Text data management method, text data management system

Abstract

To provide a sentence data management method and a sentence data management system for reducing a law violation risk accompanied by external transmission of document data by associating the document data with articles of a regulation.SOLUTION: A sentence data management system is provided with a processor. The processor converts a word string 1 of a first document described by a first term being a term to be used in a sentence for determining a regulation into a word string 2 corresponding to the word string 1 described by a third term being a term different from the first term, converts the word string 3 of a second document described by the second term being a term to be used in a sentence to be externally transmitted into a word string 4 corresponding to the word string 3 described by the third term, compares the first document with the second document by determining a comparison result between the word string 1 and the word string 3 with a result obtained by comparing the word string 2 with the word string 4, and checks the possibility of a violation of regulations in external transmission to determine external transmission of the second document.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a text data management method and a text data management system. [Background technology]

[0002] Advances in cloud technology are driving the adoption of hybrid cloud configurations that combine public clouds with in-house built private clouds. Hybrid clouds allow optimal data utilization by selectively using both clouds depending on the characteristics of the data, processing, and computing resources.

[0003] In recent years, laws and regulations have strengthened demands for data control over personal information, technical information, confidential information, and other data. For example, laws such as the General Data Protection Regulation (GDPR) for personal information and the U.S. Re-export Control Regulations for technical information restrict the international transmission of data containing such information. As these laws and regulations are becoming increasingly stringent, it is expected that companies will increasingly implement data control measures to protect against the risk of violating the laws and regulations. However, implementing appropriate data control can be a burden for companies. For example, determining whether certain technical information can be transmitted internationally requires knowledge of both the technical field and the relevant laws, which increases the training and operational costs of personnel.

[0004] Therefore, methods have been proposed for data management to help comply with these laws and regulations. The method disclosed in Non-Patent Document 1 scans the contents of data stored in storage on a public cloud, and if it finds a text pattern that is likely to be personal information, it notifies the administrator and prompts them to take further action. This method is set up with detection patterns specifically for personal information, and does not comply with other laws and regulations.

[0005] Patent Document 1 proposes a method for searching for similar documents from document data. This method uses legal texts as similar documents to be searched, which could potentially be used to link documents to legal texts. However, document data containing technical or confidential information is likely to use a lot of specialized or proprietary terminology. Meanwhile, legal texts often use legal terminology, resulting in terminology discrepancies. This can lead to inappropriate searches and failed linking. Additionally, the method does not consider the hierarchical structure and reference relationships inherent in legal texts when linking documents. [Prior art documents] [Non-patent literature]

[0006] [Non-Patent Document 1] https: / / cloud.netapp.com / cloud-compliance [Patent documents]

[0007] [Patent Document 1] Japanese Patent Application Publication No. 11-110395 Summary of the Invention [Problem to be solved by the invention]

[0008] Therefore, the present invention aims to provide a text data management method and a text data management system that links document data with the provisions of rules and reduces the risk of violating rules, such as laws and regulations, when sending document data externally. [Means for solving the problem]

[0009] According to a first aspect of the present invention, there is provided a text data management method as follows. The text data management method is implemented using a computer. The text data management method converts a word sequence 1 in a first document written in a first term, which is a term used in text defining rules, into a word sequence 2 corresponding to the word sequence 1 written in a third term, which is a term different from the first term; converts a word sequence 3 in a second document written in a second term, which is a term used in text to be transmitted externally, into a word sequence 4 corresponding to the word sequence 3 written in the third term; and uses the result of comparing word sequence 2 with word sequence 4 as the result of comparing word sequence 1 with word sequence 3, thereby comparing the first document and the second document, checking for the possibility of a violation of the rules in the external transmission, and determining whether to transmit the second document externally.

[0010] According to a second aspect of the present invention, there is provided a text data management system as follows: The text data management system includes a processor. The processor executes a rule-text linking program to (1) convert word sequence 1 of a first document written in a first term, which is a term used in the text defining the rule, into word sequence 2 corresponding to word sequence 1 written in a third term, which is a term different from the first term; (2) convert word sequence 3 of a second document written in a second term, which is a term used in the text to be transmitted externally, into word sequence 4 corresponding to word sequence 3 written in the third term; (3) compare the first document with the second document by using the result of comparing word sequence 2 with word sequence 4 as the result of comparing word sequence 1 with word sequence 3; and (4) in the execution of a text external transmission program, check for the possibility of a violation of the rule in the external transmission and decide to transmit the second text externally. [Effects of the Invention]

[0011] According to the present invention, a text data management method and a text data management system are provided that link document data with the clauses of rules and reduce the risk of violating rules, such as laws and regulations, when document data is sent externally. [Brief explanation of the drawings]

[0012] [Figure 1]FIG. 1 is a configuration diagram of a computer system in a first embodiment. [Figure 2] FIG. 2 is a diagram illustrating an example of the configuration of a rule / document linking computer in the first embodiment. [Figure 3] FIG. 2 is a diagram illustrating an example of the configuration of a document transmission management computer in the first embodiment. [Figure 4] FIG. 2 is a diagram illustrating an example of the configuration of an external transmission approval processing computer in the first embodiment. [Figure 5] 3 is an example of a term conversion dictionary in the first embodiment. [Figure 6] 3 shows examples of rules and documents in the first embodiment. [Figure 7] 4 is an example of rule information in the first embodiment. [Figure 8] 10 is a flowchart showing a process for linking a rule with a document in the first embodiment. [Figure 9] 10 is a flowchart showing a process flow for transmitting a document to an external device in the first embodiment. [Figure 10] 10 is an example of an approval screen for external transmission of a document in the first embodiment. [Figure 11] FIG. 10 is a diagram showing a procedure for building a term conversion dictionary in the second embodiment. [Figure 12] 13 is a flowchart showing a process for linking a rule with a document in the third embodiment. [Figure 13] 13 is an example of a rule including a hierarchical structure and a reference relationship in the fourth embodiment. [Figure 14] 13 is a flowchart showing a process for linking a rule with a document in the fourth embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0013] A first embodiment will be described with reference to FIG. 1. FIG. 1 shows a configuration diagram 100 of a computer system to which this embodiment is applied. A computer system 110 (a document data management system) can be configured to include multiple computers installed at a certain location. In this embodiment, the computer system 110 includes a storage 120, an external transmission approval processing computer 150, a document transmission management computer 160, and a rule / document linking computer 170. Note that in FIG. 1, an external computer system 180 is a computer system installed at a location that is logically or physically different from the computer system 110. The external computer system 180 may be, for example, a computer system of another company, a computer system of an overseas department installed overseas, or a public cloud.

