I/F update unit 102-4 are reflected one by one in the panel box C5.

Further, on the GUI of the logic flow management unit 102 in FIG. 4, addition, deletion, or editing of component recognition logic and logical structure recognition logic can be performed through user instructions such as button operations. For example, C6 is a panel box that provides an I/F for users to edit logic and logic flows. C62 is a button that allows the user to insert new logic, and by clicking the button, the user can insert pre-prepared logic or user-definable logic into the panel box C6 as a logic window. It can be inserted inside. In FIG. 4, the logic window is, for example, C63. The C62 also suggests to the user where to insert the logic window based on the button position, encouraging intuitive logic creation. Also, inside logic windows that have already been inserted or will be inserted, adjustable programs of logic, parameters and/or dictionaries are presented as buttons, text boxes and input boxes. For example, C63 is an example of a logic window for a certain recognition logic, and the name of the logic is displayed in the title C65. Furthermore, if the user is editing the logic, the width of the border of the logic window can be changed and presented to the user. Furthermore, if the user wants to delete the logic, this can be achieved by clicking the close button C64. Furthermore, panel box C6 allows users to add any number of logic windows and combine them in any way. At this time, for example, the display position of the logic window can be adjusted using the scroll bar C61.

The GUI of the logic flow management unit 102 in FIG. 4 is merely an example for explanation, and the configuration, arrangement, shape, etc. of the GUI may be in any manner. Also, it may be implemented using various interfaces instead of GUI. At this time, for example, it is possible to use a character interface (CUI). Further, part of the interface may not be implemented, and various other functions necessary for logic creation support may be implemented. Voice and gestures may be used as an interface for presenting or instructing user input and output.
<Effects of Example 1>
According to this embodiment, the logic flow management section receives and outputs the logic specified by the user from the input device, and the logic flow management section visualizes the logic information using the logic visualization I/F, and the result obtained by applying the logic is is reflected in the logic application result I/F, reducing the burden on users involved in logic creation.
<Configuration of information acquisition device 100 of Example 2>
In addition to the configuration of the information acquisition device 100, the information acquisition device of the second embodiment includes a logic window for specifying components.

In the second embodiment, in the information acquisition device 100, the logic necessary for identifying a component is displayed on the logic visualization I/F C4, and the application result of the logic is displayed on the application result I/F C5. Note that for parts that overlap with the explanation in Example 1, part or all of the explanation will be omitted. Furthermore, in the first embodiment, parts of the logic flow management unit 102 shown in FIG. 4 that overlap with the GUI will be described with some or all of the drawings omitted.
<Processing of information acquisition device 100 of Example 2>
FIG. 2 shows a GUI of the logic flow management unit 102 of the information acquisition device 100 in the second embodiment. In FIG. 2, Table A is used as an input table for explanation, but the invention is not limited to Table A and can be applied to tables other than Table A as well. C6 includes the logic window D3 being edited in the logic flow editing I/F.

In logic window D3, the user sets parameters regarding the specific logic of the component in the text box. At this time, for example, as shown in D4, if you want only the cell to be a component, you can enter "Yes", otherwise you can enter "No". At this time, D4 can prompt the user for input using various methods such as drop-down lists, check boxes, and dialog displays as a GUI. In addition, D5 allows the user to specify the maximum number of character lines that can be included in a cell as a boundary condition when using a cell as a component. In D5, for example, when the user inputs "1", the cell "Amount" in Table 4 has 1 character line, so the boundary condition is satisfied and the cell is considered to be a component.

To adjust the parameters in the logic window D3, one must have knowledge of the logic being edited. Therefore, by presenting information to the user such as how many character lines are included in a cell, the burden on the user can be reduced. Therefore, regarding the parameters of the logic window D3 being edited, the logic flow management unit 102 executes the component recognition logic execution unit 102-1 and the logic visualization I/F update unit 102-3, and the logic visualization I/F update unit 102 -3 uses logic visualization I/F C4 to display the number of character lines included in each cell on the GUI, as shown in D1, and intuitively provide the user with the knowledge necessary to adjust the parameters of the logic. to be presented. D1 can have various other forms, such as using a speech bubble or displaying a straight line connecting cells, but any interface may be used as long as it can be presented appropriately to the user.

