Support program, support system and support method
The support system addresses the limitations of existing legal document technologies by using a computer-based approach to identify and suggest missing content in legal documents, enhancing the flexibility and quality of legal document creation and modification.
Patent Information
- Application Number
- JP2023213348
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2023-12-18
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2043-12-18
AI Technical Summary
Existing technologies for legal document creation and modification, such as those described in Patent Document 1, are limited in their ability to flexibly address individual document needs, as they primarily rely on pre-registered recommended text and cannot adapt to the unique requirements of specific documents.
A support system and method that utilizes a computer-based acquisition, generation, and proposal unit to identify missing items in legal documents, employing a sentence generation model to suggest tailored content based on acquired information, including description examples, auxiliary information, and evaluation results.
Enables the creation and modification of legal documents by suggesting appropriate content for missing items and clauses, adapting to individual document needs and improving the quality of legal document drafting and revision processes.
Smart Images

Figure 0007588284000001_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to an assistance program, an assistance system, and an assistance method for assisting in the creation or modification of legal documents. [Background technology]
[0002] A high level of expertise is required for the evaluation and creation of documents, particularly legal documents such as contracts and regulations. In light of this background, there are known technologies for checking legal documents and technologies for supporting the revision of legal documents based on the checking.
[0003] For example, Patent Document 1 describes a technology that accepts input of information on review results for legal documents and makes corrections. The technology described in Patent Document 1 can correct legal documents by replacing corresponding parts with recommended sentences registered in advance. [Prior art documents] [Patent documents]
[0004] [Patent Document 1] JP 2020-080132 A Summary of the Invention [Problem to be solved by the invention]
[0005] However, the technology described in Patent Document 1 has a problem that it can only handle content that is expected in advance, and it is difficult to flexibly make corrections to suit individual documents, because the corrections are made by partially replacing the recommended sentences with those that have been registered in advance. Therefore, the present invention aims to solve the above problems. [Means for solving the problem]
[0006] In order to solve the above problems, the present invention provides an assistance program for assisting in the creation or modification of a target document, wherein the target document is a legal document to be created or modified, and the program causes a computer to function as an acquisition unit, a generation unit, and a proposal unit, wherein the acquisition unit acquires item information regarding the contents corresponding to the missing items that are missing in the target document, and auxiliary information regarding the target document for making suggestions corresponding to the target document, the generation unit inputs the item information and auxiliary information into a sentence generation model, and the proposal unit outputs proposal information for assisting in the creation of the missing items in the target document based on the output of the sentence generation model.
[0007] With this configuration, it is possible to support the creation of uncreated parts of the target document. Also, by inputting example descriptions and auxiliary information into the text generation model and outputting suggested information based on the output, it is possible to support document creation according to the target document, rather than simply presenting a template.
[0008] In a preferred embodiment of the present invention, the example description includes an example description of a legal document that corresponds to the missing item, and the example description is a description or a template in a document that specifies a past agreement.
[0009] With this configuration, descriptions that address missing items can be prepared in advance, and suggestions regarding the creation or revision of the target document can be made based on these descriptions.
[0010] In a preferred embodiment of the present invention, the computer is further made to function as an evaluation unit, the evaluation unit acquiring evaluation information indicating the evaluation result of the target document, and the acquisition unit acquiring the example description based on the evaluation information.
[0011] With this configuration, it is possible to make suggestions using appropriate example descriptions depending on the evaluation results of the target document.
[0012] In a preferred form of the present invention, the evaluation information includes an item type indicating the type of the missing item, and the acquisition unit acquires the example of an entry corresponding to the item type of the missing item from a database that stores example entries indicating the contents of the entry in correspondence with the item type.
[0013] With this configuration, it is possible to make suggestions regarding the creation or revision of a target document based on example descriptions according to item type.
[0014] In a preferred embodiment of the present invention, the evaluation information includes evaluation comments regarding the missing items, and the acquisition section acquires the evaluation comments as the item information.
[0015] With this configuration, for example, suggestions regarding the creation or revision of the target document can be made based on evaluation comments that describe revision guidelines and the like.
[0016] In a preferred form of the present invention, the evaluation information includes one or more issues regarding the target document, and the issues are assigned an issue type selected from a plurality of groups including missing items and amendments to clauses, the acquisition unit determines the issue type of the issues included in the evaluation information, and for an issue of missing items, acquires the item information of the missing items related to the issue, and for an issue of amending clauses, acquires target document information indicating the contents of the target document related to the issue, and the generation unit inputs the item information and auxiliary information into the sentence generation model for an issue of missing items, while for an issue of amending clauses, inputs the target document information and information regarding the arrangement defined by the target document into the sentence generation model.
[0017] With this configuration, appropriate suggestions can be made according to the type of problem. Specifically, for missing items, assistance can be provided for creating the missing items based on example descriptions and supplementary information. On the other hand, for amending clauses, assistance can be provided for amending the parts related to the problems pointed out based on target document information indicating the contents of the target document.
[0018] In a preferred form of the present invention, the target document includes a plurality of clauses, and the types of indications indicating missing items are divided into a plurality of types including a partial deficiency, in which some of the clauses in a group are missing, and a total deficiency, in which the entire group is missing, for a group of a plurality of clauses that defines a set of regulations, and the generation unit determines an input to the sentence generation model depending on whether the indication relating to the group of a plurality of clauses is a partial deficiency or a total deficiency.
[0019] By adopting such a structure, appropriate proposals can be made even when a set of regulations is to be established in multiple clauses.
[0020] In a preferred embodiment of the present invention, when the type of indication related to the missing item is a partial missing item, the acquisition unit acquires, as the auxiliary information, the contents of other clauses in the target document that belong to the same group as the missing item.
[0021] With this configuration, it becomes possible to make a proposal that takes into account the relationships between multiple clauses.
[0022] In a preferred embodiment of the present invention, the auxiliary information includes target document information indicating at least a portion of the content of the target document.
[0023] With this configuration, it is possible to make suggestions for creating or correcting missing items based on the contents of the target document.
[0024] In a preferred embodiment of the present invention, the target document information includes information on a location in the target document where the missing item is to be inserted, and the contents of the document before and after the insertion location.
[0025] With this configuration, appropriate suggestions can be made based on the content written before and after the insertion point in the target document.
[0026] In a preferred embodiment of the present invention, the target document includes a plurality of clauses, and the target document information includes the contents of clauses in the target document that relate to the missing item.
[0027] With this configuration, appropriate suggestions can be made based on the contents of the clauses in the target document related to the missing items.
[0028] In a preferred embodiment of the present invention, the target document is a legal document including a preamble, and the target document information includes the contents of the preamble of the target document.
[0029] In legal documents, the preamble may contain definitions of terms used in the document and other premises. By structuring the document in this way, the contents of the target document can be appropriately reflected in the proposal.
[0030] In a preferred embodiment of the present invention, the auxiliary information includes definition information indicating definitions of words used in the target document.
[0031] With this configuration, the contents of the target document can be appropriately reflected in the proposal.
[0032] In a preferred embodiment of the present invention, the auxiliary information includes stylistic information indicating a writing style in the target document.
[0033] By configuring in this way, suggestions can be made in a unified style within the document.
[0034] In a preferred embodiment of the present invention, the acquisition section further acquires a correction strength selected by a user, and the generation section further inputs the correction strength into the sentence generation model.
[0035] With this configuration, it is possible to make suggestions that reflect the user's wishes.
[0036] In a preferred form of the present invention, the computer is further made to function as an identification unit, and the acquisition unit further acquires an item type indicating the type of the missing item and a correction strength selected by a user, and the identification unit identifies, from a database that stores a plurality of example descriptions corresponding to the item type, an example description to be used in creating or correcting the target document based on the correction strength, and the generation unit inputs the example description and auxiliary information identified by the identification unit into the sentence generation model.
[0037] With this configuration, suggestions can be made using example descriptions according to the user's wishes.
[0038] In a preferred embodiment of the present invention, the generation unit generates a generation command instructing to generate a draft sentence for the missing item based on the item information and auxiliary information as a sentence in a natural language, and inputs the generated command to the sentence generation model.
[0039] With this configuration, for example, a conversational text generation model that outputs a text in response to a text input can be used to make suggestions.
[0040] In a preferred embodiment of the present invention, the computer is further made to function as an evaluation unit, wherein the evaluation unit acquires evaluation information indicating an evaluation result of the target document, the evaluation information including one or more issues regarding the target document, the issues being assigned an issue type selected from a plurality of groups including missing items and amendments to clauses, the acquisition unit determines the issue type of the issues included in the evaluation information, and for an issue of missing items, acquires item information of the missing items related to the issue, and for an issue of amendments to clauses, acquires target document information indicating the description of the target document related to the issue, and the generation unit inputs the generation command to the writing generation model for an issue of missing items, instructing the writing generation model to generate a draft sentence for the missing items based on the item information and auxiliary information, while for an issue of amendments to clauses, inputs the generation command to the writing generation model to generate a proposed amendment for the issue based on the target document information and auxiliary information.
