Document processing support device and document processing support method

Abstract

To facilitate checking contents of laws and regulations by enhancing accuracy of comparison between a checklist and a law in creating and confirming the checklist when performing laws and regulation checking.SOLUTION: A document processing support device receives an input of selecting a comparison unit of a checklist, divides the checklist into comparison unit items for each comparison unit, and calculates an average size of texts in the comparison unit items. Also, the document processing support device divides a law into division items for each of a plurality of division units, calculates an average size of the division items, selects a division unit for the law whose average size of the division items is the closest to the average size of the texts in the comparison unit items of the checklist, calculates a similarity between each of the divided texts of the checklist and each of the texts of the division items of the selected division unit, and outputs a pair of a text of the comparison unit items of the checklist and a text of the division item for the law that have the greatest similarity between them in association with each other.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a document processing support device and a document processing support method, and more particularly to a document processing support device and a document processing support method suitable for creating checklists and verifying the content of laws and regulations using checklists when companies and other organizations perform checklists to ensure compliance with laws and regulations in their operations. [Background technology]

[0002] In recent years, the importance of compliance in businesses has become widely recognized. Compliance means "adherence to laws and regulations," and violations of compliance can cause significant damage to a company, and in the worst cases, can even lead to its closure. Therefore, it is considered an extremely important and indispensable aspect of business management.

[0003] In some cases, companies create and use checklists to verify compliance with laws and regulations related to their business operations. Traditionally, these checklists were created and checked manually, which meant that the creators of the checklists sometimes made errors such as omissions when comparing the items with the relevant laws, and it was time-consuming for those checking the checklists to refer to the relevant laws. Thus, creating checklists required that they reliably covered the content of the laws and regulations and that their content was accurately understood, which was a challenge for both the creators of the checklists and those checking the items using the checklists.

[0004] To address these challenges, it is necessary to map checklists to relevant legal provisions.

[0005] Document processing technology related to this is disclosed in Patent Document 1. The document processing method in Patent Document 1 discloses, for example, a technology that calculates the degree of relevance between two contract documents on a unit (article-by-article) basis and displays both contracts based on that degree of relevance (Figure 12A, paragraphs

[0066] to

[0068] ). [Prior art documents] [Patent Documents]

[0006] [Patent Document 1] International Publication No. 2021 / 214963 [Overview of the Initiative] [Problems that the invention aims to solve]

[0007] The technology described in Patent Document 1 was designed to enable the understanding of the relationships between documents on a unit-by-unit basis when comparing a contract document under consideration with a comparable contract document.

[0008] However, in the document processing method described in Patent Document 1, the units of the items used to calculate relevance are predetermined. When this technology is applied to compare checklists and laws and regulations and attempts are made to display the checklist and related laws, the size of the corresponding items (e.g., articles, paragraphs, subparagraphs) may change depending on the content included in the items of the checklist being compared, and this method cannot handle such cases. As a result, the accuracy of the comparison between checklists and laws decreases, increasing the likelihood of errors such as omissions for checklist creators, and making the comparison difficult to refer to for those using the checklist.

[0009] The object of the present invention is to provide a document processing support device and a document processing support method that, when companies and other organizations use checklists to check compliance with laws and regulations in their operations, can flexibly change the unit of comparison between the checklist and the law according to the content of the checklist, thereby improving the accuracy of the comparison between the checklist and the law for the checklist creator and presenting an easily understandable relationship between the checklist and the law for the person performing the check. [Means for solving the problem]

[0010] The configuration of the document processing support device of the present invention is preferably a document processing support device that outputs relationships between multiple documents, and accepts input for selecting a comparison unit for a first document, divides the first document into comparison unit items for each input comparison unit, stores information on multiple division units for a second document, calculates the average size of the text for each comparison unit item of the first document, divides the second document into multiple division units for each of the multiple division units for the second document, calculates the average size of the division items obtained by dividing the second document into those division units, selects the division unit for the second document that has the closest average size of the text for each comparison unit item of the first document and the average size of the division items obtained by dividing the second document into division units, calculates the similarity between the text of the comparison unit item of the first document and the division items obtained by dividing the second document into the selected division unit, and outputs pairs of the text of the comparison unit item of the first document and the division items obtained by dividing the second document into division units with a large similarity between the two. [Effects of the Invention]

[0011] According to the present invention, when a company or the like checks compliance with laws and regulations using a checklist in its business, when creating a checklist or checking the content of laws and regulations using the checklist, in accordance with the content of the checklist, the comparison unit between the checklist and the laws and regulations can be flexibly changed, improving the accuracy of the comparison between the checklist and the laws and regulations for the checklist creator, and presenting an understandable relevance between the checklist and the laws and regulations for the checker. A document processing support device and a document processing support method can be provided.

