Document creation support device and document creation support method

The document creation support system addresses the burden of repetitive comments by associating appropriate comments with similar documents, enhancing efficiency through automated comment extraction and recording.

JP2026182539APending Publication Date: 2026-11-19HITACHI SOFTWARE ENG
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
JP2025078852
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2025-05-09
Publication Date
2026-11-19

AI Technical Summary

Technical Problem

Existing document checking technologies do not adequately address the increased workload on checkers when similar comments need to be added to multiple documents, leading to a significant burden.

Method used

A document creation support system that associates appropriate comments with specified ranges of documents by identifying similar structural units and applying comments from commented documents to corresponding ranges in newly created documents using a document creation support system comprising a storage device and computing device for effective comment extraction and recording.

Benefits of technology

Enables the association of appropriate comments with created documents, reducing the workload on checkers by automating the process of identifying and applying relevant comments based on document similarity and structure.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2026182539000001_ABST
    Figure 2026182539000001_ABST
Patent Text Reader

Abstract

Assign appropriate comments to the created documents. [Solution] A document creation support system 1 comprising a storage device 22 that stores a commented document, which is a document to which comments targeting a specified range of the document are associated, and a processing unit 21 that acquires a document to be processed and performs an effective comment extraction process to identify a range of the document to be processed that corresponds to the specified range of the commented document, and a comment recording process to associate the comments related to the commented document with the corresponding range of the document to be processed.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a document creation support device and a document creation support method.

Background Art

[0002] Technologies for checking documents using generative AI (Artificial Intelligence) such as LLM (Large Language Model) have been developed. For example, Patent Document 1 discloses a check system comprising: a text checking unit that detects predetermined deficiencies from input document data using a check tool and outputs a warning text; a first determination unit that determines whether the warning text is correct using a machine learning model; and a second determination unit that re-determines the warning text determined to be correct by the first determination unit based on the user's intention.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of Invention

Problems to be Solved by the Invention

[0004] Incidentally, in document checking, a method is sometimes adopted in which a different checker adds comments to a document created by a document creator. However, when the number of documents to be checked increases, it becomes necessary to add similar comments to similar documents, which tends to increase the work burden on the checker. At present, technologies for reducing such a burden on checkers have not been sufficiently developed.

[0005] This invention has been made in view of these circumstances, and its purpose is to provide a document creation support device and a document creation support method that can associate appropriate comments with the created document. [Means for solving the problem]

[0006] One of the present inventions for solving the aforementioned problems is a document creation support system comprising: a storage device for storing a commented document, which is a document to which comments relating to a specified range of the document are associated; and a computing device that acquires a document to be processed and performs an effective comment extraction process to identify a range of the document to be processed that corresponds to the specified range of the commented document, and a comment recording process to associate the comments relating to the commented document with the corresponding range of the document to be processed. [Effects of the Invention]

[0007] According to the present invention, appropriate comments can be associated with the created documents.

[0008] Other configurations and effects will be clarified by the following description of the embodiments. [Brief explanation of the drawing]

[0009] [Figure 1] This figure shows an example of the configuration of the document creation support system 1 according to this embodiment. [Figure 2] This figure shows an example of the hardware and functions of the review support system 20. [Figure 3] This figure shows an example of the data structure of document DB300. [Figure 4] This figure shows an example of the data structure of the review comments DB400. [Figure 5] This figure shows an example of the data structure of reviewer DB500. [Figure 6] This figure shows an example of the data structure of DB600 based on the review results. [Figure 7] This is a flowchart illustrating the overview of the document creation support process. [Figure 8] This flowchart details the commented document information registration and decomposition process s3. [Figure 9] This is a flowchart illustrating the details of the document registration and decomposition process s5. [Figure 10] This figure shows an example of a review notice DB400 created by the commented document registration decomposition process s3. [Figure 11] This figure shows an example of the document DB300 created by the commented document registration decomposition process s3 and the document to be processed registration decomposition process s5. [Figure 12] This is a flowchart that explains the details of the automated review process s7. [Figure 13] This figure shows an example of the review result DB600 created by the automated review process s7 up to s93. [Figure 14] This figure shows an example of the automatic review result display screen 1400, which is displayed during the s97 process of the automatic review process s7. [Figure 15] This figure shows an example of the updated review comments DB400 based on the processing of s97 in the automated review process s7. [Figure 16] This figure shows an example of the reviewer DB500 updated based on the processing in s97 of the automated review process s7. [Figure 17] This figure shows an example of the updated review results DB600 based on the processing in s97 of the automated review process s7. [Figure 18] This is a flowchart illustrating the details of manual review process s10. [Figure 19] This figure shows an example of review comments DB400 updated by manual review process s10. [Figure 20] This figure shows an example of the review results DB600 updated by manual review process s10. [Detailed Description of the Invention]

[0010] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

[0011] The document creation support system of the present embodiment is an information processing system that supports a predetermined creator (user) in creating a document.

