JP2025118034A - Processing device, processing program, and processing method - Google Patents

Abstract

A processing device, a processing program, and a processing method are provided that are capable of processing document information more efficiently.
[Solution] A processing device is provided that has at least one processor, wherein the at least one processor is configured to acquire input document information, determine whether the acquired document information includes contract information, and execute processing to classify the acquired document information according to the result of the determination.
[Selected figure] Figure 4

Description

This disclosure relates to a processing device, a processing program, and a processing method configured to process document information.

Systems for acquiring details from contracts and conducting contract reviews have been proposed. For example, Patent Document 1 describes an information processing system for supporting lease contracts, comprising: "a scoring unit for calculating an evaluation score representing the creditworthiness of a lease property user based on data on details entered in a predetermined lease contract request form prepared in advance and credit information data on the lease property seller related to the lease contract; an evaluation unit for evaluating whether the lease contract request satisfies predetermined approval criteria based on the details entered in the lease contract request form and the evaluation score calculated by the scoring unit; a lease contract approval unit for approving the lease contract request if the evaluation unit determines that the approval criteria are met; and a contract creation unit for calculating a lease fee based on the details entered in the lease contract request form and automatically creating a lease contract with the lease property user (leaseholder)."

Japanese Patent Application Laid-Open No. 2001-195516

Based on the above-mentioned technologies, the present disclosure aims to provide, through various embodiments, a processing device, processing program, and processing method that can process document information more efficiently.

One aspect of the present disclosure provides a processing device including at least one processor, the processing device being configured to acquire input document information, determine whether the acquired document information includes contract information, and execute processing to classify the acquired document information in accordance with the result of the determination.

One aspect of the present disclosure provides a processing program that, when executed by at least one processor, causes the at least one processor to acquire input document information, determine whether the acquired document information includes contract information, and classify the acquired document information in accordance with the result of the determination.

One aspect of the present disclosure provides a processing method executed by at least one processor, the processing method including the steps of acquiring input document information, determining whether the acquired document information includes contract information, and classifying the acquired document information in accordance with the result of the determination.

This disclosure provides a processing device, processing program, and processing method that can process document information more efficiently.

The above effects are merely illustrative for the sake of convenience and are not limiting. In addition to or instead of the above effects, any effect described in this disclosure or any effect that would be obvious to one skilled in the art may be achieved.

FIG. 1 is a block diagram showing the configuration of a processing system 1 according to an embodiment of the present disclosure. FIG. 2 is a block diagram showing the configuration of the server device 100 according to an embodiment of the present disclosure. FIG. 3A is a diagram conceptually illustrating a document management table stored in the server device 100 according to an embodiment of the present disclosure. FIG. 3B is a diagram conceptually illustrating an example of document information stored in the server device 100 according to an embodiment of the present disclosure. FIG. 4 is a diagram showing a processing flow executed in the server device 100 according to an embodiment of the present disclosure. FIG. 5 is a diagram showing a processing flow executed in the server device 100 according to an embodiment of the present disclosure. FIG. 6 is a diagram showing a processing flow executed in the server device 100 according to an embodiment of the present disclosure. FIG. 7 is a diagram showing a processing flow executed in the server device 100 according to an embodiment of the present disclosure. FIG. 8A is a diagram illustrating an example of a screen output on the terminal device 200 according to an embodiment of the present disclosure. FIG. 8B is a diagram illustrating an example of a screen output on the terminal device 200 according to an embodiment of the present disclosure. FIG. 8C is a diagram illustrating an example of a screen output on the terminal device 200 according to an embodiment of the present disclosure. FIG. 8D is a diagram illustrating an example of a screen output on the terminal device 200 according to an embodiment of the present disclosure.

1. Overview of Processing System 1 The processing system 1 according to the present disclosure is a system used to determine whether acquired document information includes contract information and classify the acquired document information according to the results of the determination. In particular, the processing system 1 acquires document information input via an input interface of a user's terminal device and determines whether the acquired document information includes contract information. The processing system 1 then classifies the document information according to the results of the determination. If the acquired document information includes contract information as described above, the processing system 1 can perform appropriate processing according to the classification results, such as performing an analysis process of the contract information.

In this disclosure, document information refers to information stored with the intention of being referenced by individuals or organizations. Examples of such document information include various types of documents, such as official documents, contracts, minutes, rules, notices, regulations, invoices, estimates, reports, notifications, guide documents, and regulations. Document information may be written in any language, such as Japanese, English, Chinese, or a combination of these. Such document information may be in any format, such as document data format, presentation data format, image data format, print layout data format, or text data format in which text information is extracted from files in these formats.