[0014] The storage 120 is a device that stores the document 130. The storage 120 can be configured using a known storage medium, such as a hard disk drive (HDD), a solid state drive (SSD), an optical disk, a magnetic disk, or a magnetic tape. The document 130 stores sentences and text information. The document 130 can be stored in a file, object, database, or other format. Therefore, the storage 120 can be configured using the above-mentioned storage medium to take on a form that matches the storage format of the document 130, such as file storage, object storage, or a relational database management system (RDBMS). The rule information 140 is information that the rule-document linking calculator 170 assigns to the document 130 based on the information stored in the document 130, and stores information about the rules linked to the document 130. The rule information 140 stores, for example, information listing the relationships between specific pages or paragraphs of the document 130 and which clauses of the rules are linked to the document 130. Although the rule information 140 is stored in the storage 120 in FIG. 1, it may be stored in the rule / document linking computer 170 or the document transmission management computer 160.

[0015] The rule-document linking calculator 170 analyzes the text stored in the document 130 and links the rule information 140 to the document 130 according to its content.

[0016] When a user 195 of the computer system 110 wishes to send a document 130 stored in the storage 120 to an external computer system 180, the document transmission management computer 160 receives the transmission request and determines whether or not the transmission is permitted. At that time, the document transmission management computer 160 may inquire of the administrator 190 of the computer system 110 via the external transmission approval processing computer 150 about whether or not the transmission is permitted.

[0017] The external transmission approval processing computer 150 presents information about the document 130 for which a transmission request has been made to the document transmission management computer 160 and the associated rule information 140 to the administrator 190, and receives approval or rejection of the transmission request from the administrator 190.

[0018] Next, a specific configuration of the rule-document linking computer 170 will be described with reference to FIG. 2. FIG. 2 shows the rule-document linking computer 170. The rule-document linking computer 170 includes a CPU 210, a memory 220, and a network interface 230. The CPU 210 determines the operation of the rule-document linking computer 170 in accordance with various programs (a rule-sentence linking program 221 and a term conversion dictionary creation program 225) stored in the memory 220. The memory 220 stores the rule-sentence linking program 221, rule data 222, a rule-specific term conversion dictionary 223, a document-specific term conversion dictionary 224, and a term conversion dictionary creation program 225. The rule-sentence linking program 221 is a program used to assign rule information 140 to documents 130 stored in the storage 120. The rule data 222 includes information on various rules (such as the name of the rule, the contents of the clauses, and hierarchical relationships) referenced in the processing of the rule-sentence linking program 221. The rule-oriented term conversion dictionary 223 is data used in the processing of the rule-sentence linking program 221, and can be used as a dictionary to convert words and phrases stored in the rule data 222. The document-oriented term conversion dictionary 224 is data used in the processing of the rule-sentence linking program 221, and can be used as a dictionary to convert words and phrases stored in the document 130. The term conversion dictionary generation program 225 is a program used to generate the rule-oriented term conversion dictionary 223 or the document-oriented term conversion dictionary 224 (if both conversion dictionaries 223 and 224 are programs, the data used by those programs) from the rule data 222, the document 130, or other external text data such as books, newspapers, and websites. The network interface 230 is an interface for communicating with other computers and the storage 120 within the computer system 110. Protocols widely used for communication between computers, such as Ethernet, TCP / IP, and HTTP, can be used for communication with the computer.For communication with storage 120, protocols that send and receive data on a file-by-file basis, such as NFS (Network File System) and SMB (Server Message Block), protocols that send and receive data on an object-by-object basis, such as REST (Representation Server Transfer), and protocols that send and receive data to and from databases, such as ODBC (Open Database Connectivity), can be used.

[0019] The above-mentioned processes are executed by the CPU 210 (i.e., the processor) in accordance with the procedures written in the various programs. This also applies to the following descriptions of the computer, and the CPU (processor) is the main body of program processing.

[0020] The locations where the various data (221-225) are stored are not particularly limited as long as they can execute the predetermined processes, and can be determined appropriately. An example of the data storage location will now be described. In the example of FIG. 2, the rule-document linking computer 170 stores the rule data 222, the rule-specific term conversion dictionary 223, and the document-specific term conversion dictionary 224 in the memory 220. However, these data may be stored in other locations as long as they can be referenced by the CPU 210. The rule-document linking computer 170 may, for example, have a storage medium in the memory 220, and data may be stored in the storage medium. The rule-document linking computer 170 may also acquire these data from inside or outside the computer system 110 via the network interface 230. For example, a storage medium may be attached to the rule-document linking computer 170, and data may be stored in the storage medium.