[0041] With this configuration, a conversational sentence generation model can be used to provide appropriate suggestions according to the type of problem.
[0042] The present invention is an assistance system for assisting in the creation or revision of legal documents, comprising an acquisition unit, a generation unit, and a proposal unit, wherein the acquisition unit acquires item information regarding the description content corresponding to missing items that are lacking in a target document, which is a legal document to be created or revised, and auxiliary information related to the target document, for making suggestions corresponding to the target document, the generation unit inputs the description examples and auxiliary information into a sentence generation model, and the proposal unit transmits proposal information to assist in the creation of the missing items in the target document based on the output of the sentence generation model.
[0043] The present invention is an assistance method for assisting in the creation or revision of legal documents, and is executed by a computer having an acquisition unit, a generation unit, and a proposal unit, wherein the acquisition unit acquires item information regarding the description content corresponding to the missing items that are missing in a target document, which is a legal document to be created or revised, and auxiliary information related to the target document, which is information for making suggestions corresponding to the target document, the generation unit inputs the description examples and auxiliary information into a sentence generation model, and the proposal unit transmits proposal information to assist in the creation of the missing items in the target document based on the output of the sentence generation model. Effect of the Invention
[0044] According to the present invention, it is possible to appropriately assist in the creation or modification of a target document, including the creation of missing parts. [Brief description of the drawings]
[0045] [Figure 1] FIG. 1 is a block diagram showing a configuration of a support system according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a hardware configuration diagram of the support system according to the embodiment of the present invention. [Diagram 3] FIG. 1 is a block diagram showing a functional configuration of a support system according to an embodiment of the present invention. [Figure 4] 13 is a registration example of a description example of a support system according to an embodiment of the present invention. [Diagram 5] FIG. 4 is a sequence diagram showing a processing procedure of the support system according to the embodiment of the present invention. [Figure 6] 13 is a display example of an evaluation screen of the support system according to the embodiment of the present invention. [Figure 7] 10 is a flowchart showing a processing procedure related to a re-evaluation process of the support system according to the embodiment of the present invention. [Figure 8] 13 is a display example of a proposal screen of the support system according to the embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0046] The support system of the present invention will be described below with reference to the drawings. The drawings show preferred embodiments. However, the present invention may be embodied in many different forms and is not limited to the embodiments set forth herein.
[0047] For example, the configuration, operation, etc. of the assistance system will be described in this embodiment, but similar effects can be achieved by an assistance device, an assistance system, or a method executed by an assistance device, an assistance program, etc. In this embodiment, the assistance system takes the form of so-called cloud computing, in which the assistance program is executed in a server (assistance device) connected to the client terminal via a network in order to realize its functions in a client terminal (user terminal).
[0048] The assistance program may be provided to a computer device as a computer-readable non-transient storage medium or may be provided downloadable from an external server so that an arbitrary computer device can function as an assistance device. The document correction program may be an add-on (plug-in) program in document editing software such as Microsoft Word (registered trademark).
[0049] In this embodiment, the term "unit" may include, for example, a combination of hardware resources implemented by a circuit in the broad sense and software information processing that can be specifically realized by these hardware resources. In this embodiment, "information" is represented, for example, by a physical value of a signal value representing a voltage or current, a high or low signal value as a binary bit set consisting of 0 or 1, or a quantum superposition (so-called quantum bit), and communication and calculation can be performed on the circuit in the broad sense. The circuit in the broad sense is a circuitry realized by appropriately combining a circuit, a processor, a memory, and the like. For example, it is a circuitry including any of a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an LSI (Large Scale Integration), an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), and the like.
[0050] <1.Definition> First, definitions of terms will be explained. The present invention is a technology for supporting document creation or revision by utilizing a sentence generation model. The sentence generation model refers to any model capable of outputting a sentence. As a sentence generation model, a model generated by machine learning can be typically used. In this embodiment, a case will be described in which a conversational sentence generation model is used, which receives a sentence and generates a sentence as a response to the input sentence.
[0051] More specifically, the sentence generation model is a machine learning model for natural language processing, and has a function of generating sentences based on input information. For example, the sentence generation model is a machine learning model based on a Transformer model having a self-attention mechanism. The Transformer model is, for example, a GPT (Generative Pre-trained Transformer) model. The sentence generation model performs learning using a large-scale data set, enabling not only general natural language generation tasks but also highly specialized sentence generation tasks.
[0052] For example, the sentence generation model may be a model trained with a large-scale data set using the GPT-3 or GPT-4 algorithm developed by OpenAI (for example, a ChatGPT model generated by combining training with a large-scale data set and reinforcement learning using a reward prediction model). In addition, although an example of input to the model is text information in this embodiment, image information or audio information may also be input.
[0053] In the present invention, the target document refers to a document that a user is to create or modify. In this embodiment, the target document is assumed to be a legal document such as a contract, a rule, or a regulation. Some legal documents include units divided into articles, clauses, and items, and in this embodiment, each of these units is called a "clause." In this embodiment, the legal document is an example of a contract. Examples of contracts include documents that exert contractual effects, such as non-disclosure agreements, sales contracts, and real estate lease contracts. Furthermore, this embodiment is not limited to contracts, and the legal document may be, for example, a rule, a regulation, or a notice. Examples of rules include documents of legal norms such as work regulations. Examples of notices include documents of policy declarations such as privacy. In addition, legal documents include documents that are somehow related to law, such as notices and powers of attorney.
[0054] When evaluating legal documents, the evaluation is performed from various viewpoints. Examples of the viewpoints of evaluation include whether necessary items are present in the target document, whether there are unfavorable conditions, whether unnecessary items are included, and the like. In the present invention, an "item" is one of the viewpoints that are the subject of such evaluation, and is a unit when checking for deficiencies in the contents of the target document. In the present invention, information about missing items, which are items that are missing in the target document, is input into a sentence generation model, and a proposal is made based on the output. For example, clauses may be used as the items of the present invention, or points of contention stipulated in the legal document may be used as the items of the present invention.
[0055] In legal documents including contracts, there are matters that are open to discussion or negotiation, either now or in the future, regarding the contents, necessity, and other matters from a legal point of view. These are generally called points of contention. Specifically, the points of contention in the present invention can include the contract period, the subject of the contract, and the specific types of exceptions to confidential information in a non-disclosure agreement. In addition, when a set of provisions is set by a plurality of clauses, the set of provisions may be treated as one point of contention. In this embodiment, an example is explained in which the points of contention are "items," that is, a form in which deficiencies are confirmed for each point of contention.
[0056] Here, the "missing items" in the present invention broadly include items that have not been created or are insufficient, regardless of the creation state of the target document. In other words, the missing items are not limited to those that are considered to be missing in an already existing target document, but also include items that should be included in a document to be created in the future. Therefore, the present invention can be used not only in the document revision stage, but also in the stage before a specific document draft is created, for example, to create a draft text corresponding to an item to be included in the target document.
[0057] In the present invention, auxiliary information refers to information related to a target document, and is information for making suggestions according to the target document. For example, target document information indicating at least a part of the contents of the target document, background information related to the background of the target document, stylistic information indicating the writing style of the target document, definition information indicating the definitions of words used in the target document, evaluation information indicating an evaluation of the target document, and the like are used as auxiliary information. The background information, stylistic information, and definition information may be generated and acquired based on the target document information. Furthermore, when using the target document information of only a part of the contents of the target document rather than the entire text, it is preferable to use information indicating a part of the contents that can be used to modify the item information described later to suit the target document.
[0058] As the background information, it is particularly preferable to use information on the agreement stipulated by the target document. More specifically, for example, the background information includes the parties to the agreement, the relationship with the parties, the purpose of the agreement, the requests of the parties regarding the agreement, information indicating the contents to be agreed upon, the type of the agreement, the requests of the parties regarding the agreement, etc.
[0059] As the writing style information, for example, information indicating whether the writing style of the target document is polite (desumasu style) or plain (de aru style) can be used. The writing style information may be acquired by determining whether the writing style is polite or plain based on the target document information.
[0060] The definition information is, for example, information indicating whether the company is classified as A or B, and the definitions of terms used in the target document. The definition information may be specified by input by the user, or may be acquired based on the target document information. Specifically, the definition information can be acquired based on rules such as, for example, the contents in parentheses indicate a definition, or a description in a specific format in the preamble is regarded as a definition.
