Information processing device, program and information processing method

Abstract

To make it more efficient to acquire contract information.SOLUTION: An information processing device 2 according to one aspect of the present disclosure includes: a contract information acquisition unit 100a that acquires first contract information (estimation target contract information) regarding a first contract (estimation target contract) that indicates at least contract contents of a first party (estimation target party); and a first estimation unit (position estimation unit 104a) that estimates the position of the first party (estimation target party) in the first contract (estimation target contract) based on a trained model (position estimation model) that uses the first contract information (estimation target contract information) as input and uses the position of the first party (estimation target party) in the first contract (estimation target contract) as output.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to an information processing device, a program, and an information processing method. [Background technology]

[0002] There are currently technologies for checking the contents of contracts. For example, Patent Document 1 discloses an electronic contract creation support system that digitizes contract documents, thereby eliminating the need to move paper media, eliminating the hassle of physically moving contract documents and the risk of losing them, and also enabling multiple contracting parties to simultaneously confirm the contract contents, thereby improving the efficiency of the contract content confirmation process and shortening the time required to conclude a contract. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2023-88560 Summary of the Invention [Problem to be solved by the invention]

[0004] The present disclosure provides an information processing device, a program, and an information processing method that improve the efficiency of acquiring contract information. [Means for solving the problem]

[0005] An information processing device according to one embodiment of the present disclosure includes a contract information acquisition unit that acquires first contract information regarding a first contract that indicates at least the contract contents of the first party, and a first estimation unit that estimates the position of the first party in the first contract based on a trained model that takes the first contract information as input and outputs the position of the first party in the first contract.

[0006] A program according to another aspect of the present disclosure causes a computer to function as an output means that outputs the position of the first party in the first contract, estimated based on first contract information regarding the first contract indicating at least the contract contents of the first party, and a trained model that takes the first contract information as input and outputs the position of the first party in the first contract.

[0007] An information processing method according to another aspect of the present disclosure causes a computer to acquire first contract information relating to a first contract indicating at least the contract content of the first party, and estimate the position of the first party in the first contract based on the first contract information and a trained model that takes the first contract information as input and outputs the position of the first party in the first contract. [Effects of the Invention]

[0008] According to the present disclosure, it is possible to efficiently obtain contract information. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a schematic diagram of the system 1. [Figure 2] FIG. 1 is a functional block diagram of the system 1. [Figure 3] 4 is a flowchart showing the operation of the information processing device 2 according to the first embodiment. [Figure 4] 10 is a flowchart showing the operation of the information processing device 2 related to the generation of a position estimation model. [Figure 5] FIG. 10 is a diagram for explaining an example of learning target contract information. [Figure 6] FIG. 1 is a conceptual diagram of a position estimation model. [Figure 7] 10 is a flowchart showing the operation of the information processing device 2 relating to estimation of the position of the party to be estimated. [Figure 8] FIG. 10 is a diagram illustrating an example of estimated contract information. [Figure 9] FIG. 1 is a conceptual diagram of a position estimation model. [Figure 10]10 is a flowchart showing the operation of the information processing device 2 according to the second embodiment. [Figure 11] 10 is a flowchart showing the operation of the information processing device 2 related to the generation of a relationship estimation model. [Figure 12] 10 is a flowchart showing the operation of the information processing device 2 relating to estimation of the position of the party to be estimated. [Figure 13] FIG. 2 is a diagram illustrating an example of hardware of an information processing device 2. DETAILED DESCRIPTION OF THE INVENTION

[0010] <1. Overview> Generally, contracts are prepared to show that multiple parties have agreed to the specific contractual content from their respective perspectives. In order for a contract reviewer to understand the positions of at least some of the parties, the reviewer must understand the contractual content themselves. For example, if a contract is about a confidentiality agreement between two parties, "XX Co., Ltd." and "XX Co., Ltd.", the reviewer must obtain information about the contract, such as through a contract review request, or read and understand the contractual content, in order to determine whether "XX Co., Ltd." is the discloser or the recipient. However, such work often places a burden on the reviewer.

[0011] According to the present disclosure, it is possible to solve such problems. Figure 1 is a conceptual diagram of the operation of a system 1 according to the present disclosure.

[0012] The system 1 according to the present disclosure receives as input (S1) the name of a party (hereinafter referred to as the "presumed party") whose position is to be presumed, among multiple parties in a contract to be presumed (hereinafter referred to as the "presumed contract"), and the text data of the presumed contract. In the example of Figure 1, the name of the presumed party is "XX Co., Ltd.", and the text data of the presumed contract includes the statement "XX Co., Ltd. (hereinafter referred to as "Party A")...Party A will disclose information to Party B."

[0013] Next, the system 1 performs predetermined preprocessing and feature extraction on these inputs. The preprocessing includes, for example, identifying that the abbreviation for "XX Co., Ltd." in the presumption target contract is "A" based on the description "XX Co., Ltd." (hereinafter referred to as "A") included in the text data of the presumption target contract. The feature extraction includes, for example, extracting a group of words having a predicate-argument structure with "A" as the argument from the text data of the presumption target contract. Hereinafter, the results obtained by the preprocessing and feature extraction will be referred to as "presumption target contract information." That is, in one embodiment, the presumption target contract information includes features indicating "what and how A (i.e., XX Co., Ltd.") did in the presumption target contract."

[0014] For example, in Figure 1, the text data of the presumed target contract contains the statement, "Party A will disclose information to Party B." Therefore, the presumed target contract information will contain the feature, "Party A will disclose information to Party B."

[0015] Next, the system 1 inputs the contract information to be estimated into the position estimation model. The position estimation model learns one or more pieces of training data in which a label indicating the position of a party indicated by the abbreviation "Party A" in the contract to be trained is associated with the contract information to be trained, which is obtained as a result of predetermined preprocessing and feature extraction for the contract to be trained (hereinafter referred to as the "contract to be trained"). That is, in one embodiment, the contract information to be trained includes features indicating "what Party A did and how" in the contract to be trained. Furthermore, the position estimation model learns the relationship between "what Party A did and how" and the position of "Party A" in the contract to be trained.

[0016] For example, in Figure 1, the position estimation model learns multiple pieces of training data including first training data and second training data. In the first training data, the label "Party A is the discloser" is associated with the contract information to be learned that includes the feature "Party A discloses information to Party B." In the second training data, the label "Party A is the recipient" is associated with the contract information to be learned that includes the feature "Party A receives information disclosure from Party B."

[0017] The contract to be studied does not have to indicate the contractual content of "XX Co., Ltd." In other words, in the contract to be studied, "Party A" is merely used as an abbreviation for a specified party, and does not have to indicate "XX Co., Ltd."

[0018] Next, the system 1 outputs the result of inputting the inferred target contract information into the position inference model (S2). As described above, in the example of FIG. 1, the inferred target contract information includes the feature "Party A discloses information to Party B." With reference to the first learning data and the second learning data, the inferred target contract information is closer to the training target contract information associated with the label "Party A is the discloser" compared to the training target contract information associated with the label "Party A is the recipient." Therefore, the position inference model outputs that "Party A" (i.e., "XX Co., Ltd.") is the discloser in the inferred target contract.

