Text structure analysis system

Abstract

To provide a sentence structure analysis system that accurately classifies documents for each headline.SOLUTION: A sentence structure analysis system 10 includes an information processor 20 that analyzes a document having a plurality of headlines. A processing circuit 21 of the information processor 20 acquires, on the basis of positions in a line direction, the plurality of headlines from the document. The processing circuit 21 regards, out of the plurality of acquired headlines, a document ranging from a first headline to a preceding paragraph of a second headline which is a headline next to the first headline, as one block, and classifies, out of the plurality of headlines, sentences ranging from the second headline to a preceding paragraph of a third headline which is a headline next to the second headline, as one block.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a text structure analysis system.

Background Art

[0002] Patent Document 1 discloses an example of a text analysis device that analyzes a hierarchical structured document with headings such as "chapter", "section", and "item". The device divides the document into sections by analyzing the sentences constituting the document. Specifically, the device detects styles and expressions that are likely to be the titles of sections. Then, the device obtains the sections from the document by evaluating the detected styles and expressions, and then divides the document into sections.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In a document with headings, there may be no uniformity in the way the headings are given or there may be misrecords in the headings. Even if the above-described processing is performed on such a document, there is a risk that the document cannot be correctly divided into sections.

Means for Solving the Problems

[0005] A document structure analysis system for solving the above problem is a system for analyzing a document with multiple headings. When the direction in which characters are arranged in the document is defined as the line direction, the document structure analysis system includes a processing circuit. Based on the position in the line direction, the processing circuit obtains multiple headings from the document, and divides the document into blocks, with the text from the first heading up to the preceding section of the second heading (which is the heading following the first heading) being considered as one block, and the text from the second heading up to the preceding section of the third heading (which is the heading following the second heading) being considered as another block.

[0006] In a document, multiple headings are listed at approximately the same position along a given line. By utilizing this regularity in the document structure, it is possible to extract headings. Therefore, the above-described document structure analysis system retrieves multiple headings from a document based on their position in the line direction. By retrieving headings while considering the document structure in this way, the accuracy of heading extraction can be increased. As a result, it becomes possible to accurately categorize the document by heading. [Brief explanation of the drawing]

[0007] [Figure 1] Figure 1 is a block diagram illustrating the schematic of the document structure analysis system according to the embodiment. [Figure 2] Figure 2 shows an example of a document analyzed by the document structure analysis system shown in Figure 1. [Figure 3] Figure 3 shows an example of a document analyzed by the document structure analysis system shown in Figure 1. [Figure 4] Figure 4 is a flowchart showing the series of processes performed by the text structure analysis system shown in Figure 1. [Figure 5] In Figure 5, (A) and (B) are display screens showing headings acquired by the processing circuit of the text structure analysis system in Figure 1. [Figure 6]In Figure 6, (A) and (B) are diagrams showing display screens that show headings acquired by the processing circuit in Figure 5. [Modes for carrying out the invention]

[0008] An embodiment of the text structure analysis system will be described below with reference to Figures 1 to 6. As shown in Figure 1, the document structure analysis system 10 is a system for analyzing documents. The document structure analysis system 10 comprises an information processing device 20 and a terminal device 30. The terminal device 30 is configured to send and receive various types of information with the information processing device 20 via a communication line 11 such as a LAN (Local Area Network).

[0009] <Documents to be analyzed> Referring to Figures 2 and 3, the document 100 that the document structure analysis system 10 analyzes will be described.

[0010] Document 100 is composed of multiple sentences. Document 100 is a hierarchical document with multiple headings such as chapters, sections, and subsections. For example, document 100 is a legal document. In document 100 as described above, the direction in which the characters are arranged is called the "row direction," and the direction perpendicular to the row direction on the page is called the "column direction." In this embodiment, the row direction is called the "X direction," and the column direction is called the "Y direction."

[0011] Document 100 has headings for each level of hierarchy. Each of these headings is assigned a heading number corresponding to its level. The heading number includes a number indicating the heading's order. The highest level of the hierarchy, the first level, is represented by the heading "Chapter." First-level headings are assigned the chapter number as their heading number. For example, the heading for Chapter 6 has the heading number "6" at the beginning, followed by a descriptive text for the heading.

[0012] The headings at the second level, which is one level below the first level, are called "sections." Second-level headings are assigned heading numbers consisting of the chapter number and the section number. For example, the heading for Chapter 6, Section 20 has the heading number "6" at the beginning, followed by "20." Additionally, a descriptive text for the heading is placed after the heading number.