[0061] Auxiliary information other than the target document information (such as background information, writing style information, and definition information) may be, for example, term sheet information input by a term sheet used in contract negotiations. A term sheet is a table that summarizes the contents of an agreement with the other party regarding the main items of a contract or the contents of a proposal, and term sheet information is information including the contents described in the term sheet. In addition, the auxiliary information may be question information based on the questions and question results generated by question software that asks questions about the contents to be included in a legal document (e.g., a contract).
[0062] The present invention makes suggestions regarding the creation or correction of a target document by inputting the above-mentioned auxiliary information into a text generation model together with item information regarding the description contents corresponding to the missing items.
[0063] <2. System configuration> Fig. 1 is a block diagram showing the configuration of the support system of this embodiment. As shown in Fig. 1, the support system 0 includes a support device 1 and a user terminal 2. The support device 1 and the user terminal 2 are configured to be able to communicate with each other via a network NW. The support device 1 operates as a server, and the user terminal 2 operates as a client terminal. In this embodiment, the network NW is an IP (Internet Protocol) network, but there are no limitations on the type of communication protocol, the type of network, etc.
[0064] The support device 1 is further connected to a generation device 3 via a network NW. The generation device 3 is an external server device that provides a text generation model. The support device 1 may store the text generation model and may be configured to be capable of making suggestions regarding the creation or modification of a target document without using an external device, or may be configured to execute a part or all of the processing executed by the support device 1 in this embodiment in the user terminal 2.
[0065] One or more information processing devices 10 (computer devices), such as a general-purpose server or a personal computer, can be used as the assistance device 1, the generation device 3, and the database server. In this embodiment, the assistance device 1 is an information processing device 10 in which a computer program (assistance program) for executing the assistance method is installed, but a terminal device 90 may function as the assistance device 1 by installing the assistance program in the terminal device 90.
[0066] Furthermore, a terminal device 90 (computer device) such as a personal computer, a smartphone, or a tablet terminal can be used as the user terminal 2. The user terminal 2 has an application (web browser application or dedicated application) for making a request to the support device 1 and receiving a response, and uses the processing results of the support program executed in the support device 1 via this application.
[0067] The support device 1 transmits data in a predetermined format as various display information to the application of the user terminal 2 based on a request transmitted from the application of the user terminal 2. The data in the predetermined format is, for example, data in a JSON (JavaScript (registered trademark) Object Notation) format or an XML format. The application of the user terminal 2 renders the display data based on the received data, and displays various information based on the proposed information on the display of the user terminal 2.
[0068] For example, when a display process of various information is executed on a web browser application of the user terminal 2, the web browser application first transmits an HTTP (Hypertext Transfer Protocol) request to a web server and receives an HTTP response including the file requested by the HTTP request. The received file is display information including a revision proposal, indicates the contents or instructions of the web page to be displayed, and is described using a markup language such as HTML (Hypertext Markup Language). A script, which is a program executed on the web browser application, may be attached to the received file. The web browser application draws (renders) a web page in the display area based on the received file. If a script is attached to the file, the browser executes the script and displays the web page according to the execution result.
[0069] <3. Hardware configuration> Fig. 2(a) is a hardware configuration diagram of the information processing device 10. As shown in Fig. 2, the information processing device 10 has a control unit 101, a storage unit 102, and a communication unit 103, which are used to cause each unit and each process to function.
[0070] The control unit 101 has a processor such as a CPU capable of executing an instruction set, and executes an OS and programs. The storage unit 102 includes a volatile memory such as a RAM capable of storing an instruction set, and a non-volatile recording medium such as an HDD or SSD capable of recording an OS, an XX program, a DBMS (database server), and the like. The communication unit 103 has an interface for physically connecting to a network, and controls communication with the network NW to input and output information.
[0071] Fig. 2(b) is a hardware configuration diagram of the terminal device 90. As shown in Fig. 2, the terminal device 90 has a control unit 901, a storage unit 902, a communication unit 903, an input unit 904, and an output unit 905, which are used to perform the functions of each unit and each process.
[0072] The control unit 901 has a processor such as a CPU capable of executing an instruction set, and executes an OS, programs, and the like. The storage unit 902 includes a volatile memory such as a RAM capable of storing an instruction set, and a non-volatile recording medium such as an HDD or SSD capable of recording an OS, a determination program, and the like. The communication unit 903 has an interface for physically connecting to a network, and controls communications with the network NW to input and output information. The input unit 904 includes an operation input device capable of input processing, such as a touch panel or a keyboard, and an audio input device capable of audio input, such as a microphone. The output unit 905 includes a display device capable of display processing, such as a display, and an audio output device, such as a speaker.
[0073] <4. Functional configuration> 3 is a diagram showing the functional configuration of the support device 1, the user terminal 2, and the generation device 3 in this embodiment. The support device 1 has an acquisition unit 11, an evaluation unit 12, an identification unit 13, a generation unit 14, a proposal unit 15, and a database DB. This is a concrete implementation of information processing by software using hardware (such as the control unit 101). For example, the control unit 101 reads out a document correction program stored in the storage unit 102, expands it in a temporary storage area (such as a RAM) of the storage unit 102, etc., executes it, and controls the communication unit 103, etc., thereby implementing the functions of the support device 1.
[0074] The user terminal 2 also includes a display unit. The user terminal 2 receives the information transmitted by the proposal unit 15, and the display unit performs display based on the information. Display examples will be described later.
[0075] The acquisition unit 11 acquires item information corresponding to the missing items of the target document. The item information is information on the description content corresponding to the missing items, and specifically includes description examples showing specific description content corresponding to the items, explanations on the items, sentences or paragraphs showing the gist of the items, evaluation comments on the missing items of the target document, etc. In this embodiment, description examples registered in advance in association with the items are acquired from the database DB. The acquisition unit 11 also acquires auxiliary information. In this embodiment, the user terminal 2 accepts input of auxiliary information from the user, and the acquisition unit 11 acquires the auxiliary information via the user terminal 2. In this embodiment, the acquisition unit 11 acquires target document information indicating the contents of the target document being created as auxiliary information. In addition, the acquisition unit 11 of this embodiment further acquires an item type indicating the type of the missing item and a correction strength selected by the user.
[0076] The evaluation unit 12 acquires evaluation information indicating the evaluation result of the target document, and passes it to the acquisition unit 11. The evaluation unit 12 of this embodiment acquires the evaluation information by performing an evaluation based on the target document information acquired by the acquisition unit 11, but may be configured to acquire the evaluation information from an external device or the user terminal 2. Details of the evaluation process will be described later.
[0077] The identification unit 13 identifies an example of description to be used for creating or correcting the target document from among the examples of description corresponding to the item type of the missing item in the target document, from a database DB that stores a plurality of examples of description in association with the item type. The identification unit 13 may further identify the example of description based on the correction strength selected by the user.
[0078] The generation unit 14 inputs the item information and auxiliary information acquired by the acquisition unit 11 into the sentence generation model and acquires the output. The acquisition unit 11 in this embodiment generates a generation command, which instructs the generation of a draft sentence for the missing item based on the item information and auxiliary information, as a sentence in a natural language and transmits it to the generation device 3. The transmission of this generation command corresponds to the input of the item information and auxiliary information into the sentence generation model in the present invention. The details of the generation process of the generation command will be described later.
[0079] The suggestion unit 15 outputs suggested information for assisting in the creation of missing items in a target document based on the output of the sentence generation model in response to a generation command. In this embodiment, the suggested information is output by transmitting it to the user terminal 2. In this embodiment, the suggested information includes a draft sentence corresponding to the missing items generated by the sentence generation model, and is presented to the user on a suggestion screen (described later) displayed on the user terminal 2.
[0080] The database DB stores a plurality of description examples in association with item types. In this embodiment, the description examples stored in the database DB are descriptions in legal documents created in the past in the process of agreement, and descriptions in past legal documents are registered in association with item types. In addition, the description examples in the present invention are not limited to legal documents that were actually created in the process of past agreements. For example, a company's own legal document template may be registered in the database DB as a description example corresponding to the item type.
[0081] An item type is a type that classifies items according to their purpose and specifies what each item defines. For example, as described in 1. Definition, in this embodiment, an issue is an item. Therefore, in this embodiment, the type of issue or issue is the item type, and in the database DB, example descriptions are registered in correspondence with the type of issue or issue, and each example description shows an example of the content of the description within the range related to each issue. Note that if clauses are used as items, the content of each clause can be registered as an example description.
[0082] For example, for each contract type, the types of issues that may be included in the contract can be specified. More specifically, for a contract type such as a "confidentiality agreement," groups of issues are assumed to be types of issues, such as exceptions to confidential information, liability for damages, and anti-social clauses. Note that a group here corresponds to a group of clauses that stipulate a set of regulations. Each type of issue includes one or more individual issues. For example, in the type of "exceptions to confidential information," the specific items stipulated as exceptions are an example of an "issue." For each issue, multiple examples may be registered.