[0021] Next, a specific configuration of the document transmission management computer 160 will be described with reference to FIG. 3. FIG. 3 shows the document transmission management computer 160. The document transmission management computer 160 includes a CPU 310, a memory 320, and a network interface 330. The CPU 310 and the network interface 330 have the same roles (functions) as the CPU 210 and the network interface 230, respectively, of the rule / document linking computer 170. The memory 320 stores a document external transmission program 321, a rule / destination information table 322, and an external transmission history table 323. When a user 195 wishes to send a document 130 to an external computer system 180, the document external transmission program 321 is a program that receives a transmission request from the user 195 and executes the transmission based on the results of a rule / document linking process flow 800, which will be described later. The rule / destination information table 322 is a table that lists the correspondence between rule clauses and the external computer systems 180 to which the document 130 corresponding to the clauses can be sent. The external transmission history table 323 is history information relating to documents that have been transmitted or whose transmission has been cancelled by the document external transmission program 321 in the past.

[0022] Next, a specific configuration of the external transmission approval processing computer 150 will be described with reference to FIG. 4. FIG. 4 shows the external transmission approval processing computer 150. The external transmission approval processing computer 150 comprises a CPU 410, a memory 420, a network interface 430, and a display 440. The CPU 410 and the network interface 430 have roles (functions) equivalent to those of the CPU 210 and the network interface 230 of the rule / document linking computer 170, respectively. The memory 420 stores an external transmission approval processing program 421. The external transmission approval processing program 421 operates when the document transmission management computer 160 executes the document external transmission program 321 to inquire of the administrator 190 about whether or not the document 130 can be externally transmitted. The display 440 is used by the external transmission approval processing computer 150 to display information to the administrator 190 by executing the external transmission approval processing program 421. However, the structure and layout of the display 440 are not particularly limited as long as it is capable of displaying information to the administrator 190. The display 440 does not necessarily have to be built into the external transmission approval processing computer 150. For example, by executing the external transmission approval processing program 421, information required for screen display may be sent via the network interface 430 via a protocol such as HTTP (Hypertext Transfer Protocol), and another computer that receives the information may display the screen for the administrator 190.

[0023] Next, the term conversion dictionary will be described with reference to Fig. 5. Fig. 5 shows examples of the rule-oriented term conversion dictionary 223 and the document-oriented term conversion dictionary 224. Both are expressed in the form of arranging pairs of source expressions (510, 560) and destination expressions (520, 570).

[0024] The source notation 510 in the rule-oriented term conversion dictionary 223 is an excerpt of a notation that appears in a rule. The corresponding destination notation 520 indicates the notation to which the notation that matches the source notation 510 that appears in the rule is converted. For example, conversion entry 531 indicates that if the notation "equipment having a calculation function" appears in the rule data 522, it should be converted to the notation "calculator." The destination notation 520 may match in multiple conversion entries. For example, the conversion entries (531, 532, 533) all have the same destination notation 520, "calculator." This means that the "equipment having a calculation function," "server," and "computer" written in the source notation 510 can all be considered to be the same "calculator." The destination notation 520 does not have to be an expression written in natural language. For example, it may be replaced by an ID number assigned to an expression, such as Wordnet (see https: / / wordnet.princeton.edu / ), or by a representation of words in the form of a numerical vector, such as Word2Vec (see https: / / arxiv.org / abs / 1301.3781). Furthermore, the correspondence between source representation 510 and destination representation 520 may be described as a rule or algorithm for generating destination representation 520 from source representation 510, other than the form of listing conversion entries in one-to-one correspondence as shown in the example of Figure 5.

[0025] The source notation 560 and the destination notation 570 in the document-oriented term conversion dictionary 224 are similar to the source notation 510 and the destination notation 520 in the rule-oriented term conversion dictionary 223, except that they are targeted at documents rather than rules.

[0026] The source representations (510, 560) use the vocabulary to which they are applied. For example, the source representation 510 mainly uses terms that appear in rules, and the source representation 560 mainly uses terms that appear in documents. The destination representations (520, 570) use common terms that are not biased towards any particular rule or document. The common terms can be, for example, general terms. This is intended to eliminate differences in vocabulary between rules and documents.

[0027] An example of a rule and a document is now described. Figure 6 shows an example of a rule 600 and a document 650.

[0028] The rule 600 includes legal terms and is expressed as a list of clauses. The legal terms can be considered as an example of first terms, which are terms used in sentences that define the rules, and the rule 600 can be considered as an example of a first sentence written in the first terms. Figure 6 illustrates two clauses (clause 610 and clause 620) of the rule 600. Each clause is expressed using natural language, mainly using words, phrases, and sentences. Chemical formulas and symbols can also be used as long as they can be converted into natural language notation.

[0029] In addition, the rules 600 may include clauses about the recipients 630 to which they apply. The example in Figure 6 includes a clause indicating that the rules 600 are rules governing the provision of information to country A. This information is extracted by the document transmission management computer 160 and stored in advance in the rules and recipient information table 322.

[0030] Document 650 is expressed as a list of paragraphs. The example in Figure 6 shows a sentence containing technical terms, and paragraphs (660, 670) are shown. The technical terms can be considered as an example of second terms, which are terms used in sentences to be sent externally, and sentence 650 can be considered as an example of a second sentence written using the second terms. Each paragraph is expressed in natural language, mainly using words, phrases, and sentences. Chemical formulas and symbols can also be used as long as they can be converted into natural language notation.

[0031] Next, the rule information will be described with reference to Fig. 7. Fig. 7 shows an example of rule information 140 that is assigned to the document 130.

[0032] The rule information 140 is in the form of an entry (720, 721) that lists an in-document position 710, a rule 711, a clause 712, a score 713, and term correspondence information 714. Each entry indicates information about the link between a part of the document and a clause of the rule.

[0033] The document location 710 indicates location information within the document indicated by the entry. Examples of location information that can be used include page numbers, paragraph numbers, chapters, sections, paragraphs, and sentences. The rule 711 and clause 712 indicate information that identifies the rule and its clause indicated by the entry. The score 713 stores a value indicating the strength of association (e.g., relevance or similarity) between the document indicated by the entry and the clause of the rule. The term correspondence information 714 indicates information on word pairs that match between words in the document and words in the clause of the rule as a result of using the rule-oriented term conversion dictionary 223 and the document-oriented term conversion dictionary 224. The term correspondence information 714 is auxiliary information and may not be necessary when configuring the rule information 140.