Contract information is one type of document information described above and contains legally binding content. For example, it is used to demonstrate at least one expression of intent between two or more parties (who may be individuals or organizations), or to prove the fact of an agreement between the parties, particularly the formation of a contract. Examples of such contract information include non-disclosure agreements, outsourcing agreements, master transaction agreements, development outsourcing agreements, license agreements, gift agreements, purchase and sale agreements, lease agreements, settlement agreements, and joint research agreements, as well as various other documents that can prove the facts or formation of the above contracts, such as memoranda, regulations, rules, agreements, applications, purchase orders, requests, memoranda, and acknowledgements. As described for document information, such contract information can be written in any language, including Japanese, English, Chinese, or a combination of these. Such contract information can be in any format, such as document data, presentation data, image data, print layout data, or text data extracted from files in these formats. In the following, we will use a business outsourcing contract as an example of contract information, but of course this is not limited to this.

2. Configuration of Processing System 1 FIG. 1 is a block diagram showing the configuration of a processing system 1 according to an embodiment of the present disclosure. According to FIG. 1 , the processing system 1 includes at least a server device 100 and a terminal device 200, which are communicatively connected via a wired or wireless network. The server device 100 determines whether document information acquired from the terminal device 200 includes contract information and classifies the acquired document information according to the results of the determination. The terminal device 200 receives user input via an input interface to input document information, and displays various information received from the server device 100 (e.g., a document determination screen).

In this disclosure, the processing device refers to the server device 100, the terminal device 200, or a combination thereof. In other words, although the following describes the case where the server device 100 functions as a processing device, the terminal device 200 can also function as a processing device. In addition, in this disclosure, the storage and processing performed by the processing device may be distributed to other terminal devices, other server devices, etc. In other words, the processing device is not limited to those configured from a single housing, but includes the server device 100, the terminal device 200, other server devices, other terminal devices, or a combination thereof.

Furthermore, in Figure 1, only one terminal device 200 is shown. However, multiple users can use the processing system 1. In other words, the number of terminal devices 200 is not necessarily limited to one, and multiple terminal devices 200 may be included depending on the number of users.

3. Configuration of the Server Device 100 FIG. 2 is a block diagram showing the configuration of the server device 100 according to an embodiment of the present disclosure. According to FIG. 2, the server device 100 includes a processor 111, a memory 112, and a communication interface 113. These components are electrically connected to each other via control lines and data lines. The server device 100 does not need to include all of the components shown in FIG. 2; some components may be omitted, or other components may be added. For example, an external memory, a database device, a server device, or the like connected in a communicable manner as memory may be used. Furthermore, some processing may be distributed and executed among processing devices, including other server devices. In other words, the server device 100 is not limited to a single device, but may be distributed across multiple devices depending on the information handling and processing load.

The processor 111 functions as a control unit that controls the other components of the processing system 1 based on the processing program stored in the memory 112. Based on the processing program stored in the memory 112, the processor 111 determines whether the acquired document information contains contract information, and executes a process to classify the document information according to the result of that determination. Specifically, the processor 111 executes processes such as "processing to acquire input document information" and "processing to classify the acquired document information according to the result of the determination" based on the processing program stored in the memory 112. The processor 111 is primarily composed of one or more CPUs, but may also be combined with a GPU, FPGA, etc. as appropriate.

The memory 112 is composed of RAM, ROM, non-volatile memory, HDD, SSD, etc., and functions as a storage unit. The memory 112 stores instructions and commands for various controls of the processing system 1 according to this embodiment as processing programs. Specifically, the memory 112 stores programs to be executed by the processor 111, such as "processing for acquiring document information input via an input interface" and "processing for classifying acquired document information according to the results of judgment." In addition to these programs, the memory 112 also stores various information stored in document management tables, etc. Note that this information does not need to be constantly stored in the memory 112 within the server device 100, and may be stored in a database device installed remotely. In this case, the database device is also included in the memory 112.

The communication interface 113 functions as a notification unit for sending and receiving various information to and from the terminal device 200 connected via a wired or wireless network. Examples of the communication interface 113 include wired communication connectors such as USB and SCSI, broadband wireless communication devices such as wireless LAN, Bluetooth (registered trademark), and LTE, wireless communication transmission/reception devices such as infrared, and various connection terminals for printed circuit boards and flexible circuit boards. The communication interface 113, for example, receives document information from the terminal device 200 and sends output information of the document judgment screen to the terminal device 200.