[0012] A document creator creates various types of documents. For a document created by the document creator, a predetermined comment creator (hereinafter referred to as a reviewer) adds a comment targeting a specified range (comment range) of the document at any time to conduct a review. The document to which such comments have been added (hereinafter referred to as a commented document) is stored in a predetermined database.

[0013] The document creation support system uses these commented documents to automatically create a document with added comments, by adding appropriate comments from the comments attached to the commented documents to the appropriate ranges of a newly created document by the creator (hereinafter referred to as a target document for processing). In the present embodiment, adding a comment to a document refers to directly adding data to the document (for example, using the review function of an application), but other modes are also possible. For example, the document creation support system 1 may create comment data separate from the document, and associate the document and the comment data with each other via a database or the like.

[0014] In this embodiment, commented documents and documents to be processed are defined as documents having one or more structural units consisting of one or more sentences. In this embodiment, these structural units are defined as "chapters," but this is not to be limited to them, and each document may have a hierarchical structure such as "sections" or "parts."

[0015] Figure 1 shows an example of the configuration of the document creation support system 1 according to this embodiment. The document creation support system 1 is composed of one or more user terminals 10 and information processing devices of the review support system 20.

[0016] User terminal 10 is an information processing device used by creators who create documents using the review support system 20.

[0017] The review support system 20 is comprised of information processing devices including a web server 30, an application server 40, a database server 50, and a storage device 60.

[0018] The web server 30 is a server device that sends and receives data with the user terminal 10. The application server 40 is a server device that stores and executes various programs. The database server 50 is a server device that stores the database accessed by the application server 40. The storage 60 stores various data such as documents.

[0019] The above information processing devices are connected to each other by a wired or wireless communication network 5, such as the Internet, LAN (Local Area Network), WAN (Wide Area Network), or dedicated line.

[0020] Figure 2 shows an example of the hardware and functions provided by the review support system 20.

[0021] The review support system 20 comprises a computing unit 21 such as a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit), a storage device 22 such as RAM (Random Access Memory), ROM (Read Only Memory), HDD (Hard Disk Drive), or SSD (Solid State Drive), an input device 23 consisting of a keyboard, mouse, touch panel, etc., an output device 24 for screen display consisting of a monitor (display), etc., and a communication device 25 consisting of a NIC (Network Interface Card), wireless communication module, USB (Universal Serial Interface) module, or serial communication module, etc. The user terminal 10 also has a similar hardware configuration.

[0022] Next, the review support system 20 has the following functional units: a document registration decomposition unit 201, a vector value calculation unit 202, a valid comment extraction unit 203, and a comment recording unit 204.

[0023] The document registration decomposition unit 201 divides the commented document and the document to be processed into chapters and sentences within each chapter.

[0024] The vector value calculation unit 202 calculates the similarity between texts (between documents, between chapters, between sentences).

[0025] The valid comment extraction unit 203 identifies the range of the document to be processed that corresponds to the comment range of the commented document. Specifically, the valid comment extraction unit 203 calculates the similarity between the specified range of the commented document and each range in the document to be processed, and identifies the range where the calculated similarity satisfies predetermined conditions.

[0026] More specifically, the valid comment extraction unit 203 first selects the chapter to which comments will be added and the sentences within that chapter from the document to be processed.

[0027] The valid comment extraction unit 203 identifies similar documents from among multiple commented documents that are commented documents whose similarity to the document to be processed satisfies predetermined conditions.

[0028] Furthermore, the valid comment extraction unit 203 identifies similar constituent units (similar chapters) from each chapter of the identified similar documents, which are constituent units whose similarity to the selected chapter in the document to be processed satisfies predetermined conditions.