[0013] The heading at the third level, which is one level below the second level, is called a "subheading." Third-level headings are assigned heading numbers consisting of the chapter number, section number, and subheading number. For example, the heading for Chapter 6, Section 20, Subheading 1 has the heading number "6" at the beginning, followed by "20," and then "1." In addition, a descriptive text for the heading is placed after the heading number.

[0014] Furthermore, in the hierarchical document 100, the position of headings in the X direction is consistent. In the examples shown in Figures 2 and 3, the first character of the heading, specifically the position of the chapter number in the X direction, is consistent.

[0015] <Information Processing Device> As shown in Figure 1, the information processing device 20 has a processing circuit 21. The processing circuit 21 is, for example, a microcomputer. In this case, the processing circuit 21 has a CPU 22, a first memory 23, and a second memory 24. The first memory 23 stores a control program executed by the CPU 22. The CPU 22 executes this control program, and the processing circuit 21 analyzes the document 100. Specifically, the processing circuit 21 generates multiple block BKs by dividing the document 100 into headings of the same hierarchical level. Here, a "block BK" refers to a block of text.

[0016] The second memory 24 is a rewritable memory. This second memory 24 stores the regular pattern PT1 and the over-extraction pattern PT2. When a heading with a correctly assigned heading number is defined as a "normal heading", the normal pattern PT1 indicates the rules for marking the heading numbers of normal headings. In the present embodiment, the normal pattern PT1 for the heading numbers assigned to the first-level headings is ".chapter number". The normal pattern PT1 for the heading numbers assigned to the second-level headings is ".chapter number.section number". The normal pattern PT1 for the heading numbers assigned to the third-level headings is ".chapter number.section number.item number". That is, the pattern where the end of the heading number ends with "." is the normal pattern PT1.

[0017] The over-extraction pattern PT2 is a pattern for extracting not only the above-mentioned normal headings but also non-normal headings. A non-normal heading is a heading to which a heading number is assigned in accordance with a similar pattern that is different from the normal pattern PT1 but similar to the normal pattern PT1. The over-extraction pattern PT2 for the heading numbers assigned to the first-level headings includes, in addition to ".chapter number", for example, "chapter number" and "chapter number#". The over-extraction pattern PT2 for the heading numbers assigned to the second-level headings includes, in addition to ".chapter number.section number", for example, "chapter number.section number" and "chapter number.section number#". The over-extraction pattern PT2 for the heading numbers assigned to the third-level headings includes, in addition to ".chapter number.section number.item number.", for example, "chapter number.section number.item number" and "chapter number.section number.item number#". That is, the similar pattern is a pattern that includes chapter numbers, section numbers, and item numbers but does not end with "." at the end of the heading number.

[0018] <Terminal device> The terminal device 30 has a display unit 31 and an operation unit 32 as user interfaces. The display unit 31 displays information transmitted from the information processing device 20, such as the analysis result of the document 100. The operation unit 32 is operated by an operator based on the information displayed on the display unit 31. The operation unit 32 includes, for example, a keyboard and a mouse. The terminal device 30 transmits information corresponding to the operation of the operation unit 32 by the operator to the information processing device 20.

[0019] <Article parsing process> Referring to FIGS. 4 and 5, the article parsing process executed by the article structure analysis system 10 will be described. The article parsing process is a series of processing flows for analyzing the document 100 acquired by the information processing apparatus 20 and classifying the document 100 for each heading at the same hierarchy.

[0020] In step S11, the processing circuit 21 of the information processing apparatus 20 reads the normal pattern PT1 from the second memory 24. In the next step S13, the processing circuit 21 reads the over-extracted pattern PT2 from the second memory 24.

[0021] Subsequently, the processing circuit 21 acquires a plurality of headings at the same hierarchy from the document 100 based on the position in the X direction. For example, the processing circuit 21 acquires the headings of the second hierarchy and the headings of the third hierarchy from the document 100.

[0022] Specifically, in step S15, the processing circuit 21 acquires a normal heading, which is a heading assigned with a heading number conforming to the normal pattern PT1, from the document 100 based on the normal pattern PT1. For example, the normal pattern of the third hierarchy is ".chapter number.section number.item number.". Therefore, assuming that the document to be analyzed is the document shown in FIG. 2, the processing circuit 21 acquires the heading assigned with "6.20.4." as the heading of the third hierarchy.