Brief Description of the Drawings

[0012] [Figure 1] It is a functional configuration diagram of a document processing support device. [Figure 2] It is a hardware / software configuration diagram of a document processing support device. [Figure 3A] It is a diagram showing a specific example of a checklist. [Figure 3B] It is a diagram showing some articles of the Enforcement Order of the Building Standards Law. [Figure 4A] It is a diagram showing a specific example when the divided items are in the unit of each article. [Figure 4B] It is a diagram showing a specific example when the divided items are in the unit of each paragraph. [Figure 4C] It is a diagram showing a specific example when the divided items are in the unit of each item. [Figure 5] It is a diagram showing the result of morphological analysis of a part of the Enforcement Order of the Building Standards Law. [Figure 6] It is a diagram showing the morphological vector of the entire document (checklist and Article 24, Paragraph 2, and Article 25, Paragraph 1 of the Enforcement Order of the Building Standards Law). [Figure 7A] It is a diagram showing the word appearance frequency vector of the text of the check items of the checklist. [Figure 7B] It is a diagram showing the word appearance frequency vector of the text of the check items of the law checklist. [Figure 8] It is a diagram showing the document number vector with the number of documents containing a word as an element. [Figure 9]This flowchart shows the details of the process by which the document processing support device according to Embodiment 1 displays a checklist and related laws and regulations in association with it. [Figure 10] This flowchart shows the details of the legal division process. [Figure 11] This flowchart shows the details of the checklist-based legal compliance process. [Figure 12] This figure shows an example of the checklist - legal association screen. [Figure 13] This flowchart shows the details of the process by which the document processing support device according to Embodiment 2 displays a checklist in association with related laws and regulations. [Figure 14] This is a diagram showing the configuration of the document processing support system according to Embodiment 3. [Modes for carrying out the invention]

[0013] Hereinafter, one embodiment of the present invention will be described with reference to Figures 1 to 14.

[0014] [Embodiment 1] Hereinafter, a first embodiment of the present invention will be described with reference to Figures 1 to 12.

[0015] (I) Configuration of the document processing support system The document processing support device according to Embodiment 1 is configured to operate as a standalone unit.

[0016] First, we will explain the functional configuration of the document processing support device using Figure 1. The document processing support device 100 is a device that provides users with the creation of checklists for checking laws and regulations, and utility functions for associating checklists with laws and regulations. As shown in Figure 1, it has a checklist input unit 201, a law input unit 202, a checklist division unit 203, a law division unit selection unit 204, a law division unit 205, a similarity calculation unit 206, a similar item acquisition unit 207, a checklist-law relationship display unit 210, and a storage unit 220.

[0017] The checklist input unit 201 is a functional unit that inputs a checklist from an external device or an input device of a document processing support system and imports it into the storage unit 220 as checklist data 221.

[0018] The legal input unit 202 is a functional unit that inputs legal documents from an external device or an input device of a document processing support device and imports them into the storage unit 220 as legal data 222.

[0019] The checklist splitting unit 203 is a functional unit that, based on the user's selection, selects a comparison unit for the checklist and splits the text of the checklist data 221 into comparison unit items (details will be described later) according to the selected comparison unit.

[0020] The legal division unit selection unit 204 is a functional unit that selects the legal division unit (details will be described later) that corresponds to the comparison unit of the checklist.

[0021] The legal division unit 205 is a functional unit that divides the text of a law into division items (details will be described later) based on the legal division unit selected in the legal division unit selection unit.

[0022] The similarity calculation unit 206 is a functional unit that calculates the similarity between the text of the comparison unit items in the checklist divided by the checklist division unit 203 and the text of the divided items in the law.

[0023] The similarity item acquisition unit 207 is a functional unit that, for each combination of the text of the comparison unit item in the divided checklist and the text of the divided item in the divided law, calls the similarity calculation unit 206 and acquires the divided item of the law according to the result.

[0024] The checklist-legal information display unit 210 is a functional unit that displays the comparison units of the checklists, which are associated with similarity, and the division items of the laws and regulations on the checklist-legal information screen.

[0025] The memory unit 220 is a functional unit that stores data and programs necessary for the document processing support device 100.

[0026] The memory unit 220 stores checklist data 221, legal data 222, checklist-legal association data 223, and other data such as work data, which is not shown in the diagram.

[0027] The checklist-legal association data 223 is data that associates the text of the comparison unit items in the divided checklist with the text of the divided items in the divided legal provisions. Although not shown in the diagram, the checklist-legal association data 223 may include, for example, numbering the text of the comparison unit items in the divided checklist and the divided items in the divided legal provisions, and storing them as pairs of numbers.

[0028] Next, the hardware and software configuration of the document processing support device 100 will be explained using Figure 2. The hardware configuration of the document processing support device 100 can be implemented using a general information processing device such as the personal computer shown in Figure 2.

[0029] The document processing support device 100 consists of a CPU (Central Processing Unit) 301, main memory 302, network I / F (Interface) 303, display device 304, input / output device 305, and auxiliary storage device 310, all connected by a bus.

[0030] The CPU 301 controls each part of the document processing support device 100 and loads and executes the necessary programs in the main memory 302.

[0031] The main memory 302 is typically composed of volatile memory such as RAM, and stores the programs executed by the CPU 301 and the data it references.

[0032] Network I / F306 is an interface for connecting to a network.

[0033] The display device 304 is a device that displays information, such as an LCD (Liquid Crystal Display), in a form that is visible to the user.

[0034] The input / output device 305 includes input devices such as keyboards and pointing devices such as mice, and output devices such as printers.