[0019] According to the present disclosure, it is possible to estimate the positions of at least some of the parties in a contract to be estimated. This reduces the time and effort required for a reviewer to understand the content of the contract to be estimated. As a result, it is possible to efficiently obtain information about the contract.

[0020] In the example of Figure 1, the type of contract to be estimated (for example, whether it is a confidentiality agreement or a service contract) is not particularly accepted as an input, but the type of contract to be estimated may also be accepted as an input. This may enable a more appropriate estimation of the position.

[0021] The detailed configuration and operation of the system 1 will be described below.

[0022] <2. Functional configuration> The functional configuration of the system 1 of this embodiment will be described with reference to Fig. 2. The system 1 includes an information processing device 2, a terminal device 3, and a communication network 5. The information processing device 2 and the terminal device 3 are configured to be able to communicate with each other via the communication network 5.

[0023] [Information processing device 2] The information processing device 2 is a server device in the case where the terminal device 3 is a client terminal device. In one embodiment, the information processing device 2 is a web server device.

[0024] The information processing device 2 includes a control unit 10, a storage unit 12, a network interface unit 14, and a bus 16. The control unit 10, the storage unit 12, and the network interface unit 14 are electrically connected via the bus 16.

[0025] (Control unit 10) The control unit 10 functions as an acquisition unit 100, a learning unit 102, an estimation unit 104, an identification unit 106, and a transmission unit 108 by executing various programs stored in the storage unit 12, which will be described later.

[0026] -Acquisition part 100- The acquisition unit 100 includes a contract information acquisition unit 100a and a type information acquisition unit 100b.

[0027] ~Contract information acquisition unit 100a~ The contract information acquisition unit 100a acquires presumed contract information relating to the presumed contract indicating the contract details of the presumed parties. In one embodiment, the contract information acquisition unit 100a acquires the presumed contract information by the following steps (1) to (3).

[0028] (1) Acquisition of text data of the contract to be estimated The contract information acquisition unit 100a acquires text data of the contract to be estimated. The contract information acquisition unit 100a acquires, for example, text data of the contract to be estimated uploaded from the terminal device 3, or text data of the contract to be estimated stored in another cloud server device. The text data of the contract to be estimated may be any of a value stored in a string type variable, a TXT file, an XML file, an HTML file, a PDF file, etc.

[0029] (2) Preprocessing the acquired text data The contract information acquisition unit 100a performs predetermined preprocessing on the text data of the contract to be estimated acquired in (1). In one embodiment, the preprocessing includes normalizing character codes, dividing text into sentence units, standardizing abbreviations, replacing parallel abbreviations, and analyzing dependency structures.

[0030] The unification of abbreviations includes a process of unifying abbreviations that can be used in a contract to predetermined abbreviations. For example, when abbreviations other than "Party A" and "Party B" are included in the text data of the contract to be estimated, the unification of abbreviations includes changing them to "Party A" or "Party B."

[0031] The replacement of parallel abbreviations includes changing a part of the text data of the presumed contract in which the abbreviations of the presumed party and the abbreviations of the opposing party are written in parallel in a predetermined order to a description that does not depend on the order of these abbreviations. The replacement of parallel abbreviations includes, for example, changing a part of the text data of the presumed contract that is written as "Party A and Party B" or "Party A or Party B" to "both parties."

[0032] (3) Extraction of features based on the results of preprocessing The contract information acquisition unit 100a extracts features from the information obtained as a result of the preprocessing in (2). In one embodiment, the extraction of features includes extraction of predicate-argument structures, extraction of word groups including the inferred target parties, anonymization of abbreviations, etc.

[0033] Extracting the predicate-argument structure involves extracting a group of words having a predicate-argument structure with the abbreviated names of the parties as arguments from information obtained by analyzing the dependency structure of the text data of the contract to be estimated. For example, if the abbreviated names of the presumed party and the opposing party are "A" and "B," respectively, extracting the predicate-argument structure involves extracting a group of words indicating "what A did and how" and a group of words indicating "what B did and how." The word groups can also be called N-grams.

[0034] Extracting a word group including the presumed party includes extracting a word group including an abbreviation indicating the presumed party from the word group obtained by extracting the predicate-argument structure. If the presumed party is "XX Co., Ltd." and its abbreviation is "A," extracting a word group including the presumed party includes, for example, extracting a word group indicating "A did what and how" from the word group indicating "A did what and how" and the word group indicating "B did what and how."

[0035] The anonymization of abbreviations involves replacing the abbreviations of the parties with a character string that does not have a specific meaning. The character string in this case may also be referred to as a token. In one embodiment, the anonymization of abbreviations involves anonymizing the abbreviations of one party to the presumption target agreement using a first method and anonymizing the abbreviations of the other party using a second method. For example, if a group of words indicating "Party A did what and how" and a group of words indicating "Party B did what and how" are extracted by extracting the predicate-argument structure described above, the anonymization of abbreviations involves replacing the characters "Party A" and "Party B" in these groups of words with "TARGET_ALIAS" and "ANOTHER_ALIAS," respectively.

[0036] Furthermore, the contract information acquisition unit 100a may further acquire learning target contract information related to the learning target contract in a manner similar to the manner in which it acquires the estimation target contract information. That is, in one embodiment, the contract information acquisition unit 100a may acquire the learning target contract information by (1) acquiring text data of the learning target contract, (2) performing preprocessing on the acquired text data, and (3) extracting features based on the results of the preprocessing.

[0037] ~Type information acquisition unit 100b~ The type information acquisition unit 100b acquires type information regarding the type of contract to be estimated. Examples of types of contracts include non-disclosure agreements (NDAs), contracts for the provision of services during a transition period in business transfers, etc. (TSA), transaction contracts, outsourcing contracts, software development outsourcing contracts, OEM (Original Equipment Manufacturing) contracts, staffing contracts, license agreements, agency contracts, M&A contracts, capital alliance contracts, consulting contracts, joint development contracts, joint application contracts, and joint venture contracts.

[0038] Furthermore, the type information acquiring unit 100b may further acquire type information relating to the type of the contract to be studied.

[0039] In this disclosure, "obtaining information" includes making the information processable in the control unit 10. "Obtaining information" may mean receiving the information from another device, reading the information from the storage unit 12, or obtaining the information as a result of a predetermined process.

[0040] -Study Section 102- The learning unit 102 includes a position estimation model learning unit 102a and a relationship estimation model learning unit 102b.

[0041] ~ Position estimation model learning unit 102a ~ The position estimation model training unit 102a trains the position estimation model with one or more pieces of training data in which a label indicating the position of the first other party in the training contract is associated with training contract information related to the training contract that indicates the contract content of the first other party. For example, as described with reference to Figure 1, the position estimation model training unit 102a trains the position estimation model with training data in which a label "Party A is the discloser" is associated with training contract information that "Party A discloses information to Party B."

[0042] In one embodiment, the position estimation model training unit 102a further trains the position estimation model with one or more pieces of training data in which a label indicating the position of a second other party in the contract to be trained is associated with the contract information to be trained. That is, the position estimation model training unit 102a may learn not only the position of one party in the contract to be trained, but also the position of the other party. The position estimation model training unit 102a may train the position estimation model with such training data using either of the following methods (1) and (2).