In the logic window D3, when D4 or D5 receives input from the user, the component recognition logic execution unit 102-1, the logic visualization I/F update unit 102-3, and the logic application result I/F update unit 102-4 The logic application result I/F update unit 102-4 updates the logic application result I/F C5. For example, in FIG. 5, in C5, the result of identifying the constituent elements is presented to the user by overlapping bounding boxes or rectangles according to the shapes of the constituent elements. At this time, for example, as shown in D2, a component identified as a cell can be overlaid with a translucent rectangle to fit the entire cell. In addition, for example, for a component identified as a character line, such as D3, a translucent rectangle can be superimposed on the component to fit the entire character line.
<Effects of Example 2>
According to this embodiment, the logic flow management unit receives and outputs the logic specified by the user from the input device, and the logic flow management unit transmits information necessary for parameter adjustment of the component recognition logic using the logic visualization I/F. The results of applying component recognition logic are visualized and reflected in the logic application result I/F, reducing the burden on users involved in logic creation.
<Configuration of information acquisition device 100 of Example 3>
The information acquisition device of the third embodiment includes, in addition to the configuration of the information acquisition device 100, an interface for classifying the top, front side, or value of the component.

In the third embodiment, in the information acquisition device 100, the logic necessary for specifying a component is displayed on the logic visualization I/F C4, and the application result of the logic is displayed on the application result I/F C5. Note that for parts that overlap with the explanation in Example 1, part or all of the explanation will be omitted. Furthermore, in the first embodiment, parts of the logic flow management unit 102 shown in FIG. 4 that overlap with the GUI will be described with some or all of the drawings omitted.
<Processing of the information acquisition device 100 of Example 3>
FIG. 6 shows an explanatory diagram of an interface for classifying the front side of the constituent elements of the logic flow management unit 102 of the information acquisition device 100 in the third embodiment. Further, in FIG. 6, the logic flow management unit 102 includes a logic window E6 that supports the user in creating front side recognition logic. Further, in FIG. 6, Table A is used as an input table for explanation, but the invention is not limited to Table A and can be applied to tables other than Table A.

At E6, the user can input on E7 the minimum and maximum values of the alignment range when analyzing the front side. For example, for the input value "-1 to 1" of E7, the minimum value is set to -1 and the maximum value is set to 1 when analyzing the front side in the vertical direction. Note that the threshold value may be in any unit. At this time, it is possible to use pixels or millimeters, for example. In addition, to analyze the front side in the vertical direction, for example, obtain the X coordinate of the character line for all components, and group the components whose X coordinate is within the range of the threshold value. can. For example, in Table A, "current assets" and "fixed assets" are located at the same group). Similarly, "1. Cash and deposits", "2. Prepaid expenses", "1. Tangible fixed assets" and "2. Intangible assets" are also included in the same group (second group).

In E6, when analyzing the front side, the user can specify the dictionary included on the front side as a clue in each group. For example, the input values ".*Asset" and ".*Expense" of E8 specify the clue on the front side as a regular expression. Similarly, a dictionary included in the value can be specified as a clue to exclude parts that are not on the front side. For example, for the input values "\d+$" and "yen$" of E9, the value clue is specified as a regular expression. The regular expressions specified in E7 and E9 calculate the number of occurrences in each group and estimate the plausibility of the front side of each group. For example, the regular expression ".*assets" specified in E8 applies to the character strings of all components included in the first group, so the proportion of the front dictionary applicable to the components included in the group ( (referred to as the front side ratio) is 100%. Similarly, in the second group, the front side ratio is 100%. In addition, for example, the regular expression specified in E9 does not apply to the strings of all the components included in the first group, so the value dictionary applies to the components included in the group. The percentage (referred to as value percentage) is 0%. Similarly, the value ratio of the second group is also 0%. In other words, it can be said that the higher the face ratio and the lower the value ratio, the higher the probability of the face being the face.