[0023] In step S17, the processing circuit 21 obtains headings from document 100 that have heading numbers corresponding to similar patterns, based on the over-extraction pattern PT2. For example, suppose the third-level over-extraction pattern PT2 is "Chapter number.Section number.Item number.", "Chapter number.Section number.Item number" and "Chapter number.Section number.Item number#". Assume that the document 100 to be analyzed is the document shown in Figure 2. In this case, the processing circuit 21 obtains headings with "6.20.1", "6.20.2", "6.30.3", "6.20.4." and "6.20.5" as third-level headings. Of the multiple headings obtained here, the heading with "6.20.4." was also obtained in step S15. Therefore, the processing circuit 21 removes the heading of normal pattern PT1 from the multiple headings obtained based on the over-extraction pattern PT2. As a result, the processing circuit 21 can obtain non-regular headings from document 100, which are headings assigned heading numbers according to similar patterns.

[0024] In the subsequent step S18, the processing circuit 21 divides the multiple headings into at least one group by analyzing the regularity of the heading numbers assigned to the multiple headings obtained in steps S15 and S17. For example, the processing circuit 21 analyzes the regularity of the multiple heading numbers to determine which level of heading the heading number belongs to and whether or not the heading number conforms to a regular pattern. Then, if the processing circuit 21 has obtained, for example, headings from the second level and headings from the third level, it classifies the headings from the second level and the headings from the third level into separate groups. Also, if the processing circuit 21 has obtained, for example, regular headings with heading numbers conforming to a regular pattern and non-regular headings with heading numbers conforming to a similar pattern, it classifies the regular headings and non-regular headings into separate groups.

[0025] The following describes the analysis process when multiple third-level headings are obtained from document 100 shown in Figure 2. In this case, all of the headings obtained by the processing circuit 21 are third-level headings. Furthermore, some of the headings are assigned heading numbers that conform to a regular pattern, while the rest are assigned heading numbers that do not conform to a regular pattern. Therefore, the processing circuit 21 divides the obtained headings into a first group consisting of regular headings that are assigned heading numbers that conform to a regular pattern, and a second group consisting of non-regular headings that are assigned heading numbers that do not conform to a regular pattern. The processing circuit 21 then transmits the analysis results to the terminal device 30.

[0026] When these analysis results are transmitted from the information processing device 20 to the terminal device 30, the document structure analysis system 10 proceeds to step S19. In step S19, the terminal device 30 displays the multiple headings obtained by the execution of the processes in steps S15 and S17 on the display unit 31. At this time, the terminal device 30 performs display processing that also takes into account the analysis results performed by the terminal device 30 in step S18.

[0027] Here, referring to Figure 5, we will explain what is displayed on the display unit 31 when a third-level heading is obtained from the document 100 shown in Figure 2. The terminal device 30 displays the display screen 31A shown in Figure 5(A) and the display screen 31B shown in Figure 5(B) on the display unit 31. Display screen 31A is a screen that displays regular headings with heading numbers according to the regular pattern PT1. That is, display screen 31A displays headings classified into the first group described above. On display screen 31A, "Clause (3)" indicates that a third-level heading is being displayed. For example, to indicate that a Nth-level heading is being displayed, "Clause (N)" is displayed. Note that "N" is an integer of 1 or more.

[0028] Display screen 31B is a screen that displays headings numbered according to similar patterns. In other words, display screen 31B displays headings classified into the second group described above. Note that even on display screen 31B, "Clause (3)" indicates that it is displaying a heading of the third level.

[0029] In the multiple display screens 31A and 31B, checkboxes CB are placed next to "Clause (3)" and next to headings, respectively. As shown in Figures 5(A) and (B), multiple checkboxes CB are checked by default. The fact that a checkbox CB next to a heading is checked indicates that the heading in question has been correctly acquired by the processing circuit 21. Therefore, if the processing circuit 21 has acquired an incorrect heading, the operator must uncheck the checkbox CB next to that heading by operating the operation unit 32. For example, if the processing circuit 21 has acquired an incorrect heading that begins with "6.20.2", the operator must uncheck the checkbox CB next to the heading that begins with "6.20.2" by operating the operation unit 32.

[0030] For example, if the checkbox CB next to "Clause (3)" is unchecked, the checkboxes CB next to all the headings displayed below it will also be unchecked.