[0035] The auxiliary storage device 310 is a storage device with a larger capacity compared to the main storage device 302, such as an HDD (Hard Disk Drive) that stores information on a magnetic storage medium or an SSD (Solid State Drive) that stores information on a semiconductor storage medium.

[0036] The auxiliary storage device 310 of this embodiment stores a program for executing the document processing support process of this embodiment. The document processing support device 100 has the following programs installed: checklist input program 311, legal input program 312, checklist division program 313, legal division unit selection program 314, legal division program 315, similarity calculation program 316, similar item acquisition program 317, and checklist-legal relationship display program 320.

[0037] The checklist input program 311, the legal input program 312, the checklist division program 313, the legal division unit selection program 314, the legal division program 315, the similarity calculation program 316, the similar item acquisition program 317, and the checklist-legal relationship display program 320 are programs for executing the functions of the checklist input unit 201, the legal input unit 202, the checklist division unit 203, the legal division unit selection unit 204, the legal division unit 205, the similarity calculation unit 206, the similar item acquisition unit 207, and the checklist-legal relationship display unit 210, respectively.

[0038] Furthermore, the auxiliary storage device 310 stores checklist data 221, legal data 222, and checklist-legal association data 223.

[0039] (II) Overview of document processing support system, checklist, and specific examples of laws and regulations Next, using Figures 3A and 3B, we will explain the overview of the document processing support device of this embodiment, as well as specific examples of checklists and laws and regulations.

[0040] The document processing support device of this embodiment works as follows: the user specifies a desired comparison unit (1 line, 2 lines, check item, etc.) for a given checklist, divides the law into division units (article, paragraph, item) of a size corresponding to that comparison unit, and displays the text of the division items of the relevant law for the text of the comparison unit items in each checklist.

[0041] When determining the appropriate division unit for the legal provisions based on the specified unit of comparison, the division unit selected is the one whose average size of the legal provision's division items is closest to the average size of the comparison unit items in the checklist divided into the unit of comparison.

[0042] Then, the similarity between the text of the comparison unit items in the divided checklist and the text of the divided items in the law is calculated (details of the similarity calculation will be described later), and the divided items in the law with the highest similarity, or those that exceed a predetermined threshold, are displayed as the divided items related to the checklist.

[0043] Alternatively, a predetermined number of legal division items with a high degree of similarity to the text of the comparison unit items in the divided checklist may be displayed as related.

[0044] In this embodiment, the law in question is the Building Standards Act Enforcement Order (Cabinet Order No. 338 of 1950: Cabinet Order based on the provisions of the Building Standards Act (Act No. 201 of 1950)), and the explanation will take as an example the case in which a construction company creates a checklist used when designing buildings in accordance with it.

[0045] The construction company wishes to create a checklist of 600 items related to landings and stairs, as shown in Figure 3A, and to display the relevant text from the Building Standards Act Enforcement Order.

[0046] Figure 3B shows Articles 24 through 26 of the Building Standards Act Enforcement Order, which may be related to Checklist 600 concerning landings and stairs, as part of the Building Standards Act Enforcement Order.

[0047] (III) Items of division of laws and regulations Next, using Figures 3B, 4A, and 4C, we will explain specific examples of how laws and regulations can be divided when the law in question is the Building Standards Act Enforcement Order, as described above.

[0048] Generally, in legal practice, legal texts are headinged with articles, paragraphs, and subparagraphs. Article numbers are indicated as subparagraph X (where X is a Chinese numeral, sometimes followed by "of Y" (where Y is a Chinese numeral)), paragraph numbers are indicated by a single Arabic numeral, and subparagraph numbers by a single Chinese numeral. Furthermore, it is customary for the first paragraph of a legal text to not have a heading.

[0049] Although not shown in the diagram, information regarding the division units of the law is assumed to be stored in the system's storage tables or similar locations beforehand.

[0050] Figure 4A shows an example where the division items are divided into articles, and in this example, the legal text is divided into articles and presented as a list.

[0051] Figure 4B shows an example where the divided items are divided into individual items, and in this example, the legal text is divided into items and presented as a list.

[0052] When a higher-level division unit (in this example, an article) than the target division unit (in this example, a paragraph) does not have lower-level division units (paragraphs), the division items of the higher-level division unit (article) shall be divided into a single division unit.

[0053] Figure 4C shows an example where the division items are divided into units of each section, and in this example, the legal text is divided into sections and presented as a list.

[0054] When a higher-level division unit (in this example, an article or paragraph) of a law or regulation does not have a lower-level division unit (an article or paragraph), the division items of that higher-level division unit (article or paragraph) shall be divided into a single division unit.

[0055] (IV) Calculation of the similarity between the text of the checklist and the text of the law. Next, we will explain in detail the process of calculating the similarity between the text of the divided comparison unit items in the above checklist and the text of the divided items in the laws and regulations, using the previously shown Figures 3A, 3B, and Figures 5 through 8.

[0056] While there are various existing methods for calculating document similarity, such as those using natural language processing or machine learning, this embodiment describes a method using tf-idf.

[0057] tf-idf (term frequency - inverse document frequency) is a statistical measure that represents "how important each word (term) is within a document." Specifically, it is the product of the tf (term frequency) value, which represents "how often a certain word appears within a given document," and the idf (inverse document frequency) value, which represents "how rarely documents containing a certain word exist" among all documents.