[0043] (1) A method of associating two labels with one contract to be learned. The position estimation model training unit 102a may train training data in which both a label indicating the position of a first other party and a label indicating the position of a second other party are associated with training target contract information based on a specific training target contract. For example, if the specific training target contract is related to a confidentiality agreement between "Party A" who is the discloser and "Party B" who is the recipient, the position estimation model training unit 102a may train the position estimation model with training data in which both the label "Party A is the discloser" and the label "Party B is the recipient" are associated with training target contract information based on the specific training target contract.

[0044] (2) A method of generating two pieces of learning contract information based on one learning contract, and associating one label with each piece of learning contract information and training them. The position estimation model training unit 102a trains first training data in which a label indicating the position of a first other party is associated with training target contract information that has been preprocessed and / or feature extracted from a specific training target contract using a first method, and second training data in which a label indicating the position of a second other party is associated with training target contract information that has been preprocessed and / or feature extracted from the specific training target contract using a second method. For example, if the specific training target contract is related to a confidentiality agreement between "Party A" who is the discloser and "Party B" who is the recipient, the position estimation model training unit 102a may train the position estimation model using the following training data (A) and (B). (A) First learning data in which the label "A is the discloser" is associated with learning target contract information (i.e., information indicating "what A did and how") that includes a group of words with a predicate-argument structure with "A" as the argument, extracted from a specific learning target contract. (a) Second learning data in which the label "Party B is a recipient" is associated with the learning target contract information (i.e., information indicating "Party B did what and how") that includes a group of words with a predicate-argument structure with "Party B" as an argument, extracted from the specific learning target contract.

[0045] In one embodiment, the position estimation model training unit 102a may train the position estimation model using training data in which the type of the contract to be learned is further associated with the contract information to be learned. In other words, the position estimation model training unit 102a may generate a single position estimation model that can handle multiple contract types.

[0046] In another embodiment, the position estimation model learning unit 102a may generate a different position estimation model for each type of contract. The position estimation model learning unit 102a may generate, for example, the following different position estimation models (1) to (3). (1) A position estimation model for non-disclosure agreements, which is trained by associating training data with labels indicating the parties' positions in the training agreements (e.g., discloser or recipient) for training agreement information based on training agreements related to non-disclosure agreements. (2) A position estimation model for outsourcing contracts, which is trained by learning data in which labels indicating the parties' positions in the contract (e.g., contractor or recipient) are associated with the training target contract information based on the training target contract. (3) A position estimation model for license agreements, which is trained by associating learning data with labels indicating the parties' positions in the learning agreements (e.g., licensor or licensee) for learning agreement information based on the learning agreements.

[0047] ~Relationship estimation model learning unit 102b~ Depending on the content and type of contract, the multiple parties in the contract may each have a different position, or all of the multiple parties may have the same position. For example, in a confidentiality agreement, one party may be the discloser and the other the recipient (i.e., the multiple parties have different positions), or both parties may be both the discloser and the recipient (i.e., the multiple parties have the same position, which can also be called "bilateral disclosure").

[0048] The relationship estimation model training unit 102b generates a relationship estimation model for estimating whether the positions of multiple parties in a contract are the same or different. Specifically, the relationship estimation model training unit 102b generates a relationship estimation model by training one or more pieces of training data in which labels indicating whether the multiple parties are in the same position or different positions are associated with training target contract information related to a training target contract that indicates the contract content between the multiple parties. For example, the relationship estimation model training unit 102b trains the relationship estimation model with training data in which labels indicating either (1) or (2) below are associated with training target contract information based on a training target contract related to a confidentiality agreement. (1) One of the parties in the Study Agreement is the Discloser and the other is the Recipient. (2) The contract to be studied involves bilateral disclosure.

[0049] Hereinafter, when multiple parties in a contract have different positions, these positions will be referred to as "contrasting positions," and the contract will be referred to as a "contract with contrasting positions." Also, below, when multiple parties in a contract have the same position, the position will be referred to as a "neutral position," and the contract will be referred to as a "neutral position contract." The relationship inference model can also be said to be a model for inferring whether the contract to be inferred is a contract with contrasting positions or a contract with a neutral position.

[0050] The relationship estimation model learning unit 102b may further use the learning target contract used when generating the position estimation model to generate the relationship estimation model, or may use a different learning target contract.

[0051] The position estimation model and the relationship estimation model are typically classification models based on supervised learning. The classification model may be a learning model based on a support vector machine (SVM), a decision tree, a k-nearest neighbor method, or a logistic regression, or a learning model including a neural network. The SVM may be either a linear SVM or a nonlinear SVM. The neural network may be any of a deep neural network (DNN), a convolutional neural network (CNN), a recurrent neural network (RNN), a generative adversarial network (GAN), etc.

[0052] In this disclosure, "generating a model" includes making the model usable by the control unit 10. Making the model usable by the control unit 10 includes making the model usable so that a predetermined output is obtained as a result of providing a predetermined input to the model.

[0053] ―Estimation part 104― The estimation unit 104 includes a position estimation unit 104a and a relationship estimation unit 104b.

[0054] ~Position estimation section 104a~ The position estimation unit 104a estimates the position of the party to be estimated in the contract to be estimated based on the contract information to be estimated and the position estimation model generated by the position estimation model learning unit 102a. In one embodiment, the position estimation unit 104a estimates the position of the party to be estimated by inputting the contract information to be estimated into the position estimation model. Note that if the contract to be estimated relates to a contract between multiple parties, the party to be estimated may be someone determined from among the multiple parties based on an operation by the user of the terminal device 3.

[0055] In one embodiment, the position estimation unit 104a estimates the position of the party to be estimated further based on type information. Specifically, if the position estimation model training unit 102a has trained the position estimation model with learning data in which the type of the contract to be learned is further associated with the contract information to be learned, the position estimation unit 104a may input information in which the type of the contract to be estimated is associated with the contract information to be estimated, into the position estimation model. Also, if the position estimation model training unit 102a has generated different position estimation models for each contract type, the position estimation unit 104a may first select a position estimation model based on type information, and then input the contract information to be estimated to the selected position estimation model.

[0056] ~Relationship estimation unit 104b~ The relationship estimation unit 104b estimates whether the presumed party in the presumed contract is in a different position from the other party in the presumed contract, or in the same position, based on the presumed contract information and the relationship estimation model generated by the relationship estimation model learning unit 102b. This includes estimating whether the presumed contract is about a neutral position. This is because if the presumed contract is about a neutral position, it can be determined that the presumed party is in the same position as the other party, and if the presumed contract is about a contrasting position, it can be determined that the presumed party is in a different position from the other party.

[0057] -Specific section 106- The identification unit 106 identifies the abbreviation of the presumed party based on the description in the presumed contract surrounding the portion where the identification information of the presumed party (for example, name, etc.) is described. In one embodiment, the identification unit 106 identifies the abbreviation of the presumed party based on a predetermined rule from the portion surrounding the portion where the identification information of the presumed party is described. For example, as explained with reference to FIG. 1, if the presumed contract contains the description "XX Co., Ltd. (hereinafter referred to as "Party A") is...", the identification unit 106 can identify "Party A", which is the abbreviation of "XX Co., Ltd.", based on the rule that "the characters in the parentheses included in the parentheses following 'XX Co., Ltd.' are identified as the abbreviation."