[0031] Returning to Figure 4, the terminal device 30 displays the result of acquiring the heading on the display unit 31 and then proceeds to step S21. In step S21, the terminal device 30 determines whether or not the operator has completed the operation of the operation unit 32.

[0032] When the results of acquiring headings are displayed on the display unit 31, the operator determines whether there are any excesses or omissions in the acquisition of headings by looking at the display unit 31. If there are no excesses or omissions in the acquisition of headings, the operator performs the operation to proceed with the process.

[0033] On the other hand, if there are too many headings retrieved, the worker unchecks the checkbox CB next to the incorrectly retrieved heading, as described above. Furthermore, if there are any missing headings, the operator performs an operation to change the over-extraction pattern PT2. For example, if the third-level over-extraction pattern PT2 does not include "chapter number.section number.item number," the processing circuit 21 cannot obtain the headings with "6.20.6," and "6.20.7," from the document 100 shown in Figure 2. Therefore, the operator operates the operation unit 32 so that the third-level over-extraction pattern PT2 includes "chapter number.section number.item number,". This operation of the operation unit 32 to change the over-extraction pattern PT2 is called a "change operation."

[0034] If the terminal device 30 determines that the operator has not yet completed the operation of the control unit 32 (S21: NO), it repeats the determination in step S21 until it determines that the operation has been completed. On the other hand, if the terminal device 30 determines that the operator has completed the operation of the control unit 32 (S21: YES), it proceeds to step S23.

[0035] In step S23, the processing circuit 21 of the information processing device 20 determines whether the operator's operation on the terminal device 30 was the above-mentioned modification operation. If the operator's operation was the above-mentioned modification operation (S23: YES), the processing circuit 21 proceeds to step S15. In this case, the processing circuit 21 again performs the acquisition of the heading based on the regular pattern PT1 in step S15, and the acquisition of the heading based on the modified over-extraction pattern PT2 in step S17.

[0036] On the other hand, in step S23, if the operator's operation on the terminal device 30 is not a change operation (S23: NO), the processing circuit 21 proceeds to step S25. In step S25, the processing circuit 21 divides document 100 into multiple blocks based on the headings it has obtained from document 100. Specifically, the processing circuit 21 considers the document from the first heading up to the section before the second heading, which is the heading following the first heading, as one block BK. The processing circuit 21 also considers the document from the second heading up to the section before the third heading, which is the heading following the second heading, as another block BK.

[0037] Once the document 100 has been divided into sections according to its headings, the document structure analysis system 10 completes its series of processes. <Operation of this embodiment> Referring to Figures 2 and 5, the operation of the information processing device 20 when it analyzes the document 100 shown in Figure 2 will be explained. Here, the case in which the document 100 is divided according to the headings of the third level will be explained.

[0038] In the information processing device 20, a third-level heading is obtained with a heading number according to the regular pattern PT1, based on the position in the X direction of the document 100 and the regular pattern (S15). In addition, a third-level heading is obtained with a heading number according to the similar pattern, based on the position in the X direction of the document 100 and the over-extraction pattern PT2 (S17).

[0039] Here, the third-level over-extraction pattern PT2 is defined as including "chapter number.section number.item number.", "chapter number.section number.item number", and "chapter number.section number.item number#", but not "chapter number.section number.item number,".

[0040] In this case, the information processing device 20 retrieves headings with "6.20.1", "6.20.2", "6.20.3", "6.20.4", and "6.20.5" from document 100. On the other hand, headings with "6.20.6," and "6.20.7," are not retrieved from document 100. Therefore, headings with "6.20.6," and "6.20.7," are not displayed on the display screen 31B showing the retrieval results (S19).

[0041] In this embodiment, the operator can add "Chapter number.Section number.Item number," to the over-extraction pattern PT2 by performing a modification operation (S21). When such a modification operation is performed (S23:YES), the process of obtaining headings from document 100 is executed again (S15, S17, S18). Then, the information processing device 20 also obtains headings with "6.20.6," and headings with "6.20.7," from document 100.

[0042] As a result, the display screen 31B also displays headings labeled "6.20.6," and "6.20.7," as shown in Figure 5(B) (S19). Then, once the operator confirms that there are no discrepancies in the acquisition of headings (S23:NO), the information processing device 20 divides the document 100 according to the acquired headings (S25). As a result, as shown in Figure 2, a block BK for each heading in the third tier is generated.