[0058] In calculating the document similarity of this embodiment, the tf-idf value is used for all words between multiple documents to replace the word groups that make up each document with a vector of numerical values ​​(vectorization). Then, the numerically converted vectors for each document are considered to represent the "features of each document" (feature vector), and the cosine similarity between the feature vectors is calculated, and the similarity between the documents is determined from the calculated cosine similarity.

[0059] As examples of documents to be addressed, we will explain using the text 601 for check item 1 and the text 602 for check item 2 shown in Figure 3A, and as examples of texts for divided legal provisions, we will use the text 701 for Article 24, Paragraph 2 and the text 702 for Article 25, Paragraph 1 shown in Figure 3B.

[0060] First, if we perform morphological analysis on the text 601 of check item 1 and the text 602 of check item 2 using a morphological analysis tool such as MeCab, as shown in Figure 5, the morphological analysis results 611 of the text 601 of check item 1 and the text 602 of check item 2 are, respectively, ['landing', 'of', 'tread width', 'is', ', 'one', '·', 'two', 'meter', 'or more', 'is', 'is', 'is'], ['stairs', 'have', 'handrails', 'installed', ''].

[0061] Similarly, the morphological analysis results 711 of text 701 in Article 24, paragraph 2, and 712 of text 702 in Article 25, paragraph 1, are, respectively, ['The 'tread width' of the 'landing' of the 'straight' 'stairs' 'must' be 'one', '·', 'two', 'meters', 'or more', 'must', 'be', 'and', 'must', 'not'], and ['Stairs' must', 'have', 'handrails', 'or']. Note that headings such as "Check Item 1" and "Article 24," as well as spaces, have been removed.

[0062] Here, MeCab (Yet Another Part-of-Speech and Morphological Analyzer) is an open-source morphological analysis engine developed through a joint research unit project between the Graduate School of Informatics, Kyoto University and the Basic Research Laboratories of Nippon Telegraph and Telephone Corporation.

[0063] Next, we will identify all morphemes appearing in the entire document (checklist and Article 24, Paragraph 2 and Article 25, Paragraph 1 of the Building Standards Act Enforcement Order), and deal with 800 morpheme vectors, each corresponding to a single morpheme (word), as shown in Figure 5. In this example, the vector has 32 elements, each corresponding to ['landing', 'of', 'tread width', 'is', 'one', '·', 'two', 'meter', 'or more', 'is', 'a handrail', 'on', 'stairs', 'must', 'install', 'a handrail', 'on', 'as', 'preceding', 'must', 'install', 'right', 'to', 'do', 'must', 'install'].

[0064] Next, calculate the tf values of the text 601 of the check item 1 to be compared and the text 701 of Article 24, Paragraph 2. As already explained, the tf value is a statistical value representing "how frequently a certain word appears within a certain document". For a word t i (i = 1,..., n) in a document d j (j = 1,..., m), the tf value of the word t i is denoted as tf(t i , d j ) and is expressed by the following (Equation 1).

[0065]

Number

[0066] Here, m is the number of documents (in this example, 4), and n is the total number of all words in the m documents (in this example, 32). Also, f(t i , d j ) is the frequency of occurrence t j of the word t contained in the document d i , and the denominator of (Equation 1) is the total number of all words contained in the document d j .

[0067] When expressing the text 601 of the check item 1 as an occurrence frequency vector 801 corresponding to the elements of the vector 800, as shown in Figure 7A, it becomes [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0].

[0068] Similarly, when expressing the text 701 of Article 24, Paragraph 2 as an occurrence frequency vector 811 corresponding to the elements of the morpheme vector 800, as shown in Figure 7B, [1, 3, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0] is obtained.

[0069] On the other hand, the denominator of (Equation 1) is the total number of times all words appear in each document, so in the case of text 601 in check item 1, it is 13, and in the case of text 701 in Article 24, paragraph 2, it is 26.

[0070] Therefore, the vectors of the tf values ​​for text 601 of check item 1 and text 701 of Article 24, paragraph 2, corresponding to the elements of morphological vector 800, are [1 / 13,1 / 13,1 / 13,1 / 13,1 / 13,1 / 13,1 / 13,1 / 13,1 / 13,1 / 13,1 / 13,1 / 13,1 / 13,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0], [1 / 26,3 / 26,1 / 26,1 / 26,1 / 26,1 / 26,1 / 26,1 / 26,1 / 26,1 / 26,0,0,0,1 / 26,1 / 26,0,0,0,0,1 / 26,1 / 26,1 / 26,1 / 26,1 / 26,1 / 26,1 / 26,1 / 26,1 / 26,1 / 26,0].

[0071] Next, we calculate the IDF values ​​for text 601 of check item 1 and text 701 of Article 24, paragraph 2, which are to be compared. As already explained, the IDF value is a statistical value that represents "how infrequently" documents containing a certain word exist "among all documents". i (i=1,…,n), document d j The word t among (j=1,…,m) i The IDF value of idf(t i ,{d j If we assume}, then it can be expressed by the following (Equation 2).

[0072]

number

[0073] Here, Count(t i ,{d j}) is the set of all documents {d j Within}, the word t i This is the number of documents that contain [the specified element].