[0058] -Transmitting unit 108- The transmitting unit 108 transmits the estimation result by the estimating unit 104 to the terminal device 3. In one embodiment, if the estimated contract is a contract with opposing positions, the transmitting unit 108 transmits to the terminal device 3 an estimation result indicating which of multiple opposing positions the estimated party has. In one embodiment, if the estimated contract is a contract with a neutral position, the transmitting unit 108 transmits to the terminal device 3 an estimation result indicating that the estimated party has a neutral position.

[0059] (Storage unit 12) The storage unit 12 stores various pieces of information that need to be stored for the operation of the information processing device 2. The storage unit 12 stores various programs executed by the control unit 10. In one embodiment, the storage unit 12 at least temporarily stores the position estimation model and the relationship estimation model generated by the learning unit 102.

[0060] The storage unit 12 stores information regarding the positions of the parties in each type of contract. For example, for a confidentiality agreement, the storage unit 12 stores that the positions of the parties are "discloser" or "recipient." For example, for a license agreement, the storage unit 12 stores that the positions of the parties are "licensor" or "licensee."

[0061] Furthermore, the memory unit 12 stores, for each type of contract, whether or not a neutral position is possible within that type. For example, with respect to a confidentiality agreement, the memory unit 12 stores information indicating that one party may be the discloser and the other party may be the recipient (i.e., opposing positions), and both parties may be the discloser and the recipient (i.e., a case where both parties disclose and are neutral). For example, with respect to a service outsourcing contract, the memory unit 12 stores information indicating that one party may be the client and the other party may be the recipient (i.e., opposing positions), and that a neutral position is not possible. Note that, in this disclosure, "a neutral position is not possible within a contract type" does not necessarily mean that a neutral position cannot be adopted within that type, but also includes the fact that a neutral position is rarely adopted within that type due to custom or social convention.

[0062] (Network Interface Unit 14) The network interface unit 14 realizes communication with other devices or systems via the communication network 5 .

[0063] [Terminal device 3] The terminal device 3 is a device used by a user to use the system 1. The terminal device 3 is configured to be able to communicate with the information processing device 2 via a communication network 5. The terminal device 3 is, for example, a personal computer, a smartphone, a tablet terminal, etc. The terminal device 3 is equipped with an input device (for example, a mouse, a keyboard, a touch panel, a camera, a microphone, etc.) and an output device (for example, a display, a speaker, etc.). In one embodiment, the terminal device 3 communicates with the information processing device 2 by executing an application such as a web browser. Note that the user may or may not be a party to the presumed contract.

[0064] [Communication Network 5] The communication network 5 realizes communication between the information processing device 2 and the terminal device 3. The communication network 5 realizes communication based on, for example, the TCP / IP protocol.

[0065] <3.Operation> [First embodiment] 3 to 9, the operation of the information processing device 2 according to the first embodiment will be described. In the first embodiment, the contract is described as being about contrasting positions.

[0066] 3 is a flowchart showing the operation of the information processing device 2 according to the first embodiment. The information processing device 2 generates a position estimation model (S10), and estimates the position of the party to be estimated in the contract to be estimated based on the position estimation model (S20). S10 will be described below with reference to FIGS. 4-6, and S20 will be described with reference to FIGS. 7-9.

[0067] (Generation of position estimation model (S10)) Fig. 4 is a flowchart showing an example of an operation of generating a position estimation model by the information processing device 2. Through the series of processes shown in Fig. 4, the information processing device 2 causes the position estimation model to learn the relationship between the contents of descriptions about the parties in the contract to be learned and the positions of the parties.

[0068] The information processing device 2 first acquires text data of the contract to be studied, and also acquires the type of the contract to be studied (S100). In this example, the information processing device 2 acquires text data of a contract to be studied that relates to a confidentiality agreement in which "Party A" is the discloser and "Party C" is the recipient. Figure 5(a) shows an example of the text data of the contract to be studied acquired by the information processing device 2.

[0069] 4, the information processing device 2 performs preprocessing on the text data of the contract to be learned (S102). In this example, the preprocessing includes normalizing the character code, dividing the contract into sentences, replacing abbreviations with "Party A" and "Party B," and replacing abbreviations written in parallel with "both parties."

[0070] Figure 5(b) shows an example of information after preprocessing. In this example, the text data of the contract to be learned is divided into four sentences, (1) to (4) in Figure 5(b). Furthermore, the parts written as "Party C" and "Party A and Party B" in the text data of the contract to be learned have been replaced with "Party B" and "Both Parties" in Figure 5(b), respectively.

[0071] 4, the information processing device 2 analyzes the dependency structure of each sentence based on the information after preprocessing and extracts a predicate-argument structure with the abbreviations of the parties as arguments (S104). After that, it executes in parallel a process of extracting a group of words containing "Party A" from the result of S104 (S106a) and a process of extracting a group of words containing "Party B" from the result of S104 (S106b).

[0072] Figures 5(c1)-(c2) respectively show an example of a group of words containing "Party A" extracted in S106a and a group of words containing "Party B" extracted in S106b. As shown in Figure 5(c1), in S106a, the group of words "Party A discloses confidential information to Party B" is extracted. Also, as shown in Figure 5(c2), in S106b, the group of words "Party A discloses confidential information to Party B" and the group of words "Party B keeps the confidential information confidential" are extracted.

[0073] Returning to Fig. 4, the information processing device 2 performs abbreviation anonymization on the extracted word group. Specifically, the information processing device 2 performs a process (S108a) for replacing "Party A" and "Party B" with the character strings "TARGET_ALIAS" and "ANOTHER_ALIAS", respectively, on the result of S106a, and a process (S108b) for replacing "Party B" and "Party A" with the character strings "TARGET_ALIAS" and "ANOTHER_ALIAS", respectively, on the result of S106b. That is, in S108a, "Party A" is replaced with "TARGET_ALIAS", and in S108b, "Party B" is replaced with "TARGET_ALIAS". Hereinafter, the result of S108a will be referred to as learning contract information D1, and the result of S108b will be referred to as learning contract information D2. 5(d1)-(d2) show examples of the learning target contract information D1 and the learning target contract information D2, respectively.

[0074] 4, the information processing device 2 associates the label "TARGET_ALIAS is the discloser" with the learning target contract information D1 (S110a), and associates the label "TARGET_ALIAS is the recipient" with the learning target contract information D2 (S110b). Hereinafter, the result of S110a will be referred to as learning data E1, and the result of S110b will be referred to as learning data E2.

[0075] Next, the information processing device 2 trains the training data E1 and the training data E2 into a position estimation model corresponding to the type of the contract to be trained (S112). In this example, since information that the type of the contract to be trained is a non-disclosure agreement is acquired in S100, the information processing device 2 trains the training data E1 and the training data E2 into a position estimation model related to a non-disclosure agreement.

[0076] The information processing device 2 generates a position estimation model by executing the processes of S110-S112 for each of the multiple learning target contracts.

[0077] Figure 6 shows a conceptual diagram of the generated position inference model. The graph in Figure 6 is a two-dimensional plot of plots corresponding to the contract information to be learned, focusing on feature v1 and feature v2 out of the multiple feature quantities contained in the contract information to be learned. In addition, a discrimination boundary separates multiple plots corresponding to the contract information to be learned that is associated with the label "TARGET_ALIAS is the discloser" from multiple plots corresponding to the contract information to be learned that is associated with the label "TARGET_ALIAS is the recipient." Determining such a discrimination boundary is an example of generating a position inference model.