[0034] 8 shows a processing flow 800 for linking rules and documents. The rule-document linking computer 170 executes the flow 800 in accordance with the rule-document linking program 221.

[0035] In flow 800, loop 810 is a process that targets only rules. Therefore, loop 810 and steps 820 and after may be performed at different times. For example, loop 810 can be performed in advance when the document to be processed is not yet ready. Furthermore, when flow 800 is performed on multiple documents, the results of loop 810 may be shared among each document.

[0036] In loop 810, the rule-document linking calculator 170 processes each rule stored in the rule data 222. In step 811, preprocessing is performed on each clause of the rule, which depends on the language used. Examples of preprocessing include removing unnecessary symbols, dividing rules into paragraphs or pages, morphological analysis for Japanese, normalization of kana, kanji, and alphanumeric characters, and case normalization and lemmatization for English. Furthermore, if the rule contains content written in a format other than natural language, such as a diagram or table, the processing also includes extracting the description and caption as natural language. In step 812, if a word in the rule clause matches the source expression 510 in the rule-specific term conversion dictionary 223, it is replaced with the corresponding destination expression 520 of the corresponding entry, and a word string (word string 2) using common terms (third terms) is generated.

[0037] Steps 820 and onward in flow 800 are processes targeted at documents. In step 820, the rule-document linking calculator 170 performs preprocessing on the document. The contents of this preprocessing are the same as the process performed on each clause of the rule in step 811. In step 830, if there is a description in the document that matches the source notation 560 in the document-oriented terminology conversion dictionary 224, it is replaced with the corresponding destination notation 570 of the relevant entry, and a word string (word string 4) using a common term (third term) is generated.

[0038] In loop 840, the rule-document linking calculator 170 processes the document processed in steps 820 and 830 for each clause of the rule processed in loop 810. In step 841, a word string using common terms generated from the rule clause obtained in step 812 is compared with a word string using common terms generated from the document obtained in step 830, and a score (a value equivalent to relevance or similarity) is calculated. Existing algorithms can be used to compare word strings. For example, precision and recall calculations in N-grams or BLEU (BiLingual Evaluation Understudy) can be used. Furthermore, if numerical vectors are used as the destination expressions (520, 570) in the rule-oriented term conversion dictionary 223 and the document-oriented term conversion dictionary 224, the distance between the numerical vectors can also be used. This value may also be calculated while weighting each word. For example, one calculation method is to lower the weights for parts of speech that do not have significant meaning in a sentence, such as particles and prepositions, or for frequently appearing words such as "a" and "this." It is also possible to store the comparison results for each word, which indicate which words in the clauses of the rules and which words in the document have a particularly high degree of relevance or similarity.

[0039] In step 850, the rule-document linking calculator 170 regards the clauses of the rules that are ranked in the top few or that show a value equal to or greater than a threshold value among the scores obtained by comparing each clause of the rule with the document in loop 840 as clauses linked to the document. This result is stored as an entry in the rule information 140 for each clause. In this case, the score 713 stores a value corresponding to the relevance or similarity calculated in step 841, and the term correspondence information 714 stores information on word pairs that match as a result of using the rule-oriented term conversion dictionary 223 and the document-oriented term conversion dictionary 224 for the words in the document and the words in the clauses of the rules, obtained in step 841.

[0040] FIG. 9 shows a document external transmission processing flow 900. The document transmission management computer 160 takes the lead in executing the flow 900 during the process up until the user 195 of the computer system 110 transmits the document 130 to the external computer system 180. The document transmission management computer 160 executes the document external transmission program 321 in the flow 900. Here, the processing related to the flow 900 may be triggered, for example, by the following triggers. One trigger may be when the user 195 instructs transmission to the external computer system 180, causing all steps of the flow 900 to be executed. Another trigger may be when step 810 is executed when the document 130 is stored in the storage 120 for the first time, and when the user 195 instructs transmission to the external computer system 180 in a subsequent step, causing the processing related to the flow 900 to be executed. As another trigger, steps up to step 930 may be completed before the user 195 issues an instruction to send to the external computer system 180, and steps 940 and after may be executed when the user 195 issues an instruction to send. Below, the processing of flow 900 will be explained using as a representative example the first case mentioned above, in which all steps are executed when the user 195 issues an instruction to send to the external computer system 180.

[0041] First, the user 195 instructs the storage 120 to transmit the document 130 in the storage 120 to the external computer system 180 via the document transmission management computer 160. This starts flow 900. In step 910, the document transmission management computer 160 requests the rule / document linking computer 170, which then executes flow 800, causing the rule / document linking computer 170 to assign rule information 140 to the document 130.

[0042] In the following step 920, the document transmission management computer 160 compares the rule information 140 with the rule / destination information table 322 and the location of the destination external computer system 180 to determine whether the rule information 140 contains any rules that restrict the sending of documents to the external computer system 180. If the rule information 140 contains multiple rule clauses, it is considered to contain at least one restrictive clause. If in step 920 there are no rules that restrict document sending to the external computer system 180, in step 950 the document transmission management computer 160 sends the document 130 to the external computer system 180. Then, the process proceeds to step 970, where the process of step 970 is performed. If not, the process proceeds to step 930, where the process of step 930 is performed.

[0043] In step 930, the document transmission management computer 160 requests approval from the administrator 190 for the transmission of the document 130 to the external computer system 180 via the external transmission approval processing computer 150. In this step 930, to support the approval by the administrator 190, the external transmission approval processing computer 150 presents the document 130 and the clauses of the rules associated with the document 130. A specific example of the presentation will be described later.