[0074] In this example, m is 4, word t i If we represent the number of documents containing a word using a vector corresponding to each word, as shown in Figure 8, then the document count vector 900 will be used. [2,2,2,4,2,2,2,2,2,2,1,1,2,3,3,2,2,1,1,1,1,1,1,2,1,1,1,2,2,2,2,1] Therefore, The vector of IDF values ​​corresponding to each word is: [2 / log4,2 / log4,2 / log4,4 / log4,2 / log4,2 / log4,2 / log4,2 / log4,2 / log4,2 / log4,1 / log4,1 / log4,2 / log4,3 / log4,3 / log4,2 / log4,2 / log4,1 / log4,1 / log4,1 / log4,1 / log4,1 / log4,1 / log4,2 / log4,1 / log4,1 / log4,1 / log4,2 / log4,2 / log4,2 / log4,2 / log4,1 / log4].

[0075] Next, we calculate the tf-idf value. Document d j The word t among (j=1,…,m) i The tf-idf value is tf-idf(t i d j If we assume that, it can be expressed by the following (Equation 3).

[0076]

number

[0077] As a result, the vector of tf-idf values ​​corresponding to each word in text 601 of check item 1 is: [2 / 13log4,2 / 13log4,2 / 13log4,4 / 13log4,2 / 13log4,2 / 13log4,2 / 13log4,2 / 13log4,2 / 13log4,2 / 13log4,1 / 13log4,1 / 13log4,2 / 13log4,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0] and so on. Furthermore, the vector of tf-idf values ​​corresponding to each word in text 701 of Article 24, paragraph 2 is: [1 / 13log4,3 / 13log4,1 / 13log4,2 / 13log4,1 / 13log4,1 / 13log4,1 / 13log4,1 / 13log4,1 / 13log4,1 / 13log4,0,0,0,3 / 26log4,3 / 26log4,0,0,0,0,1 / 26log4,1 / 26log4,1 / 26log4,1 / 26log4,1 / 13log4,1 / 26log4,1 / 26log4,1 / 13log4,1 / 13log4,1 / 13log4,0].

[0078] Next, the similarity between text 601 of check item 1 and text 701 of Article 24, paragraph 2 is calculated using the cosine similarity when the respective tf-idf value vectors are used as feature vectors. Since each element of the tf-idf value vector is greater than or equal to 0, the cosine similarity will be between 0 and 1, and the closer it is to 1, the more similar the respective feature vectors are considered to be.

[0079] In general, an N-dimensional vector v = (x1, x2, ..., x N ) and w=(y1,y2,…, y N The cosine similarity CosSim(v,w) is expressed by the following equation (4).

[0080]

number

[0081] Here, (v,w) is the dot product of vector v and vector w.

[0082] Substituting the vector of tf-idf values ​​corresponding to each word in text 601 of check item 1 above into v, and substituting the vector of tf-idf values ​​corresponding to each word in text 701 of Article 24, paragraph 2 into (Equation 4) into w, their cosine similarity is: It can be calculated as 0.705 (rounded to four decimal places).

[0083] Similarly, the vector of tf values ​​corresponding to each word in text 702 of Article 25, paragraph 1 is: [0,0,0,1 / 11,0,0,0,0,0,0,0,0,0,1 / 11,1 / 11,1 / 11,1 / 11,0,0,0,0,0,0,1 / 11,0,0,0,1 / 11,1 / 11,1 / 11,1 / 11,1 / 11] and The vector of tf-idf values ​​is, [0,0,0,2 / 11log4,0,0,0,0,0,0,0,0,0,2 / 11log4,2 / 11log4,2 / 11log4,2 / 11log4,0,0,0,0,0,0,2 / 11log4,0,0,0,2 / 11log4,2 / 11log4,2 / 11log4,2 / 11log4,2 / 11log4].

[0084] Substituting the vector of tf-idf values ​​corresponding to each word in text 601 of check item 1 above into v, and substituting the vector of tf-idf values ​​corresponding to each word in text 702 of Article 25, paragraph 1 into (Equation 4), their cosine similarity can be calculated as 0.0386 (rounded to four decimal places).

[0085] Therefore, the cosine similarity between text 601 of check item 1 and text 701 of Article 24, paragraph 2 is closer to 1 than the cosine similarity between text 601 of check item 1 and text 702 of Article 25, paragraph 1. Thus, it can be determined that text 701 of Article 24, paragraph 2 is more similar to text 601 of check item 1 than text 702 of Article 25, paragraph 1.

[0086] In this embodiment, we have described an example in which the similarity of documents is determined by the cosine similarity of the tf-idf value vector for all morphemes extracted by separating them from the checklist and the laws and regulations. However, there are several variations in the method of calculating similarity, such as preparing a list of legally important words (keywords) in advance and extracting only the words included in the list to perform the similarity calculation, or focusing on the part of speech of words and using only words of a specific part of speech for the similarity calculation. These methods may also be used.

[0087] (V) Details of the document processing support system Next, using Figures 9 and 11, we will explain in detail the process by which the document processing support device displays checklists and related laws and regulations in association with them.

[0088] First, the user selects a checklist to be compared from the checklist data 221 (S101). For example, a checklist 600 concerning landings and stairs, as shown in Figure 3A.