[0078] (Contrasting Position Estimation (S20)) FIG. 7 is a flowchart showing an example of the operation of the information processing device 2 when estimating the position of the party to be estimated based on the position estimation model generated in S10.

[0079] The information processing device 2 first acquires text data of the contract to be estimated, and also acquires the type of the contract to be estimated (S200). In this example, the contract to be estimated is assumed to be related to a confidentiality agreement between "XX Co., Ltd." and "XX Co., Ltd." Figure 8(a) shows an example of the text data of the contract to be estimated acquired by the information processing device 2.

[0080] 7, next, the information processing device 2 acquires the name of the presumed party (S202). In this example, the information processing device 2 acquires "XX Co., Ltd." as the name of the presumed party based on an input to the terminal device 3 by the user.

[0081] Next, the information processing device 2 performs preprocessing on the text data of the contract to be estimated (S204). In this example, the preprocessing includes, similar to S102, normalizing the character code, dividing the contract into sentences, replacing abbreviations with "Party A" and "Party B", and replacing abbreviations written in parallel with "both parties".

[0082] An example of information after preprocessing is shown in Figure 8(b). In this example, the text data of the contract to be estimated is divided into three sentences (1) to (3) in Figure 8(b).

[0083] 7, next, the information processing device 2 identifies the abbreviation of the presumed party based on the text data of the presumed contract and the name of the presumed party acquired in S202 (S206). In this example, based on the description "XX Co., Ltd. (hereinafter referred to as "Party A")" included in the text data of the presumed contract, it identifies the abbreviation of "XX Co., Ltd." as "Party A."

[0084] Next, the information processing device 2 analyzes the dependency structure of each sentence based on the information after preprocessing, and extracts a predicate-argument structure with the abbreviations of the parties as arguments (S208). Thereafter, the information processing device 2 extracts a group of words containing "A" from the result of S208 (S210), and replaces "A" and "B" with the character strings "TARGET_ALIAS" and "ANOTHER_ALIAS", respectively (S212). Hereinafter, the result of S212 will be referred to as estimated target contract information F1. Figures 8(c) and 8(d) show an example of the result of S210 and estimated target contract information F1, respectively.

[0085] 7, next, the information processing device 2 inputs the estimated contract information F1 into a position estimation model corresponding to the type of the estimated contract (S214). In this example, since information that the type of the estimated contract is a confidentiality agreement is acquired in S200, the information processing device 2 inputs the estimated contract information F1 into the position estimation model related to the confidentiality agreement generated in S10.

[0086] Figure 8 shows a conceptual diagram of inputting estimated contract information F1 into the position estimation model. The graph in Figure 8 is the graph in Figure 6 with a plot corresponding to estimated contract information F1 superimposed on it. In this example, the plot corresponding to estimated contract information F1 is on the "discloser" side of the identification boundary. This indicates that estimated contract information F1 tends to be closer to learning contract information associated with the label "TARGET_ALIAS is discloser" compared to learning contract information associated with the label "TARGET_ALIAS is recipient".

[0087] To summarize the above, in S200-S214, the information processing device 2 determines that "Party A," the abbreviation for "XX Co., Ltd." in the presumed contract, was listed as the discloser, not the recipient, in the presumed contract. This allows the information processing device 2 to presume that the presumed party, "XX Co., Ltd.", is in the position of discloser in the presumed contract.

[0088] [Second embodiment] 10-12, the operation of the information processing device 2 according to the second embodiment will be described. In the above-described first embodiment, the contract is described as being related to contrasting positions, but in the second embodiment, the case where the contract is related to contrasting positions or a neutral position will be described.

[0089] 10 is a flowchart showing the operation of the information processing device 2 according to the second embodiment. The information processing device 2 first generates a position estimation model by the same method as that described with reference to FIGS. 4-6 of the first embodiment (S10).

[0090] (Generation of relationship estimation model (S30)) Next, the information processing device 2 generates a relationship inference model (S30). Fig. 11 is a flowchart showing an example of the operation of the information processing device 2 when generating a relationship inference model. Note that S300-S304 may be the same as the method described in S100-S104 of the first embodiment, and therefore description thereof will be omitted below. Hereinafter, the result of S304 will be referred to as learning target contract information D3.

[0091] Next, the information processing device 2 associates a label indicating whether the multiple parties in the contract to be learned are in different positions or in the same position with the contract information D3 to be learned (S306). If the contract to be learned is, for example, related to a confidentiality agreement, the information processing device 2 associates a label saying "not mutually disclosed" or a label saying "mutually disclosed" with the contract information D3 to be learned. Hereinafter, the result of S306 will be referred to as learning data E3.

[0092] Next, the information processing device 2 trains the learning data E3 into a relationship inference model corresponding to the type of the learning target contract (S308). The information processing device 2 generates a relationship inference model by executing the processes of S300-S308 for each of the multiple learning target contracts.

[0093] 10, the information processing device 2 first acquires text data of the contract to be estimated and also acquires the type of the contract to be estimated (S32). Next, the information processing device 2 determines whether the type of the contract to be estimated is a type that can have a neutral position (S34).

[0094] If the type of the contract to be estimated is one in which a neutral position is not possible (S34 NO), the information processing device 2 estimates the position of the party to be estimated in the contract to be estimated based on the position estimation model generated in S10, using a method similar to the method described with reference to Figures 7-9 of the first embodiment (S20).

[0095] (Presumption of neutral position (S40)) On the other hand, if the type of the estimated contract is one in which a neutral position is possible (S34 YES), the information processing device 2 estimates whether the estimated contract is related to a neutral position based on the text data of the estimated contract obtained in S32 (S40).

[0096] 12 is a flowchart showing an example of the operation of the information processing device 2 when it infers whether the estimation target contract is related to a neutral position based on the relationship inference model generated in S30. Specifically, the information processing device 2 executes preprocessing on the text data of the estimation target contract (S400) and extracts a predicate-argument structure with the abbreviations of the parties as arguments (S402). Thereafter, the information processing device 2 inputs the estimation target contract information F3, which is the result of S402, into the relationship inference model (S404).

[0097] Returning to FIG. 10, the information processing device 2 determines whether or not the estimation target contract is estimated to be related to a neutral position in S40 (S42).

[0098] If it is estimated that the contract to be estimated does not relate to a neutral position (S42 NO), the information processing device 2 estimates the position of the estimated party in the contract to be estimated based on the position estimation model generated in S10 using a method similar to the method described with reference to Figures 7-9 of the first embodiment (S20).

[0099] On the other hand, if the presumption target contract is presumed to be related to a neutral position (YES in S42), the information processing device 2 presumes that the presumed target party is in a neutral position, and ends the process.

[0100] <4. Effects> An information processing device 2 according to one embodiment of the present disclosure includes a contract information acquisition unit 100a that acquires contract information relating to an estimated contract that indicates at least the contract content of the estimated party, and a position estimation unit 104a that estimates the position of the estimated party in the estimated contract based on the estimated contract information and a position estimation model that has learned one or more learning data in which a label indicating the position of the first other party in the training contract is associated with the training contract information relating to a training contract that indicates the contract content of at least a first other party.