[0083] FIG. 4 is an example of example information stored in the database DB in this embodiment. The example information includes an example ID, a document ID, an item type, a correction strength, and an example. In this embodiment, the example is the description content of a portion corresponding to each item type of a legal document created in a past agreement, or a template according to the item type. The document ID is an ID indicating the document on which the example is based when the example is a legal document created in a past agreement. When the example is a template, an ID indicating that it is a template is stored as the document ID in the example information. The data structure can be changed arbitrarily.
[0084] In addition, information on the background of the document on which each example is based, which corresponds to the above-mentioned background information, such as the creation date and time of the document on which the example is based, the person who entered the document into the database DB, the legal officer in charge of the document, the contract type, the contracting party (the other party to the contract), the company's position, the subject of the contract, etc. Also, such information may be stored separately by linking it to the document ID.
[0085] The contract type is an example of the type of arrangement specified by the document, and other types of arrangements include atypical contracts, types of rules or regulations, etc. The company's position refers to its contractual position, such as information such as the discloser or recipient of information in a confidentiality agreement. The subject of the contract is an example of the "subject of the arrangement," and refers to the products, services, or business that are the subject of each contract (arrangement). For example, the target products in a sales contract would be an example of the subject of the contract.
[0086] Note that the above-mentioned units and part of the database DB of the support device 1 may be provided in a computer other than the support device 1, such as a user terminal 2 or a generating device 3. In addition, the support device 1 may include a display unit and a generating device 3.
[0087] <5. Processing Procedure> Next, referring to FIG. 5, a processing procedure for creating and amending a target document using the support system 0 of this embodiment will be described. FIG. 5 is a sequence diagram showing an example of a processing procedure by the support device 1, the user terminal 2, and the generation device 3 in the support system 0 of this embodiment. Note that the processing shown here is an example, and may be omitted or changed without changing the gist of the present invention. In the following example, a case will be described in which both the creation of an initial draft and the creation of an amendment draft of a document are performed by the present invention, but the present invention may perform only either the creation of the initial draft or the amendment. In addition, the creation of the initial draft and the creation of the amendment draft do not need to be performed by the same model, and may be performed using different models.
[0088] (1) Creating the initial draft Steps S501 to S506 are processes related to the creation of a first draft of the target document, that is, the creation of a draft document based on auxiliary information that does not include the target document information. Note that this process is not essential, and the processes from step S507 onwards may be performed using a draft of the target document created by the user himself, a draft generated using an external machine learning model, or a contract draft received from a party to whom a contract is to be concluded.
[0089] First, in step S501, the user terminal 2 accepts an input of auxiliary information from the user and transmits it to the support device 1. The auxiliary information may be, for example, background information. In step S501, since the target document is in the initial draft creation stage, the auxiliary information does not include target document information indicating the contents of the target document. On the other hand, the auxiliary information transmitted in step S501 needs to include information for identifying the item type of the item to be created (missing item). For example, the acquisition unit 11 may acquire information indicating the item type as part of the background information by having the user select the item type of the item to be created (missing item), or the support device 1 may identify the item type of the item to be created based on the background information.
[0090] Specifically, for example, when concluding a confidentiality agreement when presenting information to a business partner, examples of background information acquired in step S501 include the relationship with the other party as "discloser of information," the type of arrangement as the contract type "confidentiality agreement," and the degree of revision as the request of the parties as "want to make a strong contract (advantageous to one's position)." It is also preferable for the acquisition unit 11 to acquire the names or titles of the contracting parties and the designations in the contract (Party A, Party B), etc., as definition information.
[0091] A typical input format for this auxiliary information is assumed to be one in which the user selects from options. In addition, the auxiliary information may be input as text by the user freely writing sentences expressed in natural language. Furthermore, this auxiliary information may be acquired from term sheet information indicating the contents of a term sheet used in contract negotiations, or question information based on questions about the contents to be included in a legal document and the answers to those questions.
[0092] Next, in step S502, the identifying unit 13 first identifies an example of description to be used in creating the target document from among the examples of description stored in the database DB, which corresponds to the item type of the missing clause. If the acquiring unit 11 acquires a correction strength indicating the strength of correction as a request from the parties, the identifying unit 13 may identify an example of description to be used in creating the target document from among the examples of description stored in the database DB, which corresponds to the item type of the missing clause and corresponds to the acquired correction strength. The generating unit 14 generates a generation command to cause the sentence generation model to generate a draft sentence regarding the missing clause, based on the identified example of description and auxiliary information.
[0093] Since the sentence generation model of this embodiment is a conversational model that uses a sentence in a natural language as an input and generates a sentence as a response, the generation command is generated as a sentence in a natural language. More specifically, the generation unit 14 can generate a generation command by generating a sentence that combines a prepared instruction sentence, a wording specified based on the auxiliary information, and a description example. For example, an instruction sentence may be prepared in advance for each option of the auxiliary information and stored in the storage unit of the support device 1, an instruction sentence may be selected from the auxiliary information according to the contract type, and the generation unit 14 may generate a sentence that combines the instruction sentence, the wording specified based on the auxiliary information, and a description example as a generation command. Note that when the acquisition unit 11 acquires a correction strength indicating the strength of correction at the request of the parties, instead of or in addition to using the correction strength to specify the description example, the generation unit 14 may generate a sentence that combines words corresponding to the acquired correction strength as a generation command.
[0094] More specifically, when auxiliary information is input by selection, a word corresponding to the selected auxiliary information is first identified. For example, by setting a corresponding word in advance for each option of auxiliary information, the word corresponding to the selected auxiliary information can be identified. When a phrase corresponding to the correction strength is to be included in the generation command, the word corresponding to the correction strength is also identified in the same way. Then, the generation unit 14 generates a generation command based on those words, the description example, and an instruction statement separately registered in advance in the storage unit of the support device 1. On the other hand, when auxiliary information is input as free text, a generation command may be generated using the input auxiliary information, the description example, and the instruction statement prepared in advance.
[0095] Here, it is preferable to store a character string describing a rule of document creation corresponding to the auxiliary information as a word corresponding to the auxiliary information. For example, in a legal document, the legal meaning may differ between describing an exception in a sentence of a provision and describing a disclaimer separately after the sentence. By combining words that specify rules of such description methods according to the auxiliary information, an appropriate legal document can be created. In addition, the support device 1 may determine a document creation rule to be included in the generation command based on the auxiliary information from among a plurality of document creation rules corresponding to each auxiliary information, or may be configured to include a plurality of document creation rules in the generation command and include a statement to use an appropriate document creation rule from the plurality of document creation rules based on the auxiliary information in the generation command, thereby determining the document creation rule to be used by the text generation model.
[0096] As the instruction sentence, for example, "Create a document based on the following description example based on the following conditions," "Input auxiliary information and description examples showing the arrangement of the contract. Based on the following rules, modify the description example according to the auxiliary information to create an appropriate contract sentence," and other sentences can be stored in the storage unit of the support device 1 in advance. Alternatively, an instruction sentence can be prepared for each contract type as an option for auxiliary information, and information corresponding to the auxiliary information can be included in the instruction sentence, such as "Based on the following conditions, create a non-disclosure agreement based on the following description example." Then, by generating a sentence in which the instruction sentence is followed by words and description examples based on the auxiliary information, a generation command can be generated to instruct the sentence generation model to create a draft.
[0097] Once the generation command has been generated, the process proceeds to step S503, where the generation unit 14 transmits the generation command to the generation device 3. The transmission of the generation command to the generation device 3 in step S503 corresponds to inputting the description example and auxiliary information to the text generation model. Upon receiving the generation command, the generation device 3 inputs the generation command to the text generation model. As a result, the text generation model outputs a first draft of the contract as a draft text regarding missing items (all items required for the target document) (step S504).
[0098] In step S505, the generation device 3 transmits the draft sentence output by the sentence generation model to the support device 1. Then, in step S506, the suggestion unit 15 transmits an initial draft of the target document as proposed information to the user terminal 2 based on the draft sentence output by the sentence generation model. Through the above process, the user can obtain an initial draft of the target document by inputting auxiliary information.
[0099] (2) Acquisition of evaluation information Next, a process related to obtaining evaluation information of the target document will be described. Steps S507 to S509 are processes related to the evaluation of the target document. Note that this process is for obtaining evaluation information indicating the evaluation of the target document. In this embodiment, as described below, the evaluation unit 12 evaluates the target document, but the evaluation information may be obtained by having an external device evaluate the target document, or by receiving the evaluation information together with auxiliary information from the user terminal 2. Note that steps S506 and S507 may be omitted, and the process may be performed consistently up to the evaluation of the target document without waiting for an instruction from the user.