[0044] In step 940, the administrator 190 approves or rejects the transmission of the document 130 to the external computer system 180 based on the information presented in step 930. The external transmission approval processing computer 150 then notifies the document transmission management computer 160 of the result. If the administrator 190 approves in step 940, the document transmission management computer 160 actually transmits the document 130 to the external computer system 180 in step 950. If the administrator 190 rejects in step 940, the document transmission management computer 160 cancels the transmission of the document 130 to the external computer system 180 in step 960. Regardless of the results of steps 920 and 940, the document transmission management computer 160 records the result in the external transmission history table 323 together with the document name, date, and administrator name in step 970.

[0045] In flow 900, the document transmission management computer 160 performs the procedures from step 920 onward by appropriately referencing the rule information 140 and the rule / destination information table 322. However, if another component has an access management function, that function may be used or substituted. For example, if the storage 120 has a function for managing access rights to the document 130, the document transmission management computer 160 may prohibit reading and writing of the document 130 from a specific country by setting or canceling the access rights of the document 130 from a specific country to the storage 120 based on the result of step 910, and in step 950, instead of performing external transmission, the prohibition of access rights may be canceled to again permit reading and writing from that country.

[0046] FIG. 10 shows an example of a screen 1000 for approving external transmission of a document used in the processing of steps 930 and 940 .

[0047] The external transmission approval screen 1000 includes a transmitted document field 1010, a related rules field 1020, a past approval history field 1070, an approve button 1090, and a reject button 1091. The transmitted document field 1010 displays the contents of the document 130 to be transmitted. The related rules field 1020 displays part of the rules indicated by the rules 711 and clauses 712 in the rule information 140 assigned to the document 130.

[0048] The transmitted document column 1010 and the related rule column 1020 display supplementary information corresponding to each entry in the rule information 140, and in the example of FIG. 10 , the information of entry 720 is displayed. The highlighted item 1030 in the transmitted document column 1010 highlights the position in the document indicated by the document position 711. The highlighted item 1040 in the related rule column 1020 highlights the position in the rule indicated by the rule 712 or clause 713. Furthermore, the connection relationship 1050 indicates that the highlighted item 1030 and the highlighted item 1040 are stored in the same entry in the rule information 140. In addition, it indicates the correspondence relationship (1060, 1061) between the highlighted item 1030 and the highlighted item 1040, which is stored in the term correspondence information 714. Here, the information on the correspondence relationship can be shown in an appropriate manner, and as shown in Fig. 10, the display may be different from other manners (for example, a display taking into consideration differences in font, the presence or absence of coloring, etc.), or the correspondence relationship may be shown by a dotted line, etc. These highlighted items (1030, 1040), connection relationship 1050, and correspondence relationship between terms (1060, 1061) present to the administrator 190 viewing the external transmission approval screen 1000 that the highlighted items 1030 and 1040 are linked.

[0049] The past approval history 1070 references the external transmission history table 323 and the rules 712 and clauses 713 stored in the rule information 140 assigned to the document 130, and displays information about documents in the past transmission history that match the rules 712 and clauses 713 or the document name 1071. The past approval history 1070 includes multiple histories (1080, 1081) that list a document name 1071, a date 1072, an administrator 1073, and a result 1074. One history shows information about when a document was approved or rejected once. The document name 1071 indicates the document 130 that was the subject of approval or rejection in the history. The date 1072 indicates the date on which the approval or rejection decision was made in the history. The administrator 1073 indicates the administrator 190 who made the approval or rejection decision at the time in the history. The result 1074 indicates the result of the approval or rejection decision in the history.

[0050] If the rule information 140 has multiple entries, the related rule column 1020 and the past approval history 1070 may display information corresponding to the number of entries. In this example, only information for a single clause is shown.

[0051] The related rules column 1020 and past approval history 1070 allow the administrator 190 to grasp the details of the clauses of the rules linked to the document 130 to be sent, and by reading both together, the administrator 190 can ultimately determine whether or not the document 130 can be sent to the external computer system 180. Then, the administrator 190 presses the approve button 1090 if he / she determines that sending is possible, or the reject button 1091 if he / she determines that sending is not possible, thereby completing step 940.

[0052] According to this embodiment, when document data in a computer system is sent to an external computer system, it is possible to prevent the sending of documents that may violate laws and regulations, even if the document data and the rules are written in sentences using different terminology. To achieve this, when linking rules and documents, word strings in the sentences in the rules are converted into word strings of general terms, and similarly, word strings in the sentences in the document are converted into word strings of general terms. After aligning each word string to a general term, the comparison is performed, and the result is treated as the result of comparing the rules and the document. If the comparison identifies a law linked to the document and sending the document may violate the law, the document and the law are presented to an administrator before sending the document to the external computer system, prompting the administrator for approval. The document is sent only after approval is obtained. This supports the administrator's decision on whether to send the document externally, and can prevent the document from being sent externally depending on the decision.

[0053] Next, a second embodiment will be described. The second embodiment relates to a method for generating the rule-oriented term conversion dictionary 223 and the document-oriented term conversion dictionary 224 in the first embodiment. In describing the second embodiment, explanations that are the same as those already explained may be omitted.

[0054] Currently, many methods have been disclosed for generating thesauruses by applying machine learning to text datasets (e.g., Kasahara Kaname; Inako Nozomi; Kato Tsuneaki. Automatic Generation of Synonyms Using Textual Data. Transactions of the Japanese Society for Artificial Intelligence, 2003, 18.4: 221-232.). Additionally, methods have been disclosed for applying machine learning to text datasets to construct distributed representations of words, such as the aforementioned Word2Vec. By using these publicly known techniques on a text dataset containing a mixture of documents using multiple types of terms, it is possible to construct a term conversion dictionary for these multiple terms.