[0101] As described in <1. Overview>, the information processing device 2 can efficiently acquire contract information.

[0102] In one embodiment, the presumed target party is indicated by a first abbreviated name ("Party A") in the presumed target contract, the first other party is indicated by a first abbreviated name ("Party A") in the studied contract, the presumed target contract information is determined based on information relating to the structure of a sentence containing the first abbreviated name ("Party A") in the presumed target contract, and the studied contract information is determined based on information relating to the structure of a sentence containing the first abbreviated name ("Party A") in the studied contract. Also, in one embodiment, the presumed target contract information is determined based on information relating to the predicate-argument structure in the presumed target contract that has the first abbreviated name ("Party A") as an argument, and the studied contract information is determined based on information relating to the predicate-argument structure in the studied contract that has the first abbreviated name ("Party A") as an argument.

[0103] When estimating a position, the description in the presumed contract of "what the presumed party did and how" is more important than the literal name of the presumed party. However, for example, if the name of the presumed party in the presumed contract is "XX Co., Ltd." and the name of the party in the study contract is not "XX Co., Ltd.", the difference between the presumed contract information and the study contract information may be large due to factors unrelated to the position estimation. With these configurations, the position of the presumed party is estimated based on the structure and / or meaning of the sentence centered on the first abbreviation in each of the presumed contract and the study contract, thereby improving the accuracy of the position estimation.

[0104] In one embodiment, the presumed target contract information and the learning target contract information are determined based on information obtained by anonymizing a first abbreviated name ("Party A") using a first method. Also, in one embodiment, the presumed target contract relates to a contract between the presumed target party and the other party, and the learning target contract relates to a contract between a first other party and a second other party, the other party is indicated by a second abbreviated name ("Party B") in the presumed target contract, and the second other party is indicated by a second abbreviated name ("Party B") in the learning target contract, and the presumed target contract information and the learning target contract information are determined based on information obtained by anonymizing the second abbreviated name ("Party B") using a second method.

[0105] The abbreviations of the parties themselves do not inherently have any specific meaning. However, by convention, the abbreviations of certain parties are frequently used in certain contexts or to represent certain positions. For example, in non-disclosure agreements, the abbreviation "Party A" is frequently used to represent the position of the discloser. This configuration can eliminate bias that may be imparted to the abbreviations of the parties due to such conventions. This improves the accuracy of position estimation.

[0106] In one embodiment, the presumed target contract relates to a contract between the presumed target party and an opposing party, and the studied contract relates to a contract between a first other party and a second other party, the opposing party is indicated by a second abbreviation ("Party B") in the presumed target contract, and the second other party is indicated by a second abbreviation ("Party B") in the studied contract, and the presumed target contract information and the studied contract information are determined based on information in which the part in which the first abbreviation ("Party A") and the second abbreviation ("Party B") are written in parallel in a predetermined order has been changed to a description in which the first abbreviation ("Party A") and the second abbreviation ("Party B") are not written in any order.

[0107] In contracts, for example, expressions such as "Party A and Party B" or "Party A or Party B" are sometimes used. In parts where multiple abbreviations are written in parallel like this, the order of the abbreviations themselves is meaningless. For example, "Party A and Party B" and "Party B and Party A" have the same meaning. However, customarily, there is a certain tendency for the order in which abbreviations are written in contracts, and by learning from such contracts, a position estimation model may discover meaning in the order of abbreviations that does not actually exist. This configuration can eliminate bias due to the order of abbreviations, thereby improving the accuracy of position estimation.

[0108] In one embodiment, the estimated contract relates to a contract between the estimated party and an opposing party, the learned contract relates to a contract between a first other party and a second other party, the opposing party is indicated by a second abbreviation ("Party B") in the estimated contract, and the second other party is indicated by a second abbreviation ("Party B") in the learned contract, and the position estimation model further learns one or more learning data in which a label indicating the position of the second other party in the learned contract is associated with the learned contract information.

[0109] In some cases, multiple parties in a contract have contrasting positions. Therefore, information on the position of the other party may be useful in estimating the position of the presumed party in the contract. With this configuration, the position of the other party is taken into consideration when estimating the position of the presumed party, thereby improving the accuracy of position estimation.

[0110] In one embodiment, the information processing device 2 further includes an identification unit 106 that identifies the first abbreviated name ("Party A") based on the description in the estimated contract that is located around the part containing identification information such as the name of the estimated party.

[0111] According to this configuration, for example, the user does not need to set the first abbreviated name by himself / herself, which further improves the efficiency of reviewing the contract.

[0112] In one embodiment, the information processing device 2 further includes a type information acquisition unit 100b that acquires type information regarding the type of the contract to be estimated, and the position estimation unit 104a estimates the position of the party to be estimated based on the type information.

[0113] The type of position of a party may differ depending on the type of contract. With this configuration, the type of contract is taken into consideration when estimating the position of the party to be estimated, thereby improving the accuracy of estimating the position.

[0114] In one embodiment, the information processing device 2 further includes a relationship estimation unit 104b that estimates whether the presumed party in the presumed contract is in a different position from or in the same position as the other party in the presumed contract, based on the presumed contract information and a relationship estimation model that has learned one or more learning data in which labels indicating whether the multiple parties are in different positions or in the same position are associated with the presumed contract information regarding the training contract that indicates the contract contents between the multiple parties.

[0115] As mentioned above, depending on the type of contract, the parties may not be in opposing positions. This configuration makes it possible to deal with various types of contracts.

[0116] The above-described effects are merely examples and do not limit the scope of application of the present disclosure.

[0117] <5. Hardware Configuration> 13, an example of a hardware configuration in which the above-described information processing device 2 is realized by a computer 70 will be described. Note that the functions of each device can also be realized by dividing them into multiple devices.

[0118] As shown in FIG. 13, a computer 70 includes a processor 700 , a storage device 702 , an input I / F 704 , a data I / F 706 , a communication I / F 708 , and a display device 710 .

[0119] The processor 700 controls various processes in the computer 70 by executing programs stored in the storage device 702. For example, each functional unit included in the control unit 10 of the information processing device 2 can be realized by the processor 700 executing the programs stored in the storage device 702.

[0120] The storage device 702 is a storage medium such as a RAM (Random Access Memory), etc. The RAM temporarily stores the program code of the program executed by the processor 700 and data required when the program is executed.

[0121] The storage device 702 may also be a non-volatile storage medium such as a hard disk drive (HDD) or flash memory. The storage device 702 stores an operating system and various programs for implementing the above-described configurations. The storage medium storing the various programs may be a non-transitory computer-readable medium. The storage device 702 may also store tables that register various types of information and a DB that manages the tables. Such programs and data are loaded into the storage device 702 as needed and referenced by the processor 700.

[0122] The input I / F 704 is a device for receiving input from a user. Specific examples of the input I / F 704 include a camera, a button, a microphone, a keyboard, a mouse, a touch panel, various sensors, and a wearable device. The input I / F 704 may be connected to the computer 70 via an interface such as a USB (Universal Serial Bus).