In E6, the user can specify the values of the above-mentioned front side ratio and value ratio that can be classified as the front side. For example, the input value "80%" of E10 indicates the minimum value of the front side ratio, and the input value "20%" of E11 indicates the allowable value of the value ratio. That is, according to the input values of E10 and E11, the first group and the second group, which have a front side ratio of 80% or more and a value ratio of 20% or less, are determined to be front sides, All components included in the first group and the second group are classified as front side.

Since the dictionary and parameters specified by the user in the logic window E6 require complex logic as described above, it is difficult for the user to understand the influence that the dictionary and parameters have on the logic. Therefore, regarding the dictionary and parameters of the logic window E6 being edited, the logic flow management unit 102 executes the component recognition logic execution unit 102-1 and the logic visualization I/F update unit 102-3 to update the logic visualization I/F. The unit 102-3 uses a logic visualization I/F C4 to intuitively present to the user the knowledge necessary for adjusting logic parameters, as shown in E1. In C4, for example, the front alignment range specified by the user in E7 is visualized with a double-headed arrow symbol, and the length of the symbol indicates the range. Further, as a result of the user specifying the front side alignment range, the group configuration recognized by the front side recognition logic is presented in an arrow shape such as E3, for example. Further, in C4, for example, the front side ratio and value ratio of each group can be presented to the user in letters and rectangles like E2.

When the logic window E6 receives input from the user, it executes the component recognition logic execution unit 102-1, the logic visualization I/F update unit 102-3, and the logic application result I/F update unit 102-4, The logic application result I/F update unit 102-4 updates the logic application result I/F C5. For example, in FIG. 6, C5 presents the result of front side identification to the user by overlapping a bounding box or rectangle according to the shape of the component and changing the pattern or color of the rectangle. At this time, for example, as shown in E5, for the component identified as the front side, a transparent rectangle surrounded by a frame can be superimposed on the component to fit the entire character line. Furthermore, by presenting a different pattern to the rectangles for components other than the front side, the front side and the non-front side can be clearly distinguished and presented.

In FIG. 7, the embodiment of the logic flow management unit 102 in FIG. 6 is applied to the front head instead of the front side, and the front head recognition logic is exemplified at F6. The main difference between the logic flow management unit 102 in FIG. 7 and the logic flow management unit 102 in FIG. 6 is that the analysis direction on the top side in FIG. The actual processing is similar, except that the directions are different, and the dictionaries and parameters used are partially different. Therefore, in this embodiment, the explanation of FIG. 7 will be omitted to avoid redundant explanation.
<Effects of Example 3>
According to this embodiment, the logic flow management unit receives and outputs the logic specified by the user from the input device, and the logic flow management unit transmits information necessary for parameter adjustment of the component recognition logic using the logic visualization I/F. It visualizes and reflects the results of applying front side, front head, and value recognition logic among the component recognition logic on the logic application result I/F, reducing the burden on users involved in logic creation.
<Configuration of information acquisition device 100 of Example 4>
The information acquisition device of the fourth embodiment includes, in addition to the configuration of the information acquisition device 100, an interface for supporting creation of logical structure recognition logic between components included on the front side.

In the fourth embodiment, in the information acquisition device 100, the logic for recognizing the logical structure of the components on the front side is displayed on the logic visualization I/F C4, and the application result of the logic is displayed on the application result I/F C5. indicate. Note that for parts that overlap with the explanation in Example 1, part or all of the explanation will be omitted. Furthermore, in the first embodiment, parts of the logic flow management unit 102 shown in FIG. 4 that overlap with the GUI will be described with some or all of the drawings omitted.
<Processing of the information acquisition device 100 of Example 4>
FIG. 8 shows an explanatory diagram of an interface for recognizing the logical structure of the front side components of the logic flow management unit 102 of the information acquisition device 100 in the fourth embodiment. Further, in FIG. 8, the logic flow management unit 102 includes a logic window F1 that supports the user in creating logic that recognizes the logical structure of the components on the front side. Further, in FIG. 8, Table A is used as an input table for explanation, but the invention is not limited to Table A and can be applied to tables other than Table A.