[0055] 11 shows a dictionary construction procedure 1100 that indicates the data and processing flow when constructing a rule-oriented term conversion dictionary 223 and a document-oriented term conversion dictionary 224 from various text data sets using these term conversion dictionary construction techniques. The dictionary construction procedure 1100 may be executed as appropriate, and for example, the rule-sentence linking computer 170 may construct the dictionary by executing the term conversion dictionary generation program 225. Alternatively, the dictionary construction procedure 1100 may be executed as appropriate outside the computer system 110 to construct the dictionary, and the computer system 110 may then acquire the dictionary constructed by an appropriate technique.

[0056] In the dictionary construction procedure 1100, the original text datasets include a rule dataset 1110, a document dataset 1120, and a general text dataset 1130. The rule dataset 1110 is a collection of sentences in the rules 600, and is a text dataset in which many rule-specific terms are used. The document dataset 1120 is a collection of sentences in the documents 650, and is a text dataset in which many document-specific terms are used. The general text dataset 1130 is a text dataset in which many general sentences are used, such as newspapers, texts on the web, books, academic papers, and patent documents.

[0057] In the dictionary construction procedure 1100, a rule and general text dataset 1140 is created by combining the rule dataset 1110 and the general text dataset 1130, and a term conversion dictionary generation process 1160 using the above-mentioned existing method (i.e., a method of applying machine learning) is applied to this rule and general text dataset 1140 to generate a term conversion dictionary for rules 223. Similarly, a document and general text dataset 1150 is created by combining the document dataset 1120 and the general text dataset 1130, and a term conversion dictionary generation process 1160 is applied to this document and general text dataset 1150 to generate a term conversion dictionary for documents 224. In this case, the converted expressions 520 in the rule term conversion dictionary 223 and the converted expressions 570 in the document term conversion dictionary 224 use expressions that appear in the general text dataset 1130, or expressions that are expressed as IDs or numerical vectors.

[0058] According to this embodiment, a term conversion dictionary can be generated for rules or documents for which there is no term conversion dictionary, making it possible to use the computer system of the first embodiment. Furthermore, by including a general text dataset in the text dataset used to generate the term conversion dictionary, the notation in the general text dataset can be used as the converted notation 520 in the rule-oriented term conversion dictionary 223 and the converted notation 570 in the document-oriented term conversion dictionary 224, thereby enabling indirect comparison of rule-specific notation and document-specific notation through conversion to notation in the general text dataset. Additionally, by using the rule dataset 1110 to generate the rule-oriented term conversion dictionary 223 and the document dataset 1120 to generate the document-oriented term conversion dictionary 224, even if the same notation is used with different meanings in a rule and a document, it is possible to prevent these notations from being mistakenly confused because they are converted to different converted notations in the rule-oriented term conversion dictionary 223 and the document-oriented term conversion dictionary 224.

[0059] Next, a third embodiment will be described. The third embodiment relates to the process of linking rules and documents in the first embodiment. In describing the third embodiment, explanations that are the same as those already explained may be omitted.

[0060] In the first embodiment, the rules and documents, and the converted expressions using the terminology conversion dictionary, used the same language, although there were differences in the terms. In the third embodiment, a method is shown in which, even when the rules and documents use different languages, or when the rules and documents use the same language, the words are converted into a language different from either of them (a language different from the language of the rules and the language of the sentence) and compared. In explaining the third embodiment, explanations that are the same as those already explained may be omitted.

[0061] In the following, an example will be given in which rules and documents are written in Japanese and English is used for comparison (i.e., as the comparison language). Note that the languages used in this embodiment are not limited to Japanese and English, and any other language can be used.

[0062] 12 shows a processing flow 1200 for linking rules and documents. The processing flow 1200 for linking rules and documents is a flow in which steps 1210 and 1220 are added to the processing flow 800 for linking rules and documents in the first embodiment. Since the other steps are the same as the processing flow 800 for linking rules and documents, steps 1210 and 1220 will be explained below.

[0063] In step 1210, if the written language of the rule is different from the comparison language (first language), the rule converted into the comparison language is obtained. Any appropriate conversion method may be used; for example, if a Japanese-English translation of the rule already exists, that English translation may be used, or human or machine translation may be performed. Similarly, in step 1220, if the written language of the document is different from the comparison language, the text converted into the comparison language is obtained.

[0064] In the processing flow 1200 for linking rules and documents, the comparison language is used for the steps other than step 1210 and step 1220. The source expressions in the rule-oriented term conversion dictionary 223 and the document-oriented term conversion dictionary 224 are also created in the comparison language.

[0065] This embodiment provides the following two advantages. First, when rules and documents are written in different languages, they can be linked through bilingual translation or translation using a comparison language. Second, because the results of the rule-document linking process flow 800 can vary depending on the language, depending on factors such as language characteristics and the ease of obtaining the general text dataset 1130 described in the second embodiment, the comparison results can be optimized by selecting a language that will provide the best results from the rule-document linking process flow 800 as the comparison language and then performing the rule-document linking process flow 1200.

[0066] Next, a fourth embodiment will be described. The fourth embodiment can be applied as a means for enhancing the effect of the first embodiment, for example, when the rules in the first embodiment have a hierarchical structure or a reference structure. In describing the fourth embodiment, explanations that are the same as those already explained may be omitted.

[0067] As regulations, such as the Customs Tariff Act, list a wide variety of conditions and goods they cover, so they have a hierarchical structure and contain many references and supplementary statements to other provisions. Figure 13 shows an example of such a regulation.