[0100] In step S507, the user terminal 2 transmits an evaluation request to the support device 1. The evaluation request includes auxiliary information. The auxiliary information transmitted here includes target document information including at least a part of the contents of the target document to be evaluated, and background information. In this embodiment, a document file containing a draft of the contract is uploaded as the target document information.
[0101] In the support device 1, the acquisition unit 11 receives an evaluation request, and the evaluation unit 12 evaluates the target document based on the target document information and background information. The target document may be evaluated by any method, such as evaluation using a machine learning model or rule-based evaluation. As an example, in this embodiment, a method of evaluating clauses based on rules in accordance with background information is used.
[0102] In particular, in this embodiment, the evaluation is performed from multiple predefined viewpoints using the contract type and contractual position of the background information. The contract type may be determined by user selection, or may be estimated from the title of the sentence in the target sentence or the title of each clause. Furthermore, the title of each clause in the target document may be used in the evaluation. Specifically, the content of each clause and the relationship between multiple clauses are evaluated. Here, multiple clauses refer to a group of multiple clauses that defines a set of regulations. For example, within a group, the presence or content of clause B is determined according to the content of clause A, and clause C is confirmed based on the result, and by specifying that the evaluation process branches according to the content, the relationship between multiple clauses can be evaluated on a rule basis. The content of each clause can be evaluated on a rule basis according to a predefined rule, for example, whether a specific keyword is included, or the relationship between specific keywords (whether keyword X is used in the affirmative or negative sense, etc.).
[0103] The evaluation of the target document is performed from the viewpoints of whether it contains unnecessary descriptions (or clauses, descriptions, etc. that are disadvantageous to the party in the relationship with the other party in the background information), whether there are any aspects that should be added (missing items), whether the contents of the clauses are appropriate, etc. For example, for missing items, a rule-based judgment is made as to whether each item is included in the target document according to an item template indicating a list of items prepared in advance for each contract type and contractual position, and an item that is not included is determined as a missing item. Note that the judgment may be made according to a rule for each item, or multiple items may be judged according to a single rule. In this embodiment, particularly, an item template is prepared for each contract type, and unnecessary descriptions and missing items are evaluated according to the item template corresponding to the contract type.
[0104] In this embodiment, the above rules are linked to the issues that are to be pointed out in the case of violating the rules, and the issues to be pointed out for the target document are registered in advance in the storage unit of the support device 1, and the issues to be pointed out for the target document are identified by checking the rules. Specifically, for each of the above-mentioned evaluation rules, the issues, the types of points pointed out, the evaluation comments, and the risk level are stored in advance in the storage unit. In this embodiment, there are three types of points pointed out: "insufficient item", which means that an item (point of content) that should be included in the target document is missing; "correction of clause", which means that the contents of the clause are incomplete; and "deleted item", which means that unnecessary statements, clauses, or points of content exist. In addition, the type of "insufficient item" is further divided into "partial deficiency", which means that some clauses in a group are missing, and "total deficiency", which means that the entire group is missing, for a group of multiple clauses that define a set of regulations.
[0105] An evaluation comment is a natural language sentence or phrase that suggests modifications to the contents of the target document or points out problems, such as "Is it necessary to stipulate a consultation clause?" or "Isn't the period for which confidentiality obligations are imposed set longer than necessary?". In addition, the evaluation comment may contain an explanatory text about the clause related to the point or an example of a description corresponding to the point at issue. The risk level is information that indicates the degree of risk assumed for each point. The risk level may be fixed (unchangeable) on the system side, or may be changeable by user input.
[0106] Examples of risk level settings are as follows. Issues that are likely to become problems after the contract and that are related to important rights and obligations of the company can be set to risk level "A". For example, issues related to "intellectual property ownership", "scope of damage compensation", "definition of confidential information", etc. are candidates for issues with risk level "A". Issues that are unlikely to become problems even if they are missing can be set to risk level "C". For example, "confidential information exceptions" and "intermediate reporting obligations" are included. Among the multiple groups, issues that do not fall under A and C can be set to risk level "B". In addition to the above example, the risk level may be set in advance for each issue, or may be a risk determined each time based on the wording of the clause. For example, the evaluation unit 12 may determine whether the company is favorable or unfavorable based on the wording of the clause, and the risk level may be set based on the determination.
[0107] As described above, the evaluation unit 12 executes the evaluation process to generate evaluation information indicating an evaluation of the target document. The evaluation information includes, for each of one or more issues identified in the above process, a clause in the target document corresponding to the issue, an evaluation comment, and a risk level. Also, when obtaining evaluation information from an external device or the user terminal 2, evaluation information is obtained for each issue, including a clause in the target document corresponding to the issue, an evaluation comment, and a risk level.
[0108] The target document may be evaluated using a text generation model. The text generation model may be the same as or different from the text generation model used for draft creation or revision. For example, the acquisition unit 11 acquires the target document information and background information, and the evaluation unit 12 reads out the evaluation rules used for evaluating the target document stored in advance based on the background information, generates an evaluation command based on the target document information and the evaluation rules according to the background information, and transmits the generated command to the generation device 3 to obtain the evaluation result (evaluation information) of the target document generated by the generation device 3. The evaluation command is generated as a sentence in a natural language. This is expected to reduce the effect of the influence on the evaluation due to the notation variation and the difference in expression of the sentence. In addition, it is expected that an appropriate evaluation can be performed by using the evaluation rule (e.g., the evaluation rule according to the item template set for each contract type) read out based on the background information (e.g., the contract type) from among the multiple evaluation rules. In addition, the evaluation rule may be a rule that determines which of predetermined points of concern should be added to the target document from the content of the target document, or it may be an evaluation rule that generates a sentence as a point of concern to be added to the target document from predetermined criteria for each point of concern and the content of the target document.
[0109] Furthermore, both the evaluation by the sentence generation model and the evaluation by the rule base may be used in combination. In this case, if the evaluation by the sentence generation model and the evaluation by the rule base are different, it is preferable to prioritize the evaluation by the rule base. This makes it possible to guarantee a minimum evaluation by the rule base even if there is uncertainty in the evaluation by machine learning. Note that the support device 1 may select an evaluation rule to be included in the generation command based on the background information from among a plurality of evaluation rules corresponding to each piece of background information. Also, a configuration may be adopted in which a plurality of evaluation rules are included in the evaluation command, and an evaluation command is included to include an indication that an appropriate rule is to be used based on the background information from among the plurality of evaluation rules, thereby determining the evaluation rule to be used by the sentence generation model.
[0110] When the evaluation unit 12 finishes evaluating the target document, in step S509, the evaluation unit 12 transmits evaluation information to the user terminal 2. When the user terminal 2 receives the evaluation information, the evaluation information is subjected to display processing, and the processing result is shown on the display.
[0111] Fig. 6 shows an example of the evaluation screen displayed on the user terminal 2 as a result of the evaluation information display process. The evaluation screen W6 displays an overview of the target document and an overview of the evaluation results at the top of the screen. The contents of the target document identified by the target document information are displayed on the left side of the screen. In addition, a "Edit read contents" button is provided, and if there is an error in the contents of the target document information, the contents can be edited by selecting this button.
[0112] The right side of the screen displays the evaluation results based on the evaluation information. For each point of concern, the type of point of concern (indicated as "Omission" or "Correction" in Figure 6) indicating whether the point of concern is an unnecessary item, a missing item, or a content correction, and evaluation comments including examples of clauses corresponding to the point of concern and explanations are displayed.
[0113] For each pointed out item, a differently shaped point out mark W61 is attached for each point out type, and a serial number for each type is displayed. Similar point out marks W61 are also displayed on the left side of the screen in the corresponding part of the pointed out item in the target document together with the serial number, so that the user can recognize the position in the target document that corresponds to the pointed out item, and the corresponding part in the target document is highlighted or the like. In the example in Figure 6, circles indicate points that require amendments to clauses, and triangles indicate points that require missing items.
[0114] The indication mark W61 is displayed in a different color depending on the risk level. By operating the risk filter operation section W62, the indications to be displayed can be filtered according to the risk level. In FIG. 6, the indication is displayed when the risk level is "B" or higher, that is, indications of A or B are displayed.
[0115] Additionally, a Batch Correction button W63 and an Individual Correction button W64 are displayed together with the evaluation results. When the user selects these buttons, a correction request is sent in step S510 in Fig. 5. The range of correction instructions varies depending on the selected button.
[0116] There are three types of batch correction buttons W63: "Create revision proposals for all items at once," "Create revision proposals for item A," and "Create revision proposals for items AB." When "Create revision proposals for all items at once" is selected, a correction request is sent for all issues. When "Create revision proposals for item A" is selected, a correction request is sent for issues with a risk level of A. When "Create revision proposals for items AB" is selected, a correction request is sent for issues with a risk level of A and issues with a risk level of B. When the individual correction button W64 displayed for each issue is selected, a correction request is sent for that issue.