F1 uses the indent width (deviation width of the X coordinates between the front side components) between the front side components as a clue to identify whether a logical structure exists between the front side components. This is a logic window related to logical structure recognition logic. For example, in Table A, which is a balance sheet, "Current assets" and "1. Cash and deposits" form a logical structure of parent and child by indentation, and in this example, acquiring the logical structure is purpose.

At F1, the user can enter the maximum allowable indent width on F2 and the minimum minimum indent width on F3. For example, an input value of ``20'' for F2 allows an indent width of up to 20 between front side components, and an input value of ``3'' for F3 allows an indentation of at least 3 or more between front side components. do. In this example, if the indent width between "current assets" and "1. Cash and deposits" is 5, the logical structure is satisfied, and "Current assets" and "1. Cash and deposits" are recognized as a logical structure. Ru. Note that the width may be in any unit. In addition, F4 allows the user to specify the maximum downward deviation allowed when providing a logical structure between front side components. The input value of F4 "100" does not add a logical structure when the vertical distance between hierarchical structures is greater than 100. Therefore, it becomes possible for the user to specify the scope of application of the logic.

Since the parameters specified by the user in the logic window F1 require complex logic as described above, it is difficult for the user to understand the influence that the parameters have on the logic. Therefore, regarding the parameters of the logic window F1 being edited, the logic flow management unit 102 executes the logical structure recognition logic execution unit 102-2 and the logic visualization I/F update unit 102-3, and the logic visualization I/F update unit 102 -3 uses logic visualization I/F C4 to intuitively present the knowledge necessary for adjusting logic parameters to the user. In FIG. 8, in C4, for example, F2 informs the user of one of the parent components of the logical structure by a rectangle. F3 and F4 inform the user of the input values for the maximum indent width F2 and minimum indent width F3 in the form of a character string and a rectangle. F6 similarly informs the user of the input value of the maximum downward distance F4. Further, as shown in F5, an arrow symbol can be placed on table A according to the value of the input value, and the valid range of the value on table A can be presented to the user.

When the logic window F1 receives input from the user, it executes the logical structure recognition logic execution unit 102-2, the logic visualization I/F update unit 102-3, and the logic application result I/F update unit 102-4, The logic application result I/F update unit 102-4 updates the logic application result I/F C5. For example, in Figure 8, C5 presents the result of identifying the logical structure between front sides to the user by overlapping bounding boxes and rectangles according to the shapes of the components, overlapping arrow symbols as logical structures. .
<Effects of Example 4>
According to this embodiment, the logic flow management unit receives and outputs the logic specified by the user from the input device, and the logic flow management unit presents rectangles, character strings, and symbols on the input table to indent The information necessary for parameter adjustment of the logical structure recognition logic is visualized using the logic visualization I/F, and the results of applying the logical structure recognition logic are reflected in the logic application result I/F, reducing the burden on users involved in logic creation. Reduce.
<Configuration of information acquisition device 100 of Example 5>
The information acquisition device of the fifth embodiment includes, in addition to the configuration of the information acquisition device 100, an interface for supporting creation of logical structure recognition logic between components included on the front side.

In the fifth embodiment, in the information acquisition device 100, the logic for recognizing the logical structure of the components on the front side is displayed on the logic visualization I/F C4, and the application result of the logic is displayed on the application result I/F C5. indicate. Note that for parts that overlap with the explanation in Example 1, part or all of the explanation will be omitted. Furthermore, in the first embodiment, parts of the logic flow management unit 102 shown in FIG. 4 that overlap with the GUI will be described with some or all of the drawings omitted.
<Processing of information acquisition device 100 of Example 5>
FIG. 9 shows an explanatory diagram of an interface of the logic flow management unit 102 of the information acquisition device 100 according to the fifth embodiment, which recognizes the logical structure between the component that is the front side and the component that is the value. . In addition, in FIG. 9, the logic flow management unit 102 creates a logic window G1 that supports the user's creation of logic that recognizes the logical structure between the component that is the front side and the component that is the value. include. Further, in FIG. 9, Table A is used as an input table for explanation, but the invention is not limited to Table A and can be applied to tables other than Table A.