[0068] In FIG. 13, rule 1300 includes clauses (1310, 1320, 1330, 1340, 1350, 1360, 1370, 1380) and note 1390. Clause 1320 is a subordinate clause of clause 1310, and clause 1330 is a subordinate clause of clause 1320. This indicates that the higher the clause, the broader the classification, and the lower the clause, the more specific the classification. In this example, "non-fluorinated compounds" referred to by clause 1320 is more specific than "compounds" referred to by clause 1310, and "biphenylene, triphenylene" referred to by clause 1330 is even more specific. Similarly, clauses (1350, 1360) are listed in parallel as subordinate clauses of clause 1340. This shows that "phenolic resin" referred to in Article 1350 and "items falling under item 2(A)" referred to in Article 1360 are more specific than "resins" referred to in Article 1340. Similarly, "heavy oil" referred to in Article 1380 is more specific than "fuels and mineral oils" referred to in Article 1370.

[0069] Generally, each provision can be referenced by its number, ID, or name. For example, provision 1310 begins with "2," and provision 1320 begins with "(A)." Therefore, a specific provision 1320 can be referenced by using a notation such as "2(A)," which lists these numbers, IDs, and names in descending order, or by including descriptions such as "item" and "paragraph" that generally refer to regulatory provisions, such as "item 2(A)." These notations, as well as notations referring to nearby provisions in the hierarchy, such as "one of the following," allow a provision to reference other provisions within or outside the regulation. For example, reference 1361 "item 2(A)" in provision 1360 references another provision 1320. This means that when comparing whether a document is linked to provision 1360, the comparison should also include provision 1320, which references reference 1361. Furthermore, if a document references a provision in another document outside the regulation (a third document), that document will also be included in the comparison.

[0070] Also, one statement may supplement another. Note 1390 supplements Article 1380 by reference 1391. This means that Article 1380 also covers "petroleum used as fuel," which is included in Note 1390.

[0071] In this embodiment, a processing flow 1400 for linking rules and documents is executed. Fig. 14 shows the processing flow 1400 for linking rules and documents. The processing flow 1400 for linking rules and documents is a flow in which steps (1410, 1411, 1412) are added to the processing flow 800 for linking rules and documents in the first embodiment, and step 850 is replaced with step 1413. The steps (1410, 1411, 1412, 1413) specific to the processing flow 1400 for linking rules and documents will be described below.

[0072] In step 1410, the hierarchical structure of the clauses of the rule is acquired. If the rule is originally written in the form of a structured document using XML (Extensible Markup Language) or the like, that hierarchical structure can be regarded as the hierarchical structure of the clauses of the rule. Alternatively, the hierarchical structure can be acquired by applying a method of constructing the hierarchical structure of the rule by referring to the numbers of the articles, paragraphs, and sections in the rule, or a method of constructing the hierarchical structure by referring to the layout and indentation in the written rule.

[0073] In step 1411, references to other items and supplementary relationships in the clauses of the rules are obtained. One possible means for doing this is to extract expressions that refer to other items, such as the above-mentioned "item 2 (A)" and "any of the following," from the sentences of the clauses of the rules.

[0074] In step 1412, the following word strings are added to the word strings using common terms generated from the sentences of the clauses themselves in step 812 as word strings constituting the clauses of the rules. -Word strings using common terms generated from clauses in higher levels A word string using common terms generated from the referenced clause and clauses above and below the referenced clause A word string using common terms generated from the supplementary clause and clauses above and below the supplementary clause

[0075] When adding these word strings, weighting may be applied to the comparison results in step 841 according to their hierarchical distance from the original clause. For example, a weight multiplied by the nth power of 0.5 may be applied to word strings in clauses n levels higher during comparison in step 841. This makes it possible to control the influence of higher-level and referenced terms on the comparison results.

[0076] In step 1413, similar to step 850, a clause of a rule associated with the document is selected. However, the hierarchical structure of the rule may be used to narrow down the results. For example, if two clauses in a hierarchical relationship both exceed a threshold, the lower clauses may be retained and the higher clauses may be removed. As an example, assume that clauses 1320 and 1330 in rule 1300 both receive scores above the threshold. In this case, since clause 1330 indicates more specific content, if clause 1330 is associated with the document, there is little point in administrator 190 considering clause 1320, which is a larger category, when determining whether to allow external transmission. As another example, assume that clauses 1350 and 1360, which are parallel in the hierarchy, both receive scores above the threshold. In this case, they may be associated with clause 1340, the higher-level clause that encompasses them, and each of clauses 1350 and 1360 may be removed from the association.

[0077] According to this embodiment, it is possible to link rules and documents based on the hierarchical relationships, references, and supplementary relationships of the rules. For example, "biphenylene resin" can be linked to clause 1360. This is because the word string using common terms generated from clause 1360 in step 1412 includes both "resin" included in clause 1340, which is a superordinate item, and "biphenylene" included in clause 1330, which is a subordinate item of clause 1320, which is the reference destination of reference 1361.

[0078] Although the embodiments have been described above, the present invention is not limited to the above-described embodiments and includes various modifications. For example, it is possible to add, delete, or replace part of the configuration of the embodiment with other configurations.

[0079] A CPU is one example of a processor, but other semiconductor devices (for example, a GPU) may also be used as long as they are the main body that executes predetermined processing.

[0080] In this embodiment, an example has been described in which the computer system 110 converts and compares rules (legal rules using legal terminology) and technical text. However, the present invention is not limited to this example, and the computer system 110 can process documents that use two different vocabularies. For example, the computer system 110 may process text that makes extensive use of technical terms or proprietary terminology. For example, the computer system 110 may convert terms contained in other rules, such as company rules, into general terms for processing, and may also convert terms contained in other types of text, such as technical text, into general terms for processing.

[0081] Although the computer system 110 is configured from a plurality of computers, it may be configured from a single computer that realizes each function of the computer system 110.