[0117] In addition, the system may accept inputs requesting collective corrections according to the type of findings. For example, in the case of findings of missing items, the system may accept inputs instructing the creation of draft text for the missing items collectively, or inputs instructing the correction of clauses collectively.
[0118] In this embodiment, when the collective correction button W63 or the individual correction button W64 is selected, further requests for correction from the parties are accepted. For example, for a problem that has multiple possible correction policies, the user can select one of them to accept a request for correction. Alternatively, when a correction request is made for multiple problems using the collective correction button W63, the strength of correction may be selected, such as "strong (highly favorable to the company)", "neutral", or "weak (more favorable to the company than before the correction, but more favorable to other companies)". At this time, a command may be generated to generate only a correction proposal (when the type of the problem is "correction of clauses") or a draft sentence (when the type of the problem is "missing items") corresponding to the correction strength selected by the user, or a command may be generated to generate a correction proposal or draft sentence for all correction strengths. In addition, such a request may be input by free text.
[0119] Here, when making a request to correct some of the identified matters using the "Make revision proposal for item A" and "Make revision proposal for items AB" of the Bulk Correction button W63, or the "Individual Correction button W64," the correction request includes only the matters related to the identified matters (item information and the target document information for the corresponding part). This makes the correction command shorter when correcting some of the items such as items A and AB than when correcting all items (the entire target document) at once, and further when correcting one item by specifying one identified matter. Furthermore, by correcting parts individually, it is possible to reduce the occurrence of consistency problems that may accompany the correction of multiple pointed out items, for example, when multiple pointed out items are included in a single sentence of the target document. Furthermore, by allowing the user to individually select and modify the necessary portions, the processing time is reduced, and the time required for modification can be reduced. These actions are expected to have the effect of improving the user experience, so it is more preferable to obtain item information and auxiliary information and make corrections for some of the multiple issues, especially for the parts that correspond to the issues selected by the user.
[0120] (3) Corrective action Next, the correction process of the target document will be described. When a correction request is received as described above, in, the acquisition unit 11 acquires the target document information and evaluation information from the user terminal 2 as auxiliary information. As the target document information, the contents of the entire target document may be acquired, or the contents of only the clause corresponding to the pointed out matter that is the preamble or the correction request, or the clauses before and after that may be acquired. The clause corresponding to the pointed out matter is, for example, a clause related to the missing item when the pointed out type is a missing item, and is a clause that is the subject of correction when the pointed out type is a clause correction. In particular, when the pointed out type is a missing item, the target document information includes information indicating the insertion position of the missing item in the target document and the description before and after the insertion position.
[0121] More specifically, when the type of indication is partial deficiency, the contents of other clauses belonging to the same group as the clause related to the missing item are acquired as the target document information. When the type of indication is amendment of a clause, if the clause to be amended is included in any clause group, it is preferable that the contents of other clauses belonging to the same group are included in the target document information. The information indicating the insertion point is information indicating the location in the target document where the missing item should be inserted, for example, information such as "between clause A and clause B." At this time, the contents of the statements before and after the insertion point are, for example, the contents of clause A and clause B. The insertion point is determined by predetermining the items and insertion rules related to the missing items (for example, if clause B is missing, then below clause A related to clause B, etc.), and determining the insertion point according to the related items and insertion rules. The insertion location may be determined by defining missing items and fulfilling items (items determined to be included in the target document) based on an item template that shows a list of items prepared for each contract type or contractual position, and then using an insertion rule based on the positional relationship on the list of missing items and fulfilling items (for example, determining the insertion location based on the list position of the missing clause and the list position of the fulfilling clause that satisfies the condition).
[0122] The acquisition unit 11 of this embodiment further acquires background information and the correction strength transmitted together with the correction request in step S510. In this embodiment, the background information received from the user terminal 2 in step S501 or step S507 and the evaluation information generated by the evaluation unit 12 in step S508 are stored in the storage unit, and the acquisition unit 11 further acquires the evaluation information and background information based on them. Note that when the target document is created and evaluated by an external device, the acquisition unit 11 may acquire the target document information and the evaluation information from the user terminal 2 or the external device.
[0123] In step S511, the acquisition unit 11 acquires a description example from the database DB. The description example used to create a proposed amendment to the clause or a draft statement regarding the missing item is selected from the description examples registered in the database DB based on the item type of the issue corresponding to the pointed out matter. Specifically, the acquisition unit 11 first identifies the item type of the missing item based on the evaluation information. Then, the identification unit 13 identifies, from the database DB, a description example that matches the amendment strength acquired by the acquisition unit 11 among the description examples corresponding to the item type. Then, the acquisition unit 11 acquires the description example from the database DB. Alternatively, instead of the description example, the acquisition unit 11 may acquire an evaluation comment included in the evaluation information as item information. The evaluation comment here may be a sentence or paragraph in natural language that suggests an amendment to the contents of the target document or points out a problem, and may not include a description example.
[0124] Then, the generation unit 14 generates a generation command for causing the text generation model to generate a proposed revision of the target document or a draft of the missing items based on the description example (item information), the target document information, the evaluation information, the background information, and information on the parties' requests regarding revision. Since the text generation model of this embodiment is a conversational model that receives a text in a natural language as input and generates a text that will be an answer, the generation command is generated as a text in a natural language.
[0125] The generating unit 14 of this embodiment generates a command to be sent to the generating device 3 according to the content of the findings related to the evaluation request in step S507. If the type of the findings is a clause correction, a generating command is generated to instruct the generating device to generate a correction proposal related to the findings, based on target document information indicating the content of the part of the target document corresponding to the findings and background information. Then, by sending this to the generating device 3, the target document information and background information are used as inputs, and a correction proposal is obtained as an output from the generating device 3. Note that in addition to this, a description example may also be used as an input.
[0126] On the other hand, when the type of the pointed out item is a missing item, a generation command is generated to instruct the generation of a draft text for the missing item based on the item information, for example, the description example identified by the identification unit 13, and the auxiliary information. Then, by sending this to the generation device 3, the description example and the auxiliary information are input, and a revision proposal is obtained as an output from the generation device 3. Here, as the auxiliary information, some or all of the target document information, background information, writing style information, definition information, and evaluation information can be used.
[0127] Specifically, a sentence is generated as a generation command that describes instructions for generating a draft sentence for the correction or missing items for each pointed out item. However, for example, if there are multiple pointed out items in one clause or a group of clauses that defines a set of regulations, generating a generation command for each pointed out item may cause problems such as generating multiple contradictory correction proposals or draft sentences. Therefore, for generation commands related to clauses or groups of clauses with multiple pointed out items, the generation unit 14 of this embodiment generates a generation command sentence for each clause or group of clauses that corresponds to the multiple pointed out items.
[0128] The generation of the generation command will be specifically described below. First, for background information, when background information is input by selection, a word corresponding to the selected background information is identified. For example, by setting a corresponding word in advance for each option of background information, it is possible to identify the word corresponding to the selected background information.
[0129] Similarly, when request information is input by selection, first, a word corresponding to the selected request is identified. For example, by setting in advance corresponding words for each combination of the target pointed out matter and the request option, it is possible to identify a word corresponding to the selected request for each pointed out matter.
[0130] Then, the generating unit 14 generates a generation command based on those words, the target document information, the description example (item information), and a command statement separately registered in advance in the storage unit of the support device 1. On the other hand, when the background information and the request are input as free text, the generating command may be generated using the description example (item information), the input background information, and a command statement prepared in advance.
[0131] Here, similarly to the generation of the generation command in the creation of the first draft of the document, it is preferable to store a character string describing the rule of document creation according to the background information as a word corresponding to the background information. For example, in a legal document, the legal meaning may differ between describing an exception in a sentence of a provision and describing a disclaimer separately after the sentence. By combining words specifying the rule of such a description method according to the background information, an appropriate legal document can be created. Note that, among the multiple rules corresponding to each background information, the support device 1 may determine the rule to be included in the generation command based on the background information, or the multiple rules may be included in the generation command and a statement to use an appropriate rule from the multiple rules based on the background information may be included in the generation command to determine the rule to be used by the text generation model.
[0132] It is preferable that the instruction sentence is different depending on the type of the indication. For example, as an instruction sentence for amending a clause, sentences such as "The following indications were made on the above draft as a result of the review. Please prepare a proposed amendment based on the following conditions, taking into account the indications," and "The following indications were made on the above draft as a result of the review. Please prepare an appropriate contract document based on the indications, referring to the example below," can be stored in advance in the storage unit of the support device 1. Then, a sentence combining the example, the target document information, the instruction sentence, and words based on the background information and evaluation information can be generated as a generation command for instructing the sentence generation model to generate a proposed amendment to the clause.