[0029] Furthermore, the valid comment extraction unit 203 identifies similar sentences from each sentence of the identified similar constituent unit (similar chapter) that satisfy predetermined conditions regarding similarity to the selected sentence in the document to be processed.

[0030] The comment recording unit 204 then associates the comments related to the commented document with the corresponding range of the document to be processed.

[0031] For example, the comment recording unit 204 associates the selected sentence in the document to be processed with a comment from a commented document that targets a similar sentence identified by the valid comment extraction unit 203.

[0032] Next, the review support system 20 stores data from the following databases: the document DB 300, which stores information about each document; the review comments DB 400, which stores information about comments attached to each document; the reviewer DB 500, which stores information about the person who made the comment on each document (reviewer); and the review results DB 600, which stores detailed information about the comments attached to each document.

[0033] (Document Database) Figure 3 shows an example of the data structure of document DB300. Document DB300 has the following data items: document ID 301 for each document (commented documents and documents to be processed), chapter ID 302 for each chapter in each document, sentence ID 303 for each sentence in each document, text 304 representing the content of each sentence, vector value 305 for the text at the document, chapter, or sentence level, and the number of comments attached to each document (review point count 306).

[0034] (Review Point Database) Figure 4 shows an example of the data structure of the Review Comments DB400. The Review Comments DB400 has the following data items: the ID of each comment attached to a commented document (Review Comments ID401), the document ID402 which is the document to which each comment applies (the commented document or the document to be processed), the ID of the chapter to which each comment applies (Chapter ID403), the ID of the sentence to which each comment applies (Sentence ID404), the content of each comment (Review Comments 405), the reviewer 406 who made each comment, and the number of times each comment was accepted by the document's creator (Acceptance Count 407). Note that if the scope of a comment spans multiple sentences (chapters), multiple sentences (chapters) are associated with the comment by multiple records in Sentence ID404.

[0035] (Reviewer DB) Figure 5 shows an example of the data structure of the reviewer DB 500. The reviewer DB 500 has the following data items: each reviewer 501, the number of comments each reviewer has made (number of comments 502), the number of times each reviewer's comments have been adopted by the document creator (number of adoptions 503), and each reviewer's confidence level 504. The confidence level 504 is an evaluation value that represents the reviewer's skill. In this embodiment, the confidence level 504 is the ratio of the number of adoptions 503 to the number of comments 502.

[0036] (Review Results Database) Figure 6 shows an example of the data structure of the review results DB 600. The review results DB has the following data items: an ID assigned to the author's action for each comment (review result ID 601), an ID of the document to which each comment applies (document ID 602), an ID of the chapter to which each comment applies (chapter ID 603), an ID of the sentence to which each comment applies (sentence ID 604), a review comment ID 605 which is the ID of each comment, a review comment 606 which is the content of each comment, and information indicating whether or not each comment was accepted by the document's author (acceptance flag 607).

[0037] Each of the functions of the review support system 20 described above is realized by the hardware of the review support system 20, or by the computing unit 21 of the review support system 20 reading and executing each program stored in the storage device 22. These programs are stored in storage devices such as secondary storage devices, non-volatile semiconductor memory, hard disk drives, SSDs, or recording media readable by each device, such as IC cards, SD cards, or DVDs. Each device may be realized in whole or in part using virtual information processing resources provided using virtualization technology, process space isolation technology, etc., such as virtual servers provided by a cloud system. Furthermore, all or part of the functions provided by these devices may be realized by services provided by the cloud system via APIs (Application Programming Interfaces), etc.

[0038] Next, we will explain the processes performed by the document creation support system 1.

[0039] Figure 7 is a flowchart illustrating the overview of the document creation support process performed by the document creation support system 1. This process is executed, for example, when the review support system 20 receives a predetermined start instruction from the user terminal 10, when the document is stored in the storage 60, or at other predetermined times (for example, a predetermined time, a predetermined time interval).

[0040] The review support system 20 retrieves commented documents and documents to be processed from the storage 60 (s1). The scope of comments in commented documents can be, for example, chapter by chapter, sentence by sentence, spanning multiple sentences, spanning multiple chapters, or the entire document.

[0041] The review support system 20 then performs a commented document information registration and decomposition process s3, which analyzes the commented document to identify chapters and sentences, and to identify the scope of each comment.