[0082] The computer system 110 can convert word strings in a sentence as appropriate. For example, word strings in a sentence that add explanations, conditions, exceptions, etc. to the contents of the previous sentence, such as "proviso," may be converted. Also, word strings in a sentence that uses parentheses to provide supplementary explanations, such as "parenthetical writing," may be converted. [Explanation of symbols]

[0083] 100 Computer system configuration diagram 110 Computer Systems 120 Storage 130 documents 140 Regulation Information 150 External transmission approval processing computer 160 Document transmission management computer 170 Rules and Document Linking Calculator 180 External Computer Systems 190 Administrator 195 User

Claims

1. A document data management method carried out using a computer, comprising: the computer converts a word string 1 in a first document written in a first term that is a term used in a sentence defining a rule into a word string 2 having the same meaning as the word string 1, which is written in a third term that is a term different from the first term, using a term conversion dictionary that uses the first term as a conversion source term; converting a word string 3 of a second document written in second terms, which are terms used in a document to be sent externally, into a word string 4 having the same meaning as the word string 3 written in third terms, using a term conversion dictionary that uses the second terms as source terms; the computer performs a process of converting each clause of a rule in the first document into the word string 2; calculating a score for each clause of the rule regarding the association and / or similarity between a word string 2 and the word string 4 included in the clause of the rule; the score is a score based on N-gram, a score based on BLUE, or a score based on the distance between numerical vectors when a numerical vector is used as a destination notation in a term conversion dictionary in which the first term is a source term and a term conversion dictionary in which the second term is a source term, the computer stores the clauses of the rules that show the top few scores or scores equal to or greater than a threshold in rule information to be assigned to the second document; The computer compares the rule information with information about the destination to which the rule applies and the location of the destination system, and if it determines that the rule information does not contain any rule clause that restricts the document transmission of the second document to the system, it determines to externally transmit the document to be transmitted. A document data management method comprising:

2. 2. The document data management method according to claim 1, generating a term conversion dictionary using the first term as a conversion source term through machine learning using a group of documents described using the first term and a group of documents described using the third term; generating a term conversion dictionary using the second term as a conversion source term through machine learning using a group of documents described using the second term and a group of documents described using the third term; A document data management method comprising:

3. 2. The document data management method according to claim 1, After the word string 1 is translated into a first language, the word string 1 is converted into the word string 2 described in the third terminology of the first language; After the word string 3 has been translated into the first language, the word string 3 is converted into the word string 4 described in the third terminology of the first language. A document data management method comprising:

4. 2. The document data management method according to claim 1, When an item in the first document references another item in the first document or an item in a third document different from the first document and the second document, a word string obtained by converting the word string 1 obtained from the item of the first document by the computer using a term conversion dictionary with the first term as the source term into a word string of a third term, and adding the word string obtained by converting the word string described in the first term obtained from the referenced item by the computer using the term conversion dictionary into a word string of a third term, and processing the resulting word string as the word string 2; A document data management method comprising:

5. 2. The document data management method according to claim 1, When the first document has a hierarchical structure, a word string obtained by converting a word string 1 obtained from an item of the first document by the computer using a term conversion dictionary with the first term as the source term into a word string of a third term, and adding the word string obtained by converting a word string described in a first term obtained from an item in a higher hierarchy of the item by the computer using the term conversion dictionary into a word string described in a third term, and processing the resulting word string as the word string 2; A document data management method comprising:

6. 6. The document data management method according to claim 5, When processing the word string obtained by adding the word string described with the third term obtained from the item in the upper hierarchy as the word string 2, the obtained word string is weighted in relation to the score calculated by the computer according to its positional relationship in the hierarchy. A document data management method comprising:

7. 2. The document data management method according to claim 1, When the first document has a hierarchical structure, a word string described with a third term obtained from an item in a higher hierarchy than the first document is added to the word string obtained from the first document item; Furthermore, when an item in the first document refers to another item in the first document or an item in a third document different from the first document and the second document, a word string obtained by adding a word string described in third terms obtained from the referenced item and an item above or below the referenced item to the word string 1 obtained from the item in the first document is processed as the word string 2. A document data management method comprising:

8. 2. The document data management method according to claim 1, presenting to the administrator word strings to be converted into the second and fourth word strings that are related and / or similar based on the magnitude of the score; Implement or stop the transmission of said text as a result of the administrator's approval or denial; A document data management method comprising:

9. A program for causing a computer to execute the document data management method according to claim 1.

10. A program for causing a computer to execute the document data management method according to claim 8.

11. a processor; The processor: (1) converting a word string 1 in a first document written in a first term, which is a term used in a sentence defining a rule, into a word string 2 having the same meaning as the word string 1, which is written in a third term, which is a term different from the first term, using a term conversion dictionary that uses the first term as a conversion source term; (2) converting a word string 3 in a second document written in second terms, which are terms used in a document to be sent externally, into a word string 4 having the same meaning as the word string 3 written in third terms, using a term conversion dictionary that uses the second terms as source terms; (3) the processor performs a process of converting each clause of a rule in the first document into the word string 2; calculating a score for each clause of the rule regarding the association and / or similarity between a word string 2 and the word string 4 included in the clause of the rule; the score is a score based on N-gram, a score based on BLUE, or a score based on the distance between numerical vectors when a numerical vector is used as a destination notation in a term conversion dictionary in which the first term is a source term and a term conversion dictionary in which the second term is a source term, the processor stores the clauses of the rules that show the top few scores or scores equal to or greater than a threshold in rule information to be assigned to the second document; (4) The processor compares the rule information with information about the destination to which the rule applies and the location of the destination system, and if it determines that the rule information does not contain any rule clause that restricts the document transmission of the second document to the system, it determines to externally transmit the document to be externally transmitted. A text data management system characterized by:

12. The document data management system according to claim 11, When sending the above text externally, The processor: presenting to the administrator word strings to be converted into the second and fourth word strings that are related and / or similar based on the magnitude of the score; Implement or stop the external transmission of the text based on the result of the administrator's approval or rejection of the external transmission; A text data management system characterized by:

Images