[0133] For example, instructions regarding missing clauses may be stored in advance in the storage unit of the support device 1, such as "The above draft is missing the following points. Please create a draft for the missing points based on the following conditions and with reference to the example description," or "The following comments were made about the above draft as a result of reviewing it. Please create a draft for the clause with the missing points added based on the comments and with reference to the example description below." Furthermore, it is preferable to prepare different instructions for partial and complete deficiencies. Then, a sentence that combines the example description (item information), target document information, instructions, and words based on background information and evaluation information may be generated as a generation command for instructing the sentence generation model to generate a draft for the missing clause.
[0134] Once the generation command is generated, the process proceeds to step S512, where the generation unit 14 transmits the generation command to the generation device 3. Note that the generation command for each pointed out item, each clause, or each group of clauses may be transmitted in multiple batches. For example, the configuration may be such that the directive portion of the generation command is transmitted first, followed by words corresponding to the pointed out items or requests. Also, the configuration may be such that it is determined whether to transmit the generation command all at once or in multiple batches depending on the number of characters in the generation command.
[0135] When the generation device 3 receives the generation command, it inputs the generation command to the sentence generation model. As a result, the sentence generation model outputs a correction proposal for the clause of the target document or a draft sentence for the missing item (step S513). Note that the generation unit 14 may generate a generation command in step S511 to instruct the generation unit 14 to propose multiple correction proposals or draft sentences for the missing items for one pointed out item.
[0136] Furthermore, a generation command may be sent to generate reasons for revision based on the evaluation information, and the reasons for revision may be obtained in addition to the proposed revision or draft text. Furthermore, the reasons for revision may be generated and obtained as reasons for revision for one's own company (or for one's own client) and reasons for revision for the other party. In contract negotiations, revisions are made repeatedly between one's own company and the other party, and at this time, at least one of reasons for revision to be explained within the company and reasons for revision to be explained outside the company is often written down. Writing reasons for revision requires a high level of expertise, but the above configuration can support even users without specialized knowledge to write reasons for revision.
[0137] In step S514, the generation unit 14 acquires the proposed amendment or the proposed sentence generated by the generation device 3. Then, the proposal unit 15 generates proposal information based on the proposed amendment or the proposed sentence, and in step S515, transmits the proposal information to the user terminal 2. In the user terminal 2, a proposal screen based on the proposal information is displayed.
[0138] (4) Reevaluation process Although omitted in FIG. 5, in this embodiment, the quality of the proposal can be further improved by repeating the process of acquiring evaluation information in step S508 and the process of correcting in steps S511 to S514.
[0139] The evaluation unit 12 of this embodiment receives a proposed revision or draft sentence for the target document from the generation device 3 as an output of the sentence generation model in response to a generation command. Then, it obtains evaluation information indicating an evaluation of the entire target document when the proposed revision or draft sentence is reflected, and judges whether the evaluation satisfies a preset standard. Note that in this embodiment, the evaluation unit 12 executes a re-evaluation and generates evaluation information as a result, but it may also transmit the contents of the target document after the revision to an external evaluation device and obtain evaluation information from the evaluation device as the evaluation result.
[0140] If the evaluation satisfies the criteria, the evaluation unit 12 causes the proposing unit 15 to transmit the proposal information. On the other hand, if the evaluation does not satisfy the criteria, the generation unit 14 generates a generation command to instruct the revision of the revision or the draft sentence based on the information indicating the content of the revision or the draft sentence, the evaluation information, the background information, and the description proposal. Then, the generation unit 14 transmits the generation command to the sentence generation model, whereby the revision of the revision or the draft sentence is further generated by the sentence generation model.
[0141] The evaluation unit 12 of this embodiment is configured to repeat obtaining and evaluating a revision proposal until the evaluation of the revision proposal satisfies a predetermined criterion or a termination condition is satisfied. As the predetermined criterion, for example, it is possible to set a risk level for each of the issues mentioned below contained in the evaluation information that is equal to or lower than a predetermined risk level. As the termination condition, for example, it is assumed to set an upper limit on the number of repetitions or a lower limit on the number of characters changed by the revision. Hereinafter, the process related to obtaining evaluation information and repeating revisions will be described with reference to the flowchart of FIG. 7.
[0142] First, when the assistance device 1 receives a revision proposal or a draft sentence, which is the output of the sentence generation model, the evaluation unit 12 acquires this revision proposal or draft sentence in step S701.
[0143] Then, in step S702, reevaluation is performed. Specifically, the proposed revision is reevaluated in the same procedure as in the evaluation process described above. Evaluation information indicating the evaluation of the proposed revision is then generated in the same manner as in step S508. The reevaluation may be performed not on the entire target document, but only on the portion related to the previously pointed out issue. Specifically, the clause or group of clauses indicated by the proposed revision may be reevaluated, and the reevaluation of the other portions may be omitted. This is expected to have the effect of improving processing speed.
[0144] Next, in step S703, it is determined whether the evaluation of the proposed revision satisfies a preset criterion. In this embodiment, the risk levels of the identified items in the evaluation information are checked, and it is determined that the criterion is met if the risk levels of all identified items are equal to or lower than a predetermined risk level.
[0145] If the criteria are not met, the process proceeds to step S705, where it is confirmed whether the number of loops (number of revisions) has reached the upper limit. If the upper limit has not been reached ("Y" in step S705), the process proceeds to step S706, where the evaluation unit 12 causes the generation unit 14 to generate and transmit a generation command again, increments the loop count counter, and the process returns to step S701.
[0146] If the criteria are met in step S703, or if the number of revisions reaches the upper limit in step S705, the process proceeds to step S704, where the evaluation unit 12 causes the proposal unit 15 to generate proposal information. The proposal unit 15 then transmits the proposal information to the user terminal 2 (corresponding to step S515 in FIG. 5) and ends the process. Note that the above processing procedure is an example, and may be arbitrarily changed, for example, by switching the judgment order of steps S703 and S705.
[0147] In principle, the loop process is performed until a preset criterion is met, but if the number of repetitions becomes extremely large or if it becomes impossible to exit the loop, there is a concern that the processing load will increase. Therefore, in this embodiment, an upper limit on the number of loop processes is set as an end condition, and even if the reevaluation does not meet the criterion, if the end condition is met, the repetition of the reevaluation is terminated, the loop process is terminated, and the process proceeds to step S515. In addition, other end conditions may be set, such as, for example, the number of characters changed by comparing the documents before and after correction is equal to or less than a certain number. This makes it possible to prevent the repetition of a situation in which only minor corrections are made.
[0148] (5) Proposal screen Next, a proposal screen displayed on the user terminal 2 as a result of the display process of the proposal information will be described with reference to Fig. 8. Fig. 8 is a display example of the proposal screen in this embodiment. The proposal screen W8 in this embodiment is displayed in a similar configuration to the evaluation screen W6 shown above, and the suggestion mark W81 and risk filter operation section W82 are similar to the suggestion mark W61 and risk filter operation section W62 in the evaluation screen W6, so their explanation will be omitted.
[0149] The difference from the evaluation screen W6 is mainly in the display content on the right side of the screen, especially the part related to "correction proposals." The proposal screen W8 includes correction proposals W83 for each pointed out item as a display based on the proposed amendments or proposed text regarding missing clauses included in the proposal information. The correction proposals W83 display the above-mentioned correction proposals and also display the text so that it can be copied. Furthermore, if the support program of the present invention is provided as an add-on (add-in, plug-in) program for document editing software, a correction button may be provided instead of a copy button. In this case, when the user operates the correction button, the corrections based on the proposed amendments are made directly on the document editing software.
[0150] It is also preferable to make it easy for the user to recognize the changes from the draft by highlighting them, etc. This allows the user to obtain the contents of each pointed out item and a draft of the amendment or deficiency, so that even a user who is not familiar with legal knowledge can amend the legal document by himself / herself.
[0151] In addition, when multiple suggested revisions or draft sentences are obtained for a pointed out item, since the proposal information contains multiple suggested revisions or draft sentences for the pointed out item, it is preferable to display multiple suggested revisions W83 for the pointed out item on the proposal screen W8 as well so that the user can select from them.
[0152] The method of outputting the proposed information is not limited to displaying the proposal screen W8 described above, and the proposal unit 15 may transmit, for example, a document-format file of the entire target document to which proposed revisions have been applied to at least some of the pointed-out items to the user terminal 2. In addition, the user may be prompted to select whether or not to reflect the proposed revisions displayed on the proposal screen W8, or which proposed revisions to reflect, and the proposal unit 15 may further transmit, to the user terminal 2, a document-format file in which the entire draft has been revised based on the result.