[0043] Next, referring to Figures 3 and 6, we will explain the operation of the information processing device 20 when it analyzes the document 100 shown in Figure 3. In the information processing device 20, based on the position in the X direction of the document 100 and the regular pattern PT1, a second-level heading is obtained with a heading number in accordance with the regular pattern PT1 (S15). Specifically, the processing circuit 21 obtains the headings with "12.1.", "12.2.", "12.3.", and "12.4." as second-level headings.

[0044] In document 100 shown in Figure 3, a sentence beginning with "03" is written following "12.2.". Furthermore, a sentence beginning with "04" is written following "12.3.". And finally, a sentence beginning with "05" is written following "12.4.".

[0045] Therefore, the information processing device 20 acquires the headings labeled "12.2.03", "12.3.04", and "12.4.05" as third-level headings based on the position in the X direction of the document 100 and the over-extraction pattern PT2.

[0046] In the document structure analysis system 10, the processing circuit 21 analyzes the regularity of the heading numbers attached to the multiple headings obtained (S18). In the example described here, some of the multiple headings obtained are second-level headings, and the rest are third-level headings. Therefore, the second-level headings are classified into the first group, and the third-level headings are classified into the second group.

[0047] Then, the display unit 31 of the terminal device 30 displays a display screen 31C that shows the second-level headings shown in Figure 6(A), and a display screen 31D that shows the third-level headings shown in Figure 6(B) (S19). That is, display screen 31C displays headings classified into the first group. On the other hand, display screen 31D is assigned headings classified into the second group.

[0048] If the third-level heading displayed on the display screen 31D is an incorrectly acquired heading, the operator unchecks the checkbox CB next to "Clause (3)" on the display screen 31D. This deletes the headings with "12.2.03", "12.3.04", and "12.4.05" from the headings acquired by the processing circuit 21. As a result, document 100 is no longer divided by the headings with "12.2.03", "12.3.04", and "12.4.05" (S25). Consequently, as shown in Figure 3, blocks BK are generated for each second-level heading.

[0049] <Effects of this embodiment> (1) In the document structure analysis system 10, multiple headings of a certain hierarchy are obtained from document 100 based on the position of document 100 in the X direction. By obtaining headings while considering the structure of document 100 in this way, the accuracy of heading extraction can be increased. Therefore, the division of document 100 into headings can be performed with high accuracy.

[0050] (2) In the document structure analysis system 10, in addition to the regular pattern PT1, the over-extraction pattern PT2 is stored in the second memory 24 of the information processing device 20. Therefore, the processing circuit 21 can acquire headings with heading numbers assigned according to the regular pattern PT1, as well as headings with heading numbers assigned according to the similar pattern. This makes it less likely that headings will be missed during acquisition.

[0051] (3) The processing circuit 21 cannot acquire headings that have heading numbers assigned to patterns not included in the over-extraction pattern PT2. In this regard, the document structure analysis system 10 displays all acquired headings on the display unit 31 once the heading acquisition is complete. This allows the operator to check whether or not there are any missing headings. If the operator discovers any missing headings, they can add the over-extraction pattern PT2 through a modification operation. Once such a modification operation is performed, the heading acquisition process is executed again. By making the over-extraction pattern PT2 modifiable in this way, the effectiveness of suppressing missing headings can be increased.

[0052] (4) When the processing circuit 21 acquires multiple headings, it analyzes the regularity of the heading numbers assigned to those headings. The processing circuit 21 then divides the multiple headings into at least one group. If the multiple headings are classified into multiple groups, the terminal device 30 displays a display screen for each group on the display unit 31. For example, the display screen for the first pattern group and the display screen for the second pattern group are displayed on the display unit 31. If, for example, a heading classified into the first group was acquired in error, the operator can delete the heading classified into the first group all at once. As a result, it is possible to prevent the document 100 from being divided by an incorrectly acquired heading.

[0053] <Example of changes> The above embodiment can be implemented with the following modifications. The above embodiment and the following modifications can be combined with each other to the extent that they do not contradict each other technically.

[0054] After both the acquisition of headings based on the regular pattern PT1 and the acquisition of headings based on the over-extraction pattern PT2 have been performed, the acquisition results are displayed on the display unit 31, for example, as shown in Figure 6(A). At this time, if there are any missing headings, for example, a heading with "12.3." attached may be displayed after a heading with "12.1." attached. In this case, the processing circuit 21 can determine that there is a possibility that the heading with "12.2." attached could not be acquired. Therefore, the processing circuit 21 should output a command to the terminal device 30 to notify the operator that there may be a missing heading. In this case, the operator will operate the operation unit 32 to modify the regular pattern PT1 and the over-extraction pattern PT2 so that the processing circuit 21 can acquire the missing headings.