[0123] The data I / F 706 is a device for inputting data from outside the computer 70. A specific example of the data I / F 706 is a drive device for reading data stored in various storage media. The data I / F 706 may be provided outside the computer 70. In this case, the data I / F 706 is connected to the computer 70 via an interface such as a USB.

[0124] The communication I / F 708 is a device for performing wired or wireless data communication with devices external to the computer 70 via the communication network 5. The communication I / F 708 may be provided external to the computer 70. In this case, the communication I / F 708 is connected to the computer 70 via an interface such as a USB.

[0125] The display device 710 is a device for displaying various types of information. Specific examples of the display device 710 include a liquid crystal display, an organic EL (Electro-Luminescence) display, and a display of a wearable device. The display device 710 may be provided outside the computer 70. In this case, the display device 710 is connected to the computer 70 via, for example, a display cable. Furthermore, when a touch panel is adopted as the input I / F 704, the display device 710 can be configured as an integral part of the input I / F 704.

[0126] The components of the information processing device 2 described in the above embodiment are assumed to realize predetermined processing in cooperation with other hardware when a program stored in the storage device 702 is executed by the processor 700. In other words, these components are assumed to be software or firmware, as well as corresponding hardware, and in both of these concepts, they are also referred to as "functions," "means," "parts," "processing circuits," "units," or "modules," and can be interpreted as such.

[0127] <6. Variations> The matters described in the above embodiment can be changed as appropriate within a range that does not cause inconsistencies.

[0128] In the above embodiment, the position estimation model and the relationship estimation model are described as separate models, but this is not limited to this. Specifically, the position estimation model and the relationship estimation model may be realized, for example, by a single neural network. In this case, the neural network may receive the text data of the contract to be estimated and the type of the contract to be estimated, and output whether the position of the estimated party is (1) one of the contrasting positions, (2) the other of the contrasting positions, or (3) a neutral position. If the type of contract is, for example, a confidentiality agreement, the neural network may output whether the position of the estimated party is (1) the discloser, (2) the recipient, or (3) a mutual discloser.

[0129] In the above embodiment, the estimation target contract information and the learning target contract information have been described as feature amounts determined based on the text data of the estimation target contract and feature amounts determined based on the text data of the learning target contract, respectively, but this is not limited to this. The estimation target contract information may be any of the text data of the estimation target contract, the results of preprocessing of the text data, and feature amounts extracted based on the results of the preprocessing. Similarly, the learning target contract information may be any of the text data of the learning target contract, the results of preprocessing of the text data, and feature amounts extracted based on the results of the preprocessing.

[0130] In the above embodiment, an example in which there are two parties to the contract has been described, but the present disclosure can also be applied to cases in which there are three or more parties.

[0131] In the above embodiment, an example in which the abbreviations of the parties are anonymized has been described, but this is not limiting. Specifically, the information processing device 2 may directly anonymize identification information such as the names of the parties included in the estimation target contract and the learning target contract.

[0132] In the above embodiment, the information processing device 2 has been described as generating a position estimation model and a relationship estimation model, but this is not limited to this. At least one of the position estimation model and the relationship estimation model may be generated in another device. In this case, the information processing device 2 may receive the position estimation model and the relationship estimation model from the other device and at least temporarily store them in the storage unit 12. The information processing device 2 may further train the received position estimation model and relationship estimation model using learning data, or may perform fine tuning.

[0133] When the position estimation model and the relationship estimation model are stored in another device, the information processing device 2 may use the position estimation model and the relationship estimation model via an API (Application Programming Interface) provided by the other device.

[0134] The system 1 may include a first information processing device that generates a position estimation model (see the position estimation model learning unit 102a) and estimates the positions of the parties to be estimated (see the position estimation unit 104a), and a second information processing device that generates a relationship estimation model (see the relationship estimation model learning unit 102b) and estimates whether the contract to be estimated relates to an opposing position or a neutral position (see the relationship estimation unit 104b). In this case, the information processing device 2 can also be said to include the first information processing device and the second information processing device.

[0135] The positions of the parties estimated by the information processing device 2 of the above embodiment can be used as input for other processes. For example, the positions of the parties estimated by the information processing device 2 can be used as input for a process related to automatic review of a contract (hereinafter referred to as "automatic review process").

[0136] Here, the automatic review process may be, for example, a process in which the information processing device 2 uses natural language processing technology to evaluate the contents of the contract to be reviewed based on the difference between the contents of the contract to be reviewed and the contents of a contract template or a contract that has already been reviewed.

[0137] Generally, the results of a contract review can vary depending on the perspective of one of the parties in the contract. For example, even if a statement does not pose a risk to one party, it may pose a risk to another party. Therefore, even in automated review processes, using the parties' positions as input enables more effective contract reviews.

[0138] The information processing device 2 may estimate the position of the estimated party in the estimated contract using the position estimation unit 104a, and then perform an automatic review process on the estimated contract from the perspective of the estimated position. The information processing device 2 may, for example, evaluate the degree of risk associated with a difference between the contents of the contract to be reviewed and the contents of a contract template, based on the position estimated by the position estimation unit 104a.

[0139] Furthermore, the information processing device 2 may execute an automatic review process on the contract to be estimated, and then estimate the position of the party to be estimated in the contract to be estimated, and extract review results related to the estimated position. For example, the information processing device 2 may extract, from the differences between the contents of the contract to be reviewed and the contents of the contract template, contents that are evaluated to have a relatively high level of risk from the position estimated by the position estimation unit 104a.

[0140] Furthermore, the information processing device 2 may output display information that provides an area for displaying the results of the automatic review process to, for example, the terminal device 3. This allows the user to evaluate the results of the automatic review process of the estimated contract and review the estimated contract.

[0141] The automatic review process may be executed by the information processing device 2 or by another device that can communicate with the information processing device 2.

[0142] In addition, for example, when storing concluded contracts, the position of a user (the organization to which the user belongs) in a certain contract can be extracted, associated with the contract, and stored in a database, making it easier to manage contracts.

[0143] 7. EMBODIMENTS OF THE PRESENT DISCLOSURE The present disclosure includes, for example, the following embodiments: Note that terms used in the above embodiments are shown in parentheses.

[0144] [Appendix 1] An information processing device 2 according to one embodiment of the present disclosure includes a contract information acquisition unit 100a that acquires first contract information (estimated contract information) regarding a first contract (estimated contract) that indicates the contractual content of at least the first party (estimated contract), and a first estimation unit (position estimation unit 104a) that estimates the position of the first party (estimated contract) in the first contract (estimated contract) based on a trained model (position estimation model) that receives the first contract information (estimated contract information) as input and outputs the position of the first party (estimated contract) in the first contract (estimated contract).

[0145] [Appendix 2] In the information processing device 2 described in Appendix 1, the trained model (position estimation model) may train one or more pieces of training data in which a label indicating the position of the first other party in the second contract (trained contract) is associated with second contract information (trained contract information) regarding the second contract (trained contract) that indicates the contract content of at least the first other party.

[0146] [Appendix 3] In the information processing device 2 described in Appendix 2, the first party (presumed target party) may be indicated by a first abbreviation ("Party A") in the first contract (presumed target contract), the first other party may be indicated by a first abbreviation ("Party A") in the second contract (contract to be studied), the first contract information (presumed target contract information) may be determined based on information regarding the structure of sentences containing the first abbreviation ("Party A") in the first contract (presumed target contract), and the second contract information (contract to be studied) may be determined based on information regarding the structure of sentences containing the first abbreviation ("Party A") in the second contract (contract to be studied).