[0042] Furthermore, the review support system 20 performs a document information registration and decomposition process s5, which analyzes the document to be processed and identifies chapters and sentences.

[0043] After the execution of the commented document information registration and decomposition process s1 and the document information registration and decomposition process s3, the review support system 20 executes an automated review process s7 that adds comments to the document to be processed using the comments in the commented document.

[0044] Furthermore, the review support system 20 performs a manual review process s10, which accepts additional comments from the reviewer regarding the document being processed.

[0045] The details of each process are explained below.

[0046] <Commented Document Information Registration and Decomposition Process s3> Figure 8 is a flowchart illustrating the details of the commented document information registration decomposition process s3.

[0047] The review support system 20 selects one of the commented documents (s31).

[0048] The review support system 20 then calculates the text features of the selected commented document (s33). For example, the review support system 20 calculates the vector values ​​of the selected commented document. Any algorithm can be used to calculate the vector values, such as Bag of Words (TF-IDF, LSI, LDA) or distributed representations (Word2Vec, Retrofitting, Glove, FastText, BERT). The review support system 20 registers the calculated vector values ​​in the document DB 300. Note that the review support system 20 may also use scalar quantities as features instead of vector values.

[0049] Next, the review support system 20 identifies the text of each chapter in the selected commented document and calculates vector values ​​for the identified text (s35). For example, the review support system 20 identifies the text of each chapter by analyzing the commented document and identifying the string representing the chapter. The method for calculating the vector values ​​may be the same as in s33. The review support system 20 registers the calculated vector values ​​in the document DB 300.

[0050] Next, the review support system 20 identifies the text of each sentence in each chapter identified in s35 and calculates the vector value of the identified text (s37). For example, the review support system 20 identifies the text of each sentence by analyzing the commented document and identifying the string that represents the end of the sentence. The method for calculating the vector value may be the same as in s33. The review support system 20 registers the calculated vector value in the document DB 300.

[0051] Next, the review support system 20 calculates the vector value of each comment attached to the selected commented document in the same manner as in s33 (s39).

[0052] The review support system 20 identifies the reviewer of each comment attached to the selected commented document (s41). The review support system 20 registers the scope of each comment (sentence, chapter, document) and the reviewer who made each comment in the review reference DB 400. If a comment spans multiple sentences or chapters, those multiple sentences or chapters are registered in the review reference DB 400.

[0053] The review support system 20 checks whether there are any unselected commented documents (s43). If there are any unselected commented documents, the review support system 20 repeats the process from s33 onwards for those commented documents. If there are no unselected commented documents, the commented document registration and decomposition process s3 terminates.

[0054] <Processing target document registration and decomposition process s5> Figure 9 is a flowchart illustrating the details of the document registration and decomposition process s5.

[0055] First, the review support system 20 calculates the vector values ​​of the document to be processed, similar to s33 (s53). The review support system 20 then registers the calculated vector values ​​in the document DB 300 (s55).

[0056] Next, the review support system 20 identifies the text of each chapter of the document to be processed, similar to s35, and selects one of the identified chapters (s57). The review support system 20 calculates the vector value of the text of the selected chapter (s59). The method for calculating the vector value may be the same as in s33. The review support system 20 registers the calculated vector value in the document DB 300 (s61).

[0057] Next, the review support system 20 identifies the text of each sentence in the selected chapter, similar to s37, and selects one of the identified sentences (s63). The review support system 20 calculates the vector value of the text of the selected sentence (s65). The method for calculating the vector value may be the same as in s33. The review support system 20 registers the calculated vector value in the document DB 300 (s67).

[0058] The review support system 20 checks whether processes s65 and s67 have been performed for all sentences in the selected chapter (s69). If processes s65 and s67 have been performed for all sentences in the selected chapter, the review support system 20 performs process s71. If there are any selected sentences for which processes s65 and s67 have not been performed, the review support system 20 repeats the processes from s65 onwards for those sentences.

[0059] The review support system 20 checks whether processes s59 and s61 have been performed for all chapters of the document to be processed (s71). If processes s59 and s61 have been performed for all chapters of the document to be processed, the review support system 20 performs process s73. If there are any chapters for which processes s59 and s61 have not been performed, the review support system 20 repeats the processes from s59 onwards for those chapters.