[0055] The documents to be analyzed may be other documents besides legal documents, as long as they are hierarchical documents with headings. Examples of other documents include product instruction manuals and specifications, legal documents, and academic papers.

[0056] In the above embodiment, the processing circuit 21 divides a horizontally written document into sections according to headings, as shown in Figures 2 and 3, but it is not limited to this. That is, the processing circuit 21 may also divide a vertically written document into sections according to headings.

[0057] The document to be analyzed can have any hierarchical structure, as long as it has a hierarchical structure. For example, the document may have three levels of hierarchy, four levels, or five levels.

[0058] In the above embodiment, the processing circuit 21 divides the multiple headings it has acquired into multiple groups based on the regularity of the heading numbers. The display unit 31 of the terminal device 30 displays a separate display screen for each group, but this is not limited to this configuration. For example, the processing circuit 21 does not need to perform the process of dividing the multiple headings it has acquired into multiple groups based on the regularity of the heading numbers. In this case, the display unit 31 of the terminal device 30 may display a display screen that shows all the headings. Preferably, the terminal device 30 displays all the headings on the display unit 31 so that the heading numbers are in order.

[0059] In the above embodiment, a checkbox CB is used as a means to select the correctly acquired heading from among the multiple acquired headings, but this is not limited to this. Any selection means other than a checkbox CB may be used as long as it is possible to select either the correctly acquired heading or the incorrectly acquired heading from among the multiple acquired headings.

[0060] If the heading numbers of all the headings in the document are assigned according to a regular pattern, then it is not necessary to set up the over-extraction pattern PT2. In the above embodiment, if headings can be obtained from the document based on their position in the X direction, it is not necessary to consider patterns of heading numbers, etc.

[0061] The processing circuit 21 is not limited to one that includes a CPU and ROM and executes software processing. In other words, the processing circuit 21 may have any of the following configurations: (a), (b), and (c).

[0062] (a) The processing circuit 21 comprises one or more processors that perform various processes according to a computer program. The processor includes a CPU and memory such as RAM and ROM. The memory stores program code or instructions configured to cause the CPU to perform the processes. The memory, i.e., computer-readable media, includes any available media that can be accessed by a general-purpose or dedicated computer.

[0063] (b) The processing circuit 21 includes one or more dedicated hardware circuits that perform various processes. Examples of dedicated hardware circuits include application-specific integrated circuits, i.e., ASICs or FPGAs. ASIC is an abbreviation for "Application Specific Integrated Circuit," and FPGA is an abbreviation for "Field Programmable Gate Array."

[0064] (c) The processing circuit 21 includes a processor that executes a portion of the various processes according to a computer program, and a dedicated hardware circuit that executes the remaining processes among the various processes.

[0065] In this specification, the expression "at least one" means "one or more" of the desired options. For example, if there are two options, the expression "at least one" means "only one option" or "both of the two options." As another example, if there are three or more options, the expression "at least one" means "only one option" or "a combination of two or more arbitrary options." [Explanation of symbols]

[0066] 10...Document structure analysis system, 20...Information processing device, 21...Processing circuit, 100...Document, BK...Block.

Claims

1. A document structure analysis system that analyzes documents with multiple headings, The headings of the aforementioned documents are numbered sequentially, In the aforementioned document, when the direction in which the characters are arranged is defined as the line direction, Equipped with a processing circuit, The aforementioned processing circuit is When retrieving multiple headings from the document based on the row-direction position, Based on the regular pattern, the headings assigned heading numbers in accordance with that regular pattern are obtained from the document. Headings that are assigned heading numbers according to a similar pattern, even if they differ from the aforementioned canonical pattern, are obtained from the document. Of the multiple headings obtained, the text from the first heading up to the beginning of the second heading (which follows the first heading) is treated as one block, and the text from the second heading up to the beginning of the third heading (which follows the second heading) is treated as another block, thereby dividing the document. A text structure analysis system.

2. The aforementioned document is a legal document, The processing circuit divides the legal document into multiple blocks according to its headings. The text structure analysis system according to claim 1.

Images