4. Other Devices The configuration of the terminal device 200 includes, for example, a processor functioning as a control unit, a memory functioning as a storage unit, an input interface functioning as an input unit, an output interface functioning as an output unit, and a communication interface functioning as a communication unit. These components are electrically connected to each other via control lines and data lines. Note that the terminal device 200 does not necessarily need to include all of these components; it may be configured with some of the components omitted, or other components may be added. The terminal device 200 may be any device capable of communicating with the server device 100 via a wired or wireless network. Examples of the terminal device 200 include smartphones, tablets, laptop PCs, desktop PCs, scanners, image capture devices, and multifunction peripherals. Note that, as described above, when there are multiple users, multiple terminal devices 200 are used for each user, but the terminal devices 200 may each be a different type of terminal device. Furthermore, it is not necessary for one terminal device 200 to exist for one user; multiple users may use one terminal device 200, or one user may use multiple terminal devices 200.

The terminal device 200 inputs document information related to various documents, such as contracts, via an input interface and transmits the input document information to the server device 100 via a communication interface, for example, by having the processor process a program stored in memory. Specifically, the processor of the terminal device 200 photographs a document via a camera unit that functions as an input interface and inputs the photographed image data as document information. The processor of the terminal device 200 also accepts user input via an operation unit, such as a touch panel, that functions as an input interface, and inputs a file in the selected document format as document information. The processor of the terminal device 200 also captures document image data via a scanner unit and an operation unit, such as a touch panel, that functions as an input interface, and inputs the captured image data as document information. The processor of the terminal device 200 transmits the document information input in this manner to the server device via the communication interface.

The terminal device 200, for example, has a processor that processes a program stored in memory, and further outputs various output information received via a communication interface via the output interface. Specifically, the processor of the terminal device 200 receives output information for a document judgment screen related to document information from the server device 100 via the communication interface, and outputs the screen to an output interface such as a display.

In this embodiment, when the server device 100 acquires document information, it inputs the document information into a trained judgment model to determine whether it contains contract information. Therefore, the processing system 1 may, in some cases, further include a model generation device. The model generation device includes a processor, memory, and a communication interface, with these elements electrically connected to each other via control lines or data lines. The model generation device generates a trained judgment model by having the processor process a program stored in memory, learning pairs of training document information prepared for training and training judgment result information that labels the document information.

Note that such a model generation device is not necessarily required; the server device 100 or the terminal device 200 can also generate a trained decision model.

5. Various Information Used in Processing in the Processing System 1 Fig. 3A is a diagram conceptually illustrating a document information table stored in the server device 100 according to an embodiment of the present disclosure. According to Fig. 3A, the document management table of this embodiment stores document information, classification information, template information, attribute information, related contract information, risk information, and the like, in association with document ID information.

"Document ID information" is, for example, information unique to each piece of document information and is used to identify each piece of document information. Document ID information is generated by the server device 100, for example, each time new document information is received.

"Document information" is information stored with the assumption that it will be referenced by individuals or organizations. "Classification information" is information assigned based on the results of a determination of whether the document information contains contract information. In other words, classification information can be in any form, such as "contains contract information" or "does not contain contract information," as long as it can identify whether the document information contains contract information. It can also be in any form, such as expressions like "contains contract information" or "does not contain contract information," or specific codes that represent the above. Furthermore, if contract information is included, classification information can be stored as more detailed classifications based on the type of contract, such as outsourcing contracts or non-disclosure agreements. Similarly, if contract information is not included, classification information can be stored as more detailed classifications based on the type of non-contract document, such as employee invention regulations or work rules. "Template information" is information stored when document information includes contract information, and is information used to identify which party provided the contract stored as contract information. That is, the template information may be in any form, such as information indicating the position of "our company," "our organization," "yourself," "another company," "another organization," or "the other party," or information indicating that it is one of the templates stored in the system (such as the name of the template or an ID that can identify the template). Furthermore, the template information may not only indicate which party actually provided the contract template, but may also indicate the results of the template determination process. In the latter case, this may differ from the fact of who actually provided it.

"Attribute information" is information that is stored when document information includes contract information, and indicates the attributes of the document information. Such attribute information includes, for example, information that indicates the characteristics of the contract, as well as information separately entered by the user. Examples of attribute information include the contract title, names of the parties to the contract, transaction amount, contract start date, contract end date, whether or not automatic renewal occurs, contract period, contract termination notice deadline, and contract conclusion date. Attribute information only needs to be stored when at least contract information is included, and of course attribute information may also be stored even if contract information is not included.