[0153] As described above, the support system of this embodiment can support a user who does not have specialized knowledge to make revisions to a draft of a legal document such as a contract. In addition, for missing items, an appropriate draft can be created based on the proposed description. Furthermore, the quality of the revision of the target document can be improved by reevaluating the proposed revision or the draft and issuing an instruction to create another proposed revision based on the result of the evaluation.
[0154] In addition, by generating commands in natural language text, it is possible to create proposed revisions or draft texts using a conversational text generation model that takes text as input and generates answer texts. This makes it possible to have the model generate proposed revisions or draft texts that take into account various conditions such as the background and requests of the contract, making it possible to provide proposed revisions or draft texts for the target document that are more in line with the user's situation and wishes.
[0155] Furthermore, a generation command may be sent to the sentence generation model to generate reasons for revision based on the evaluation information, and the reasons for revision may be included in the proposal information. Furthermore, the reasons for revision may be generated as reasons for revision for one's own company (or for one's own client) and reasons for revision for the other party. In contract negotiations, one's company and the other party make repeated revisions to each other, and at this time, at least one of reasons for revision to be explained within the company and reasons for revision to be explained outside the company is often written down. Writing reasons for revision requires a high level of expertise, but the above configuration can support even users without specialized knowledge to write reasons for revision. [Explanation of symbols]
[0156] 0: Support system 1: Support device 11: Acquisition part 12: Evaluation section 13: Specific part 14: Generation part 15: Proposal Department 2: User terminal 3:Generation device 10: Information processing device 101: Control unit 102: Storage section 103: Communications Department 90: Terminal equipment 901: Control unit 902: Storage section 903: Communications Department 904: Input section 905: Output section NW: Network W6: Evaluation screen W61: Pointing out mark W62: Risk filter operation section W63: Bulk edit button W64: Individual edit button W8: Proposal screen W81: Point out mark W82: Risk filter operation section W83: Correction proposal
Claims
1. A support program for supporting the creation or modification of a target document, The subject document is a legal document to be created or amended; A computer is caused to function as an acquisition unit, a generation unit, and a suggestion unit; The acquisition unit is Item information on the contents of the document that correspond to the missing items, which are missing in the document; Acquire auxiliary information related to the target document for making a suggestion according to the target document; The generation unit inputs the item information and auxiliary information into a sentence generation model; The suggestion unit outputs suggestion information for assisting in the creation of the missing items in the target document based on the output of the sentence generation model.
2. The support program according to claim 1 , wherein the item information includes examples of legal documents corresponding to the missing items, and the examples are descriptions or templates in documents that prescribe past agreements.
3. The computer further functions as an evaluation unit; The evaluation unit acquires evaluation information indicating an evaluation result of the target document, The assistance program according to claim 1 , wherein the acquisition unit acquires the item information based on the evaluation information.
4. The evaluation information includes an item type indicating a type of the missing item, The support program according to claim 3 , wherein the acquisition unit acquires the example corresponding to the item type of the missing item from a database that stores examples showing the contents of the items in association with the item types.
5. The evaluation information includes evaluation comments regarding the deficiencies, The assistance program according to claim 3 , wherein the acquisition unit acquires the evaluation comments as the item information.
6. The evaluation information includes one or more points of criticism regarding the target document, The problem is assigned a classification selected from a plurality of groups including a missing item and a correction of a clause; The acquisition unit determines the indication type of the indication included in the evaluation information, For items that are found to be insufficient, obtain the item information of the insufficient items related to the items that are found to be insufficient; Regarding the matters to be amended, obtain target document information indicating the contents of the target document related to the matters to be amended; The generation unit is For items where there is a shortage of items, the item information and auxiliary information are input to the sentence generation model, 4. The support program according to claim 3, wherein, for a suggested item for amending a clause, the target document information and information on an agreement stipulated by the target document are input to the text generation model.
7. The target document includes a plurality of clauses; The types of indications indicating deficiencies are divided into several types for a group of multiple clauses that define a set of regulations, including partial deficiency, where some of the clauses in the group are missing, and total deficiency, where the entire group is missing. The support program according to claim 6 , wherein the generation unit determines an input to the sentence generation model depending on whether an indicated matter relating to a group of a plurality of clauses is partially insufficient or completely insufficient.
8. When the type of indication related to the item deficiency is the partial deficiency, The support program according to claim 7 , wherein the acquisition unit acquires, as the auxiliary information, contents of other clauses in the target document that belong to the same group as the clause relating to the missing item.
9. The support program according to claim 1 , wherein the auxiliary information includes target document information indicating at least a part of the content of the target document.
10. 10. The support program according to claim 9, wherein the target document information includes information on a location in the target document where the missing item is to be inserted, and information on contents written before and after the insertion location.
11. The target document includes a plurality of clauses; The support program according to claim 9 , wherein the target document information includes a description of a clause in the target document that is related to a missing item.
12. the subject document is a legal document including a preamble; The support program according to claim 9 , wherein the target document information includes a description of a preamble of the target document.
13. The support program according to claim 1 , wherein the auxiliary information includes definition information indicating definitions of words used in the target document.
14. The support program according to claim 1 , wherein the auxiliary information includes writing style information indicating a writing style in the target document.
15. The acquisition unit further acquires a correction strength selected by a user, The support program according to claim 1 , wherein the generation unit further inputs the correction strength into the sentence generation model.
16. causing the computer to further function as a specific unit; The acquisition unit further acquires an item type indicating a type of the missing item and a correction strength selected by a user, the identifying unit identifies an example of writing to be used in creating or correcting the target document from among the examples of writing corresponding to the item type in a database that stores a plurality of examples of writing corresponding to the item type, based on the correction level; The support program according to claim 2 , wherein the generation unit inputs the example descriptions and auxiliary information identified by the identification unit into the text generation model.
17. The assistance program according to claim 1 , wherein the generation unit generates a generation command instructing to generate a draft sentence for the missing item based on the item information and auxiliary information as a sentence in natural language and inputs the generated command to the sentence generation model.
18. The computer further functions as an evaluation unit; The evaluation unit acquires evaluation information indicating an evaluation result of the target document, The evaluation information includes one or more points of criticism regarding the target document, The problem is assigned a classification selected from a plurality of groups including a missing item and a correction of a clause; The acquisition unit determines the indication type of the indication included in the evaluation information, For items that are found to be insufficient, obtain item information on the insufficient items related to the items that are found to be insufficient; Regarding the matters to be amended, obtain target document information indicating the contents of the target document related to the matters to be amended; The generation unit is Regarding the item lacking, the generation command instructing the sentence generation model to generate a draft sentence regarding the lacking item based on the item information and auxiliary information is input, The assistance program according to claim 17 , further comprising: inputting, into the text generation model, the generation command for instructing the model to generate a proposed revision for a clause based on the target document information and auxiliary information, where the proposed revision is found to be necessary for the clause.
19. A support system for supporting the creation or modification of a target document, comprising: The subject document is a legal document to be created or amended; The apparatus includes an acquisition unit, a generation unit, and a proposal unit, The acquisition unit is Item information on the contents of the document corresponding to the missing items that are missing in the legal document to be created or amended; Acquire auxiliary information related to the target document for making a suggestion according to the target document; The generation unit inputs the item information and auxiliary information into a sentence generation model; The suggestion unit transmits suggestion information for assisting in the creation of the missing items in the target document based on an output of the sentence generation model.
20. A method for assisting in the creation or modification of a target document, comprising: The subject document is a legal document to be created or amended; The method is executed by a computer including an acquisition unit, a generation unit, and a suggestion unit, The acquisition unit is Item information on the contents of the document corresponding to the missing items that are missing in the legal document to be created or amended; Acquire auxiliary information related to the target document for making a suggestion according to the target document; The generation unit inputs the item information and auxiliary information into a sentence generation model; The suggestion unit transmits suggestion information for supporting the creation of the missing items in the target document based on an output of the sentence generation model.
Citation Information
Patent Citations
Document creation program and information processor
JP2019153110A
Legal document creation support system, legal document creation support method and program
JP2020035165A
Legal document data correction method, legal document data correction system and legal document data correction program
JP2020080132A
Contract check device and program for the same
JP2020166864A
Information analyzing method for performing autamatic generating of document based on artificial intelligence and apparatus therefor
KR102442350B1
Cited By
Information processing apparatus, information processing method, program, and system
JP2026081108A
Computer systems for assisting in the creation of structured documents, methods and programs for which such computer systems are executed.
JP2026109291A
Information processing device, information processing method, and program
JP2026165446A
Document analysis device, document analysis method, and document analysis program
JP7897686B1
Computer system for assisting in creation of structured document, method executed in said computer system, and program
WO2026133831A1