[0060] Figure 10 shows an example of the review comments DB400 created by the commented document registration decomposition process s3. This review comments DB400 contains the scope of each comment (sentence, chapter, and the ID of the commented document) associated with the commented document (document A) by the commented document registration decomposition process s3, along with the reviewer information for each comment. Note that the adoption count 407 is not set (details will be explained later).

[0061] Figure 11 shows an example of a document DB300 created by the commented document registration and decomposition process s3 and the document to be processed registration and decomposition process s5. This document DB300 stores the full text, each chapter, and each sentence of the document to be processed (document B) (reference numeral 320).

[0062] Furthermore, document DB300 stores the full text, chapters, and sentences of the commented document (document A) (symbol 310). In addition, document DB300 separately stores the scope of each comment in the commented document (document A) that is neither a document, sentence, nor chapter (311 if it spans chapters, 312 if it spans sentences, or 313 if it is part of a sentence), the text 304, and the vector value 305 of that scope. In other words, all such scopes that are neither a document, sentence, nor chapter are treated as sentences. Comments on such scopes are set by the manual review process s10 described later.

[0063] Each commented document in document DB300 has a review count of 306.

[0064] <Automated review process s7> Figure 12 is a flowchart illustrating the details of the automated review process s7.

[0065] First, the review support system 20 calculates the similarity between the vector values ​​of the documents to be processed and the vector values ​​of each commented document (s73).

[0066] For example, the review support system 20 calculates the cosine similarity between the vector value 305 of the record related to the document to be processed in the document DB 300 and the vector value 305 of the record related to each commented document in the document DB 300. In addition to cosine similarity, any other similarity metric such as Euclidean distance may be used.

[0067] The review support system 20 then identifies a predetermined number (for example, 10) of commented documents (similar documents) that have the highest similarity (s75). Alternatively, the review support system 20 may identify a predetermined percentage of commented documents from all commented documents that have the highest similarity, rather than a predetermined number.

[0068] Next, the review support system 20 selects one chapter in the document to be processed by selecting one of the chapter IDs 302 of the record of the document to be processed in the document DB 300 (s77).

[0069] The review support system 20 then calculates the similarity between the vector values ​​of the chapters (selected chapters) of the document to be processed selected in s77 and the vector values ​​of each chapter of each similar document identified in s75 (s79).

[0070] For example, the review support system 20 refers to the document database 300 and calculates the cosine similarity between the vector value 305 of the record relating to the selected chapter of the document to be processed and the vector value 305 of the record relating to each chapter of each commented document identified in s75. In addition to cosine similarity, any other similarity metric such as Euclidean distance may be used.

[0071] The review support system 20 then identifies a predetermined number (for example, 10) of chapters that have the highest similarity (similar chapters) (s81). Alternatively, the review support system 20 may identify a predetermined percentage of all chapters that have the highest similarity, rather than a predetermined number.

[0072] Next, the review support system 20 selects one sentence in the selected chapter of the document to be processed by selecting one of the sentence IDs 303 of the records in the selected chapter of the document to be processed from the document DB 300 (s83).

[0073] The review support system 20 then calculates the similarity between the vector values ​​of the sentences (selected sentences) of the document to be processed selected in s83 and the vector values ​​of each sentence (including those within the selection range) of each similar chapter identified in s81 (s85).

[0074] For example, the review support system 20 refers to the document database 300 and calculates the cosine similarity between the vector value 305 of the record related to the selected sentence and the vector value 305 of the record related to each sentence (selected range) in each similar chapter identified in s81. In addition to cosine similarity, any other similarity metric such as Euclidean distance may be used.

[0075] The review support system 20 then identifies a predetermined number (for example, 10) of sentences with the highest similarity (similar sentences) (s87). Alternatively, instead of a predetermined number, it may identify a predetermined percentage of sentences with the highest similarity from the entire text.

[0076] Then, the review support system 20 identifies comments (hereinafter referred to as corresponding comments) that cover the similar sentences identified in s87 (s89). Specifically, the review support system 20 refers to the review observation DB 400 and retrieves review observation 405 for the record related to sentence ID 404 of the sentence contained in the selected sentence.

[0077] The review support system 20 checks whether it has performed the processes s85-s89 for all sentences in the selected chapter (s91). If it has performed the processes s85-s89 for all sentences in the selected chapter, the review support system 20 performs the process s93. If there are any selected sentences for which the processes s85-s89 have not been performed, the review support system 20 repeats the processes from s85 onwards for those sentences.