[0089] Next, the user selects the relevant laws and regulations from the legal data 222, matching the checklist selected in S101 (S102). For example, the Building Standards Act Enforcement Order concerning landings and stairs, as shown in Figure 3B. This can be explicitly specified by the user, or the system may maintain a list of checklists and their related legal data, from which the document processing support device can automatically select. Here, it may be possible to select multiple laws and regulations, or a group of laws and regulations containing multiple laws and regulations.

[0090] Next, the user selects the comparison unit for the selected checklist (1 row, 2 rows, check items, etc.) (S103).

[0091] Next, the document processing support device 100 divides the checklist into comparison unit items according to the comparison unit of the checklist selected in S103 (S104).

[0092] Next, the document processing support device 100 performs legal division processing on the selected checklist and related laws and regulations (S105). Details of the legal division processing will be explained later using Figure 10.

[0093] Next, the document processing support device 100 performs checklist-law relationship processing for each comparison unit item of the checklist and each text-law relationship item based on the results of the law division processing in S105 (S106). Details of the checklist-law relationship processing will be explained using Figure 11.

[0094] Next, the document processing support device 100 displays the checklist-legal association screen based on the result of S106 (S107). The specific user interface of the checklist-legal association screen will be described later.

[0095] Next, we will explain the details of the legal division process using Figure 10. This process corresponds to S105 in Figure 9, and involves inputting a list of text for the comparison unit items of a checklist and the text of the laws and regulations related to that checklist, and returning them as a list of segmented items.

[0096] First, the document processing support device 100 obtains the text size of the comparison unit items in the checklist as the comparison unit and calculates the average value (S201).

[0097] Here, "text size" refers to, for example, the number of characters in the text. Alternatively, it could refer to other criteria, such as the number of morphemes analyzed by a morphological analysis tool like MeCab mentioned above.

[0098] Next, the document processing support device 100 repeats the processing in S203 and S204 for all legal division units defined in the system (S202-S205).

[0099] In the S202-S205 loop, the document processing support device 100 selects a division unit for one law (S202) and divides the law into the selected division units (S203).

[0100] Next, the document processing support device 100 calculates and stores the average size of the divided items of the divided laws and regulations (S204).

[0101] Exiting the S202-S205 loop, the document processing support device 100 selects a division unit such that the average size of the divided checklist text calculated in S201 is closest to the average size of the divided items in the law (S206).

[0102] Next, the document processing support device 100 returns a list of division items for the division unit selected in S206 (S207). The return value of the list of division items can be returned to the Call side as, for example, a String array.

[0103] Next, we will explain the details of the checklist-legal association process using Figure 11. First, the document processing support device 100 selects one of the divided texts from the divided checklist and repeats the process from S302 to S307 (S301 to S308).

[0104] In the S301-S308 loop, the document processing support device 100 selects one of the legal division items for all of the legal division items and repeats the processing in S303-S306 (S302-S307).

[0105] In the loop from S302 to S307, the document processing support device 100 performs morphological analysis on the selected divided checklist text and the selected divided items of the law (S303).

[0106] Next, the document processing support device 100 calculates a vector of tf-idf values ​​for each word in the selected divided checklist text and the divided items of the selected laws and regulations, based on the morphological analysis in S303 (S304, (Equation 1) to (Equation 3)).

[0107] Next, the tf-idf value vectors of each word calculated in S304 are used as feature vectors, and their cosine similarity is calculated (S305).

[0108] Next, the document processing support device 100 sets pairs of divided checklist text and divided legal items whose cosine similarity calculated in S305 is above a certain threshold (for example, 0.7 or higher) as associated with the checklist-legal association data 223 (S306).

[0109] Alternatively, instead of S306, the cosine similarity of the pairs of divided checklist text and legal division items may be stored, and the one with the highest value (closest to 1) may be selected between S307 and S308. The corresponding checklist text and legal division item pairs may then be set as related to the checklist-legal association data 223.

[0110] (VI) User interface provided by the document processing support system Next, we will explain the user interface of the document processing support device using Figure 12. The Checklist-Legal Association screen is a utility screen that allows users to select checklists, select laws, select the comparison unit for checklists, and instruct the checklist-legal association process. It also displays the text of the selected checklists, the selected laws, and the results of the checklist-legal association process.

[0111] The checklist-legal association screen 1000, as shown in Figure 12, includes a checklist selection list 1001, a checklist display area 1002, a checklist comparison unit selection list 1003, a legal selection list 1004, a legal display area 1005, an association start button 1006, and an association result display area 1007.

[0112] The checklist selection list 1001 is an item that displays checklists stored in the checklist data 221 held by the document processing support device as a drop-down list or similar, allowing the user to select one. Alternatively, methods such as displaying a file selection window could be considered instead of a drop-down list. Furthermore, the user may be able to retrieve and configure data from an external source.

[0113] The checklist display area 1002 is the area that displays the contents of the checklist selected in the checklist selection list 1001. In the example shown in Figure 12, the text of checklist 600 concerning landings and stairs shown in Figure 3A is displayed.