G1 is a logical structure that uses the distance and shift between the components to identify whether a logical structure exists between the component on the front side and the component that is the value. This is a logic window related to recognition logic. For example, in Table A, which is a balance sheet, "100" existing at almost the same Y coordinate as "2. Prepaid Expenses" forms a logical structure of parent and child, and in this example, the logical structure is The purpose is to obtain.

In G1, the user can specify the range (maximum distance) to search for the logical structure. For example, the maximum distance to the left as shown in G2, the maximum distance to the right as shown in G3, the maximum distance upwards as shown in G4, the maximum distance downwards as shown in G5. can be input by the user. For example, for the input values of G2, G3, G4, and G5, only the component "100" is searched and the logical structure is recognized using a search range centered on the component "2. Prepaid expenses." Note that any method of providing the search range and any search method may be used. In this case, for example, if a plurality of components are searched within the search range, the component closest to the search source component can be recognized as a logical structure.

Since the parameters specified by the user in the logic window G1 require complex logic as described above, it is difficult for the user to understand the influence of the parameters on the logic. Therefore, regarding the parameters of the logic window G1 being edited, the logic flow management unit 102 executes the logical structure recognition logic execution unit 102-2 and the logic visualization I/F update unit 102-3, and the logic visualization I/F update unit 102 -3 uses logic visualization I/F C4 to intuitively present the knowledge necessary for adjusting logic parameters to the user. In FIG. 9, in C4, as in F4 and F5 in the fourth embodiment, the numerical value of the maximum distance is indicated by a character string or a rectangle, and the maximum distance is indicated by an arrow symbol to inform the user. In addition, for example, G5 visually draws the search range as a rectangle to notify the user.

When the logic window G1 receives input from the user, it executes the logical structure recognition logic execution unit 102-2, the logic visualization I/F update unit 102-3, and the logic application result I/F update unit 102-4, The logic application result I/F update unit 102-4 updates the logic application result I/F C5. For example, in Figure 9, C5 presents the result of identifying the logical structure between the front side and the value to the user by overlapping the bounding box and rectangle according to the shape of the component, and overlapping the arrow symbol as a logical structure. .
<Effects of Example 5>
According to this embodiment, the logic flow management unit receives and outputs the logic specified by the user from the input device, and the logic flow management unit configures the configuration by presenting rectangles, character strings, and symbols on the input table. The information necessary for parameter adjustment of logical structure recognition logic using the distance between elements is visualized using logic visualization I/F, and the result of applying logical structure recognition logic is reflected in logic application result I/F to create logic. Reduce the burden on users involved.
<Configuration of information acquisition device 100 of Example 6>
In addition to the configuration of the information acquisition device 100, the information acquisition device of the sixth embodiment includes an interface for supporting the creation of logical structure recognition logic between the components using the shapes of the components as clues.

In the sixth embodiment, the information acquisition device 100 uses the shape of the component as a clue to display logic for recognizing the logical structure of the components on the front side on the logic visualization I/F C4, and applies the logic. Display the results on application result I/F C5. Note that for parts that overlap with the explanation in Example 1, part or all of the explanation will be omitted. Furthermore, in the first embodiment, parts of the logic flow management unit 102 shown in FIG. 4 that overlap with the GUI will be described with some or all of the drawings omitted.
<Processing of information acquisition device 100 of Example 6>
FIG. 10 shows an explanatory diagram of an interface of the logic flow management unit 102 of the information acquisition device 100 in the sixth embodiment, which recognizes the logical structure of the front side components using the shape of the component as a clue. Further, in FIG. 10, the logic flow management unit 102 includes a logic window H1 that supports the user's creation of logic that recognizes the logical structure of the components on the front side. Further, in FIG. 10, Table B is used as an input table for explanation, but the invention is not limited to Table B and can be applied to tables other than Table B.