[0078] The review support system 20 checks whether the processes s79-s81 have been performed for all chapters of the document to be processed (s93). If the processes s79-s81 have been performed for all chapters of the document to be processed, the review support system 20 performs the process s97. If there are any chapters for which the processes s79-s81 have not been performed, the review support system 20 repeats the processes from s79 onwards for those chapters.

[0079] In s97, the review support system 20 displays a screen (automatic review result display screen) showing the corresponding comments identified in s89 for the document to be processed, along with the scope of those corresponding comments. The review support system 20 then accepts input from the user terminal 10 to select whether or not to adopt the corresponding comments.

[0080] Figure 13 shows an example of the review result DB600 created by the processing up to s93 of the automated review process s7. This review result DB600 contains corresponding comments (review comments 606) associated with each sentence in each chapter of the document being processed. Note that no information has yet been registered in the acceptance flag 607.

[0081] (Automatic review results display screen) Figure 14 shows an example of the automated review results display screen 1400 that is displayed during the s97 process of the automated review process s7.

[0082] The automated review results display screen 1400 includes a display field 1410 for the documents to be processed and a comment selection field 1420.

[0083] The display area 1410 for documents to be processed shows the text of the document to be processed, along with the scope 1411 of the corresponding comments identified in the automated review process s7, displayed in a predetermined format (shaded in the example shown in the figure).

[0084] Then, the comment adoption selection field 1420 is displayed, corresponding to the scope 1411 of the corresponding comment. The comment adoption selection field 1420 displays selection fields 1421 and 1422 that accept input from the user terminal 10 regarding whether or not to adopt the corresponding comment (whether or not to associate the corresponding comment with the document to be processed).

[0085] When an input of "accepted" or "not accepted" is made in the comment acceptance selection field 1420, the review support system 20 updates the review comments DB 400, the reviewer DB 500, and the review results DB 600.

[0086] The automated review results display screen 1400 also includes a similarity display field 1430 that shows the similarity between the commented document and the document to be processed (similarity at the document level, similarity at the chapter level, and similarity at the sentence level).

[0087] The automated review results display screen 1400 may also display information such as the number of comments made by each reviewer (502), the number of times they were adopted (503), or their confidence level (504).

[0088] Next, Figure 15 shows an example of the updated review comments DB400 based on the processing in s97 of the automated review process s7. In this review comments DB400, the value of the number of times the corresponding comment (review comments 405) was adopted in the comment adoption selection field 1420 was adopted has increased by 1.

[0089] Figure 16 shows an example of the reviewer DB500 updated based on the processing in s97 of the automated review process s7. In this reviewer DB500, the number of adoption counts 503 for records related to reviewer 501 who created the corresponding comment adopted in the comment adoption selection field 1420 has increased, and the confidence value 504 has been updated.

[0090] Figure 17 shows an example of the review results DB600 updated based on the processing in s97 of the automated review process s7. In this review results DB600, for each corresponding comment proposed for each sentence of the document being processed, the acceptance flag 607 is set to "accepted" or "rejected" according to the "accepted" or "rejected" selected in the comment acceptance selection field 1420.

[0091] <Manual review process s10> Figure 18 is a flowchart illustrating the details of the manual review process s10.

[0092] The review support system 20 displays the automatic review result display screen 1400 or another screen displaying the document to be processed, and accepts input from the user terminal 10 for additional comments on the document to be processed and the scope of those comments (s101).

[0093] The review support system 20 calculates the vector value of the comment entered in s101 in the same manner as in s33 (s103).

[0094] The review support system 20 obtains the reviewer of the comment entered in s101 from the user terminal 10 or the like. The review support system 20 registers the sentences, chapters, and documents covered by the comment, along with the reviewer of the comment, in the review reference DB 400. If the comment spans multiple sentences or chapters, all of those sentences or chapters are registered.

[0095] The review support system 20 checks whether the user has finished entering comments (s107). For example, the review support system 20 checks whether it has received a predetermined termination instruction from the user terminal 10.

[0096] If you finish entering comments (s107:YES), the review support system 20 will terminate the manual review process s10. If you do not finish entering comments (s107:NO), the review support system 20 will repeat the process from s101 onwards.