[0114] The checklist comparison unit selection list 1003 is an item that displays the set comparison unit (1 row, 2 rows, check items, etc.) in a dropdown list or similar format, allowing the user to select one. Depending on the checklist comparison unit selected by the user, the display text in the checklist display area may be divided accordingly.

[0115] The legal selection list 1004 is an item that displays the legal provisions stored in the legal provision data 222 as a dropdown list or similar, allowing the user to select one. Alternatively, a file selection window could be displayed instead of a dropdown list. The user could also be allowed to retrieve and configure data from an external source. To support the selection of multiple legal provisions for comparison, an item such as checkboxes could be used.

[0116] The legal information display area 1005 is the area that displays the content of the legal information selected in the legal information selection list 1004. In Figure 12, parts of Articles 24 to 26, paragraph 700 of the Building Standards Act Enforcement Order are shown.

[0117] The association start button 1006 is a button that takes as input the content of the checklist selected in the checklist selection area 1002, a list of text obtained by dividing the content of the checklist into comparison units selected in the checklist and comparison unit selection area, and the content of the law selected in the law selection list 1004, and instructs the user to display each association.

[0118] The association result display area 1007 is an area that displays the association results between the checklist and the laws and regulations executed when the association start button 1006 is clicked. Alternatively, as a method of displaying the results, an interface may be provided in which clicking an item in the checklist in the checklist display area 1002 displays items of laws and regulations similar to the clicked checklist item in the laws and regulations display area 1005.

[0119] As described above, the document processing support device of this embodiment appropriately selects a method for dividing the law to be compared according to the size of each text obtained by dividing the checklist into specified comparison units, calculates the similarity between each item of the divided checklist and the law text using natural language processing, and presents the section of the law that is similar to the checklist to the user, thereby allowing the user to efficiently check the degree to which the checklist corresponds to the law and to identify any omissions or errors. This enables the creator of the checklist to create a highly accurate checklist that complies with the law and does not contain any omissions or errors, and allows those who perform checks using the checklist to efficiently refer to and verify the relevant articles of the law.

[0120] Furthermore, in this embodiment, the documents being compared are described as a checklist and a law, but the documents may be other texts. For example, they may be checklists for employees regarding company manuals for business operations, or checklists for employees regarding quality control manuals for product assembly in manufacturing.

[0121] In this embodiment, the division units for laws and regulations were described as articles, paragraphs, and subparagraphs, but it is also conceivable to use document division units that are appropriate to the characteristics of the text. For example, chapters, sections, and subparagraphs may be used.

[0122] [Embodiment 2] A second embodiment of the present invention will be described below with reference to Figures 3B and 13.

[0123] In the document processing support device of Embodiment 1, the relationship between the checklist and the law has been determined by dividing the content of the law as is and examining the similarity between the divided text of the checklist and the law.

[0124] However, the common format for writing laws and regulations is to refer to other articles. This embodiment corresponds to such a format of writing that refers to other articles.

[0125] In this embodiment, when one provision of a law or regulation refers to another provision of that law or to a provision of another law or regulation, it is possible to insert the text of the referenced provision and generate a text of the law or regulation that is equivalent in content. In describing this embodiment, the differences will be explained based on the configuration and processing of the document processing support device of Embodiment 1.

[0126] For example, as shown in Figure 3B, Article 25, Paragraph 4 of the Building Standards Act Enforcement Order states, "The provisions of the preceding three paragraphs shall not apply to the portion of stairs that is one meter or less in height." Here, "the preceding three paragraphs" is the reference part. Therefore, the document processing support device of this embodiment inserts the text of Article 25, Paragraphs 1 to 3, and generates the following text for Article 25, Paragraph 4 of the Building Standards Act Enforcement Order: "The provisions that 'stairs must be equipped with handrails,' 'side walls or equivalents must be provided on both sides of the stairs and their landings (excluding the side on which a handrail is provided),' and 'if the width of the stairs exceeds three meters, a handrail must be provided in the middle; however, this shall not apply to the portion of stairs that is one meter or less in height, if the rise is 15 centimeters or less and the tread is 30 centimeters or more.' are not applicable to the portion of stairs that is one meter or less in height."

[0127] Furthermore, Article 26, Paragraph 2 of the Building Standards Act Enforcement Order states, "The provisions of the preceding three articles (excluding the parts concerning risers and treads) shall apply mutatis mutandis to the ramp referred to in the preceding paragraph," so the referenced parts are "the preceding three articles" and "the preceding paragraph." Therefore, text is generated in which the text of Article 23 (not shown) (and the articles that Article 23 further references), Article 24, and Article 25 of the Building Standards Act Enforcement Order, which correspond to the preceding three articles of Article 26, and the text of Article 26, Paragraph 1 of the Building Standards Act Enforcement Order, which corresponds to the preceding paragraph of Article 26, Paragraph 2, are inserted.

[0128] In the process of displaying a checklist and related laws and regulations by the document processing support device of this embodiment, as shown in Figure 13, a process of inserting reference text for laws and regulations is performed between S102 and S103 of the process shown in Figure 9 of Embodiment 1 (S110), and the text of laws and regulations generated in S110 is used as the target for subsequent text processing.

[0129] According to this embodiment, the checklist and the text of the law can be associated with each other, taking into account the format of the reference description in the law.