H1 is a logic window related to logical structure recognition logic that uses the shape of the constituent elements as a clue to identify whether a logical structure exists between the constituent elements on the front side. In this embodiment, it is assumed that in Table B, the "general" and "characteristic X" cells are each recognized as a component. Here, "general" and "characteristic and "Characteristic X" may not be recognized as a logical structure. Therefore, using the shape of the cell as a clue, the logical structure is identified by the extent to which the vertical lengths of the parent cell "General" and the child cell "Characteristics X" overlap when viewed in the horizontal direction. .

In H1, the user can specify the threshold for the overlapping ratio (overlap rate) of parent and child cells and the direction of analysis. For example, as shown in H2, the direction of analysis can be specified as "right" or "down". At this time, the analysis direction is the direction in which the overlapping degree of the child cell with respect to the parent cell is calculated, and if the analysis direction is "right", the logical structure of the front side and the front side, and if the analysis direction is "bottom", it is the direction in which the overlapping degree of the child cell is calculated. Identify the logical structure of the head and head of the table. In this example, only the identification of the logical structure between the front side and the front side will be explained, but it is also possible to recognize logical structures such as head and head, head and value, and front and value using a mechanism similar to this example. .

H3 can calculate the overlap rate with the parent cell using the child cell as a reference according to the analysis direction specified in H2, and can specify the boundary conditions for the calculated overlap rate. For example, according to the input value of H3, a logical structure is assigned when the overlap rate with the parent cell exceeds 80% using the child cell as a reference.

Since the parameters specified by the user in the logic window H1 require complex logic as described above, it is difficult for the user to understand the influence of the parameters on the logic. Therefore, regarding the parameters of the logic window H1 being edited, the logic flow management unit 102 executes the logical structure recognition logic execution unit 102-2 and the logic visualization I/F update unit 102-3, and the logic visualization I/F update unit 102 -3 uses logic visualization I/F C4 to intuitively present the knowledge necessary for adjusting logic parameters to the user. In Figure 10, C4 informs the user of the shape of the cell processed by logic, as shown by H4, and informs the user of the overlap rate using a string, a rectangle, and a double arrow symbol, as shown by H5. I can do it.

When the logic window H1 receives input from the user, it executes the logical structure recognition logic execution unit 102-2, the logic visualization I/F update unit 102-3, and the logic application result I/F update unit 102-4, The logic application result I/F update unit 102-4 updates the logic application result I/F C5. For example, in FIG. 10, C5 presents the user with the result of identifying the logical structure between the two front sides, overlapping the arrow symbols as a logical structure.
<Effects of Example 6>
According to this embodiment, the logic flow management unit receives and outputs the logic specified by the user from the input device, and the logic flow management unit performs necessary adjustment of the parameters of the logical structure recognition logic using the shape of the component as a clue. The information is visualized using the logic visualization I/F, and the results of applying the logical structure recognition logic between the front side and the front side are reflected in the logic application result I/F, reducing the burden on the user involved in logic creation.
<Configuration of information acquisition device 100 of Example 7>
In addition to the configuration of the information acquisition device 100, the information acquisition device of the seventh embodiment includes a logic selection unit 200 for supporting logic creation by allowing the user to select logic from a logic list prepared in advance.
<Processing of information acquisition device 100 of Example 7>
FIG. 11 shows the logic flow management unit 102 and logic selection unit 200 of the information acquisition device 100 in the seventh embodiment.

The user displays the logic selection section 200 by clicking C62. Note that any means may be used to display the logic selection section 200. At this time, for example, the logic selection section 200 may be displayed on the right side of the logic flow editing I/F, or it may be implemented as a separate application.