[0097] Figure 19 shows an example of the review comments DB400 updated by manual review processing s10. This review comments DB400 has the scope (review comments 405) set for each comment added to the document being processed (document B). For example, the review comments DB400 has the scopes set for sentences 411, chapters 412, sentences 413, and parts of sentences 414. In addition, the reviewers 406 for each comment have been added to the review comments DB400. The adoption count 407 is not set.

[0098] Figure 20 shows an example of the review results DB600 updated by the manual review process s10. This review results DB600 contains comments (review comments 606) for the document being processed and the scope of those comments (document ID 602, chapter ID 603, sentence ID 604). Note that the acceptance flag 607 is not set.

[0099] Furthermore, the document DB300 is also updated by the manual review process s10. In this case, for example, as shown in Figure 11, the document to be processed is a commented document (Document A), and the scope of the comments (311 if it spans chapters, 312 if it spans sentences, 313 if it is part of a sentence, etc.) is set. Note that the number of review points 306 is not set.

[0100] As described above, the review support system 20 of this embodiment identifies the range of the document to be processed (document B) that corresponds to the range of comments in the commented document (document A), and associates the identified comments with the corresponding range of the document to be processed.

[0101] In this way, by associating the comments for a specified range in a commented document with the corresponding range in the document to be processed, it is possible to associate the appropriate comments with the appropriate locations in the document to be processed.

[0102] As described above, the review support system 20 of this embodiment makes it possible to associate appropriate comments with the created documents.

[0103] Furthermore, the review support system 20 of this embodiment calculates the similarity between a specified range (document, chapter, sentence, or selected range) of a commented document and each range (document, chapter, sentence) in the document to be processed, and identifies the range whose calculated similarity satisfies predetermined conditions.

[0104] In this way, by identifying similar ranges between the commented document and the document to be processed, it is possible to associate comments from the commented document with the appropriate ranges of comments with the document to be processed.

[0105] Specifically, the review support system 20 of this embodiment identifies similar documents from a plurality of commented documents that satisfy predetermined conditions (such as being above a threshold) in similarity to the document to be processed, and identifies similar chapters from each chapter of the identified similar documents that satisfy predetermined conditions in similarity to the chapter selected in the document to be processed. Furthermore, the review support system 20 identifies similar sentences from each sentence of the identified similar chapters that satisfy predetermined conditions in similarity to the sentence selected in the document to be processed, and associates the comments of the commented documents targeting the similar sentences with the selected sentences in the document to be processed.

[0106] In this way, when associating comments with a sentence in a chapter of a document to be processed, it is possible to identify commented documents that are similar to the document to be processed as a whole, have similar chapters, and have similar sentences, and then associate the comments on the similar sentences in those commented documents with sentences in the document to be processed. This allows for the appropriate comments to be associated with the appropriate locations while taking into account the structure of the document.

[0107] Furthermore, the review support system 20 of this embodiment identifies a predetermined number or percentage of commented documents that satisfy predetermined conditions for similarity to the document to be processed as similar documents, and identifies a predetermined number or percentage of chapters from each chapter of the similar documents that satisfy predetermined conditions for similarity to a selected chapter in the similar documents as similar chapters. Then, the review support system 20 identifies a predetermined number or percentage of sentences from each sentence of the similar chapters that satisfy predetermined conditions for similarity to a selected sentence in the document to be processed as similar sentences.

[0108] In this way, by selecting only a predetermined number or percentage of documents, chapters, and sentences that are highly similar to the document being processed, it is possible to identify appropriate comments that should be associated with the document being processed.

[0109] Furthermore, the review support system 20 of this embodiment receives input from the creator regarding whether or not to associate corresponding comments with a predetermined range of the document to be processed, calculates an evaluation value for the comment creator who created the comments related to the commented document based on the input selection, and outputs the calculated evaluation value to the output device 24.

[0110] In this way, by updating the reviewer's evaluation score based on whether or not the responding comments from the creator of the document being processed are accepted, it is possible to estimate the reviewer's review ability.

[0111] Furthermore, the review support system 20 of this embodiment accepts comments for the document to be processed and the range of those comments from the user, and stores the document to be processed with comments associated with the specified range as a commented document in the document DB 200.

[0112] In this way, by accepting comments from reviewers for a specified range of the document to be processed and marking it as a commented document, the content of those comments can be reflected in future automated review processes (s7).