[0130] [Embodiment 3] Hereinafter, Embodiment 3 of the present invention will be described with reference to Figure 14.

[0131] The first embodiment of the document processing support device demonstrates an example of associating checklists with laws and regulations using a standalone information processing device, such as a PC.

[0132] In this embodiment, the method for associating checklists with laws and regulations is the same, but the system configuration is different, and this is done using a cloud system (document processing support system).

[0133] As shown in Figure 14, the document processing support system of this embodiment consists of a document processing support device 100, a similarity calculation device 110, an internet connection 10, an intranet 20, a firewall 15, and a user terminal 5.

[0134] The document processing support device 100, similar to Embodiment 1, is a device that provides the user with the creation of a checklist for checking laws and regulations, and utility functions for associating the checklist with the laws and regulations.

[0135] In this embodiment, the document processing support device 100 calculates the similarity of documents using the functions of an external similarity calculation device 110. Of course, the document processing support device 100 may also include a function for calculating the similarity of documents, similar to Embodiment 1.

[0136] User terminal 5 is a device such as a smartphone, tablet, laptop PC, or desktop PC used by the user to create checklists and input commands and data for comparing checklists with legal regulations. It is assumed that user terminals are equipped with a standard web browser. Note that there may be multiple user terminals 5, such as different terminals used by each user.

[0137] The example shown in Figure 14 illustrates a case where the document processing support device 100 and the similarity calculation device 110 user terminal 5 are connected via an intranet 20. The intranet 20 connection can be either wired or wireless, such as Wi-Fi.

[0138] The intranet 20 is a network that connects the Internet 10 via a document processing support device 100, a similarity calculation device 102, and a firewall 15, enabling data communication between devices.

[0139] Furthermore, the intranet 20 is connected to the external internet 10 via a firewall 15. Here, the firewall 15 is a device or software that connects the internet 10 and the intranet 20, forwarding messages to the appropriate device depending on the content of the communication, or blocking inappropriate communications.

[0140] According to the document processing support system of this embodiment, users can verify the relationship between checklists and laws and regulations by using cloud services. [Explanation of Symbols]

[0141] 5...User terminal, 15...Firewall, 10...Internet, 20...Intranet, 100...Document processing support device, 110...Similarity calculation device 201...Checklist input section, 202...Legal information input section, 203...Checklist division section, 204...Legal information division unit selection section, 205...Legal information division section, 206...Similarity calculation section, 207...Similar item acquisition section, 210...Checklist-Legal information display section, 220...Storage section, 221...Checklist data, 222...Legal data, 223...Checklist-Legal association data

Claims

1. A document processing support device that outputs the relationships between multiple documents, The system accepts input to select the comparison unit for the first document, and then divides the first document into comparison unit items according to the entered comparison unit. It retains information about multiple division units for the second document, The average size of the text for each comparison unit item in the first document is calculated. For each of the multiple division units of the second document, the second document is divided into multiple division units, and the average size of the divided items obtained by dividing the second document into those division units is calculated. The division unit for the second document is selected that is closest in size to the average size of the text for each comparison unit item in the first document and the average size of the division items obtained by dividing the second document into division units. The similarity between the text of the comparison unit item in the first document and the division items divided into division units for the selected second document is calculated, respectively. A document processing support device characterized by associating and outputting pairs of the text of the comparison unit item of the first document, which has a high degree of similarity between the two, with the division items of the second document, which has been divided into division units.

2. The document processing support device according to claim 1, characterized in that the first document is a checklist.

3. The document processing support device according to claim 1, characterized in that the second document is a law or regulation.

4. The document processing support device according to claim 1, characterized in that the similarity is calculated by the cosine similarity between a vector of tf-idf (term frequency - inverse document frequency) values ​​for the words in the first document and a vector of tf-idf values ​​for the words in the second document.

5. The document processing support device according to claim 3, characterized in that, when there are referenced provisions of the aforementioned laws and regulations, it replaces the text of the referenced provisions with the text of the referenced provisions to generate a new document, divides the second document into multiple division units for each of the multiple division units of the second document, and calculates the average size of the division items into which the second document has been divided into those division units.

6. A document processing support method using a document processing support device that outputs relationships between multiple documents, The process involves accepting input to select the comparison unit for the first document, and dividing the first document into comparison unit items according to the entered comparison unit, A step of storing information about multiple division units for the second document, A step of calculating the average size of the text for each comparison unit item in the first document, The steps include dividing the second document into multiple division units and calculating the average size of the division items obtained by dividing the second document into those division units, The steps include selecting the division unit for the second document that has the closest average size of the division items obtained by dividing the second document into division units, based on the average size of the text for each comparison unit item in the first document, A step of calculating the similarity between the text of the comparison unit item in the first document and the division items divided into division units for the selected second document, A document processing support method characterized by having the step of associating and outputting a pair of divided items, which have been divided into division units for the second document, with the text of the comparison unit item of the first document, which has a high degree of similarity between the two.

7. In the step of calculating the similarity, The document processing support method according to claim 6, characterized in that the similarity is calculated by the cosine similarity between a vector of tf-idf (term frequency - inverse document frequency) values ​​for words in the first document and a vector of tf-idf values ​​for words in the second document.

Images