The logic selection unit 200 presents the user with a list of logics that can be added to the logic flow. Furthermore, the logic selection section 200 may include a button for inserting the addable logic, and explanatory text and diagrams of the addable logic. For example, by clicking the button for inserting logic, the logic specified by the user can be added to the logic flow.
<Effects of Example 7>
According to this embodiment, the logic flow management section receives and outputs the logic specified by the user from the input device, and the logic flow management section allows the user to select the logic prepared in advance from the logic list selection section. Reduce the burden on users involved in logic creation.
<Example of hardware configuration of computer 300 that implements information acquisition device FIG. 12 shows an example of hardware configuration of computer 300 that implements information acquisition device 100. The computer 300 includes an input device 304, an output device 305, a communication interface 306, a storage device 301, an arithmetic device 302, and a bus 303.

The input device 304 is a device through which the user inputs the input table 101a, component recognition logic 101b, logical structure recognition logic 101c, or commands to the information acquisition device 100. Input from the input device 304 may be stored in the storage device 301. As the input device 304, for example, a keyboard, a mouse, a touch panel, a microphone, a camera, and a scanner may be used to issue commands to the logic flow management unit 102.

The output device 305 presents the logic flow management unit 102, logic visualization I/F, logic application result I/F, and logic flow I/F output by the information acquisition device 100 to the user. The output device 305 may be, for example, a display.

The communication interface 306 is connected to a network and sends and receives various data necessary for the operation of the computer 300. At this time, the input device 304 and the output device 305 can be connected through a communication interface 306.

The arithmetic unit 302 operates the computer 300 according to electronic instructions and executes a logic flow. Computing device 302 may also include single or multiple computing or processing devices. The arithmetic device 302 may be any device that operates according to electronic instructions. At this time, for example, it is possible to use a CPU (Central Processing Unit).

The storage device 301 functions as a work area for the arithmetic device 302. The storage device 301 records programs for executing the information acquisition device 100 and data. Further, the storage device 301 may be implemented using any kind of medium. At this time, various media such as non-volatile memory, hard disk, or volatile memory can be considered.

101 Input unit 102 Logic flow management unit 102-1 Component recognition logic execution unit 102-2 Logical structure recognition logic execution unit 102-3 Logic visualization I/F update unit 102-4 Logic application result I/F update unit 103 Output unit

Claims (7)

A logic creation support system for recognizing the logical structure of documents,
The logic flow management unit is connected to the input/output device, and manages the logic and logic flow of component recognition and logical structure recognition.For component recognition, the component recognition logic specified by the user is transferred to the input device. The logical structure recognition unit receives and outputs the output and the logical structure recognition logic specified by the user in the logic flow management unit from the input device, and the logic flow management unit outputs the logic visualization and logic application results through the interface. reflected in,
A logic creation support system characterized by: The logic creation support system according to claim 1,
The logic flow management unit visualizes information necessary for parameter adjustment of the component recognition logic using a logic visualization I/F, and reflects the result of applying the component recognition logic to the logic application result I/F.
A logic creation support system characterized by: The logic creation support system according to claim 1,
The logic flow management unit visualizes the information necessary for parameter adjustment of the front side, front head, and value recognition logics among the component recognition logics using the logic visualization I/F, and displays the results of applying the component recognition logics. is reflected in the logic application result I/F,
A logic creation support system characterized by: The logic creation support system according to claim 1,
The logic flow management unit visualizes information necessary for parameter adjustment of the logical structure recognition logic by indentation using a logic visualization I/F, and reflects the result of applying the logical structure recognition logic on the logic application result I/F.
A logic creation support system characterized by: The logic creation support system according to claim 1,
The logic flow management unit visualizes the information necessary for parameter adjustment of the logical structure recognition logic using the distance between components using the logic visualization I/F, and converts the result of applying the logical structure recognition logic into a logic application result I/F. Reflect in /F,
A logic creation support system characterized by: The logic creation support system according to claim 1,
The logic flow management unit visualizes the information necessary for parameter adjustment of the logical structure recognition logic using the shape of the component as a clue, using the logic visualization I/F, and converts the result of applying the logical structure recognition logic into a logic application result I/F. Reflect in /F,
A logic creation support system characterized by: The logic creation support system according to claim 1,
The logic flow management section receives and outputs the logic specified by the user from the input device, and the logic flow management section selects the logic prepared in advance by the user from the logic list selection section.
Logic creation support system.

Images