[0113] The above description of embodiments is for the purpose of facilitating understanding of the present invention and does not limit it. The present invention can be modified and improved without departing from its spirit, and equivalents thereof are included.

[0114] For example, some of the hardware components of each device in this embodiment may be provided in other devices.

[0115] Furthermore, each program of each device may be provided in other devices, a program may consist of multiple programs, or multiple programs may be integrated into a single program.

[0116] In this embodiment, the confidence level 504 is calculated by the ratio of the number of times the issue was adopted 503 to the number of times it was pointed out 502, but other calculation formulas may also be used.

[0117] Furthermore, in this embodiment, the review support system 20 associates comments in a commented document that span across sentences or chapters with the sentences in the document being processed. However, the system may also identify the sentence or chapter spans corresponding to a comment in a commented document by calculating the vector value of each range (a range that spans across sentences or chapters) in the document being processed as needed. [Explanation of Symbols]

[0118] 1 Document creation support system, 20 Review support system, 201 Document registration decomposition unit, 202 Vector value calculation unit, 203 Effective comment extraction unit, 204 Comment recording unit

Claims

1. A storage device that stores a commented document, which is a document to which comments relating to a specified range of the said document are associated, and A valid comment extraction process that retrieves the document to be processed and identifies the range of the document to be processed that corresponds to the specified range of the commented document, A computing device that performs a comment recording process, which associates the comments relating to the commented document with the corresponding range of the document to be processed. A document creation support system equipped with the following features.

2. The aforementioned computing device is In the above-mentioned valid comment extraction process, the similarity between the specified range of the commented document and each range in the document to be processed is calculated, and the range that satisfies the calculated similarity conditions is identified. The document creation support system according to claim 1.

3. The storage device stores multiple commented documents, each having one or more constituent units consisting of one or more sentences, and a document to be processed, each having one or more constituent units consisting of one or more sentences. The aforementioned computing device is In the above valid comment extraction process, From the aforementioned multiple commented documents, similar documents that are commented documents whose similarity to the document to be processed satisfies predetermined conditions are identified. Select a constituent unit and a statement within that constituent unit from the document to be processed, From each of the constituent units of the identified similar documents, similar constituent units are identified that satisfy predetermined conditions regarding the similarity to the selected constituent unit in the document to be processed. From each sentence of the identified similar constituent unit, similar sentences are identified that satisfy predetermined conditions regarding similarity to the selected sentence in the document to be processed. In the comment recording process, the selected sentence in the document to be processed is associated with the commented document that corresponds to the identified similar sentence. The document creation support system according to claim 2.

4. The aforementioned computing device is In the above valid comment extraction process, From the aforementioned multiple commented documents, commented documents whose similarity to the document to be processed satisfies predetermined conditions are identified, and a predetermined number or percentage of the identified documents are identified as similar documents. From each component unit of the identified similar documents, a component unit whose similarity to the selected component unit in the document to be processed satisfies a predetermined condition is identified, and a predetermined number or proportion of the identified component units are identified as the similar component units. From each sentence of the identified similar constituent unit, sentences whose similarity to the selected sentence in the document to be processed satisfies predetermined conditions are identified, and a predetermined number or proportion of the identified sentences are identified as similar sentences. The document creation support system according to claim 3.

5. The aforementioned computing device, in the comment recording process, The system receives input from the user indicating whether or not to associate the identified comment with the corresponding range of the document to be processed. Based on the input selection, it calculates an evaluation value for the comment creator who created the comment related to the commented document, and outputs the calculated evaluation value to the output device. The document creation support system according to claim 1.

6. The aforementioned computing device is The system further performs manual comment processing, which involves receiving comments and the range of those comments from the user for the document to be processed, and then storing the document to be processed, with the comments corresponding to the specified range, as a commented document in the storage device. The document creation support system according to claim 1.

7. A document creation support method using an information processing device comprising a storage device for storing a document, which is a commented document to which comments relating to a specified range of the document are associated, and a computing device, wherein the computing device is A valid comment extraction process that retrieves the document to be processed and identifies the range of the document to be processed that corresponds to the specified range of the commented document, The process involves performing a comment recording process that associates the comments relating to the commented document with the corresponding range of the document to be processed. Document creation support methods.

Citation Information

Patent Citations

  • Document check system and document check program

    JP2024160574A