[0147] [Appendix 4] In the information processing device 2 described in Appendix 3, the first contract information (estimated target contract information) may be determined based on information regarding the predicate-argument structure in which the first abbreviated name ("A") is an argument in the first contract (estimated target contract), and the second contract information (study target contract information) may be determined based on information regarding the predicate-argument structure in which the first abbreviated name ("A") is an argument in the second contract (study target contract).

[0148] [Appendix 5] In the information processing device 2 described in Appendix 3 or 4, the first contract information (estimated contract information) and the second contract information (learned contract information) may be determined based on information obtained by anonymizing the first abbreviation ("A") using a first method.

[0149] [Appendix 6] In the information processing device 2 described in Appendix 5, the first contract (estimated contract) may relate to a contract between a first party (estimated party) and a second party, and the second contract (study contract) relates to a contract between a first other party and a second other party, and the second party may be indicated by a second abbreviation ("Party B") in the first contract (estimated contract), and the second other party may be indicated by a second abbreviation ("Party B") in the second contract (study contract), and the first contract information (estimated contract information) and the second contract information (study contract information) may be determined based on information obtained by anonymizing the second abbreviation ("Party B") using a second method.

[0150] [Appendix 7] The information processing device 2 described in any one of Appendices 3 to 6 may further include an identification unit that identifies the first abbreviation ("Party A") based on the description of the first contract (presumed contract) surrounding the part containing the identification information of the first party (presumed party).

[0151] [Appendix 8] In the information processing device 2 described in any one of Supplementary Note 1 to Supplementary Note 7, the first party (presumed party) may be identified from multiple parties in the first contract (presumed contract) based on a user operation.

[0152] [Appendix 9] The information processing device 2 described in any one of Supplementary Notes 1 to 8 further includes a type information acquisition unit 100b that acquires type information regarding the type of the first contract (contract to be estimated), and the trained model (position estimation model) may further be input to the type information.

[0153] [Appendix 10] The information processing device 2 described in any one of Supplementary Notes 1 to 9 may further include a second estimation unit (relationship estimation unit 104b) that estimates whether the first party (estimated party) in the first contract (estimated contract) is in a different position from or in the same position as the second party in the first contract (estimated contract) based on another trained model (relationship estimation model) that receives first contract information (estimated contract information) as input and outputs whether the first party (estimated party) in the first contract (estimated contract) is in a different position from or in the same position as the second party in the first contract (estimated contract).

[0154] [Appendix 11] In the information processing device 2 described in Appendix 10, another trained model (relationship estimation model) may learn one or more learning data associated with third contract information relating to a contract that indicates the contents of a contract between multiple parties, the labels indicating whether the multiple parties are in different positions or the same position.

[0155] [Appendix 12] A program according to another aspect of the present disclosure causes a computer 70 to function as an output means for outputting the position of the first party (presumed target party) in the first contract (presumed target contract), estimated based on first contract information (presumed target contract information) regarding the first contract (presumed target contract) indicating at least the contract contents of the first party (presumed target party), and a trained model (position estimation model) that takes as input the first contract information (presumed target contract information) and outputs the position of the first party (presumed target party) in the first contract (presumed target contract).

[0156] [Appendix 13] An information processing method according to another aspect of the present disclosure causes a computer 70 to acquire first contract information (estimated target contract information) regarding a first contract (estimated target contract) indicating the contractual content of at least the first party (estimated target party), and estimate the position of the first party (estimated target party) in the first contract (estimated target contract) based on the first contract information (estimated target contract information) and a trained model (position estimation model) that takes the first contract information (estimated target contract information) as input and outputs the position of the first party (estimated target party) in the first contract (estimated target contract). [Explanation of symbols]

[0157] 1...system, 2...information processing device, 3...terminal device, 10...control unit, 12...storage unit, 70...computer, 100...acquisition unit, 100a...contract information acquisition unit, 100b...type information acquisition unit, 102...learning unit, 102a...position estimation model learning unit, 102b...relationship estimation model learning unit, 104...estimation unit, 104a...position estimation unit, 104b...relationship estimation unit, 106...identification unit, 108...transmission unit

Claims

1. a contract information acquisition unit that acquires first contract information relating to a first contract that indicates at least the contract content of the first party; a first estimation unit that estimates the position of the first party in the first contract based on a trained model that receives the first contract information as an input and outputs the position of the first party in the first contract; An information processing device comprising:

2. 2. The information processing device according to claim 1, wherein the trained model learns one or more pieces of training data in which second contract information relating to a second contract indicating the contract content of at least a first other party is associated with a label indicating the position of the first other party in the second contract.

3. The first party is identified in the first agreement by a first abbreviation; The first other party is identified in the second agreement by the first abbreviated name; the first contract information is determined based on information about a sentence structure including the first abbreviation in the first contract, The information processing device according to claim 2 , wherein the second contract information is determined based on information about a sentence structure including the first abbreviation in the second contract.

4. the first contract information is determined based on information about a predicate-argument structure in the first contract that has the first abbreviation as an argument, The information processing device according to claim 3 , wherein the second contract information is determined based on information about a predicate-argument structure in the second contract that has the first abbreviation as an argument.

5. The information processing device according to claim 3 , wherein the first contract information and the second contract information are determined based on information obtained by anonymizing the first abbreviation using a first method.

6. The first contract is a contract between the first party and the second party, the second contract is a contract between the first other party and a second other party; The second party is designated by a second abbreviation in the first agreement; The second other party is identified in the second agreement by the second abbreviated name; The information processing device according to claim 5 , wherein the first contract information and the second contract information are determined based on information obtained by anonymizing the second abbreviation using a second method.

7. The information processing device according to claim 3 , further comprising: an identification unit that identifies the first abbreviation based on a description in the first contract surrounding a portion in which the identification information of the first party is described.

8. The information processing device according to claim 1 , wherein the first party is identified from a plurality of parties in the first contract based on a user operation.

9. a type information acquisition unit that acquires type information regarding the type of the first contract, The information processing device according to claim 1 , wherein the first estimation unit estimates a position of the first party further based on the type information.

10. 2. The information processing device according to claim 1, further comprising: a second estimation unit that estimates whether the first party in the first contract is in a different position from or the same as the second party in the first contract, based on another trained model that receives the first contract information as an input and outputs whether the first party in the first contract is in a different position from or the same as the second party in the first contract.

11. The information processing device according to claim 10, wherein the other trained model learns one or more pieces of training data in which third contract information relating to a contract indicating the contents of a contract between multiple parties is associated with a label indicating whether the multiple parties are in different positions or in the same position.

12. Computer, A program that functions as an output means that outputs the position of the first party in the first contract, estimated based on first contract information regarding the first contract indicating at least the contract content of the first party, and a trained model that receives the first contract information as input and outputs the position of the first party in the first contract.

13. On the computer, Obtaining first contract information relating to a first contract indicating at least the contract content of a first party; Estimating the position of the first party in the first contract based on the first contract information and a trained model that uses the first contract information as input and outputs the position of the first party in the first contract; An information processing method for executing the above.

Images