"Related contract information" is information stored when document information includes contract information, and indicates other contracts related to the contract indicated by the contract information. For example, if the contract indicated by the contract information is a service outsourcing contract, a memorandum concluded to change the content of that service outsourcing contract corresponds to the related contract. When there is a related contract as described above, the related contract information stores information that can identify the relationship, such as the document ID information assigned to the related contract. Note that related contract information only needs to be stored when at least contract information is included, and naturally related contract information may be stored even if contract information is not included. Furthermore, when there is no related contract, related contract information is not stored, or information indicating that there is no related contract is stored.

"Risk information" is information stored when document information includes contract information, and indicates the results of a risk assessment of the contract indicated by the contract information. Such risk information includes information indicating the results of a comprehensive assessment based on the entire contract, information indicating the results of a risk assessment on an article-by-article or paragraph-by-paragraph basis for individual clauses included in the contract, and information indicating the results of a risk assessment on the words, phrases, or sentences that make up individual clauses. Examples of risk information include classification expressions such as "high," "medium," or "low" that indicate the level of risk, numerical values, sentence expressions that indicate the level of risk, risk content, methods for dealing with risk, and combinations of these. Note that risk information only needs to be stored when at least contract information is included; it may naturally be stored even if contract information is not included. Furthermore, if no risk information exists, either no risk information is stored, or information indicating the absence of risk information is stored.

In addition, such a document management table may also store other information as appropriate, such as user ID information indicating the user who entered each document information, authority information indicating the authority of that user, and organization ID information identifying the organization to which the user belongs.

Figure 3B is a conceptual diagram illustrating an example of document information stored in a server device 100 according to an embodiment of the present disclosure. Figure 3B shows document information B1, which includes contract information, as an example of document information. Document information B1 is composed of multiple pieces of part information, as shown as part information S1 to S20. Part information is document information divided into predetermined units. In the example of Figure 3B, page units are used as units for dividing document information, but of course, this is not limited to page units, and any unit can be used, such as unit, item, page, sentence, phrase, or word.

The example of document information B1 shown in Figure 3B includes parts information S1-S4, which is the main text of the service contract; parts information S5-S7, which is an official document (e.g., a registry) of the company (Corporation Y) that is the party (other party) to the service contract; parts information S8-S17, which is a non-disclosure agreement concluded with the parties to the service contract; and parts information S18-S20, which is the minutes of negotiations between the parties to the contract. In other words, document information B1 is configured as an integrated data file, with parts information S1-S4, parts information S5-S7, parts information S8-S17, and parts information S18-S20 imported in image data format, for example.

In this embodiment, a determination is made as to whether document information B1 contains contract information, but it is also possible to make this determination on a part information basis that makes up document information B1. In the example of Figure 3B, part information S1-S4 and part information S8-S17, which are contract information, are included. Therefore, document information B1 as a whole may be determined to contain contract information. Alternatively, part information S1-S4 and part information S8-S17, which are contract information, may be extracted from document information B1, and part information S1-S4 may be stored as new document information B1-1 and part information S8-S17 may be stored as new document information B1-2 in the document management table.

The determination may also be output on a page-by-page basis, or as information indicating a range of the same document information. Specifically, a determination may be output that part information S1, S2, S3, and S4 of document information B1 contain contract information, or a determination may be output that the range of part information S1 to S4 contains contract information.

6. Processing Flow Executed by Server Device 100 Figure 4 is a diagram showing a processing flow executed by the server device 100 according to an embodiment of the present disclosure. Specifically, Figure 4 is a diagram showing a series of processing flows from receiving document information from the terminal device 200, determining whether the document information includes contract information, and outputting the results as a document determination screen. This processing flow is mainly performed by the processor 111 of the server device 100 reading and executing a program stored in the memory 112.

4, in this embodiment, the processor 111 acquires document information from, for example, the terminal device 200 via the communication interface 113 (S111). In this case, the terminal device 200 receives input of document information via the input interface through processing by the processor of the terminal device, and transmits the document information input via the communication interface to the server device 100. Such input of document information is, for example,
This can be done using a variety of methods and procedures, such as: - Photographing a document via a camera unit that functions as an input interface and inputting the photographed image data as document information; - Accepting user input via an operation unit such as a touch panel that functions as an input interface and inputting a file in the selected document format as document information; - Importing image data of a document via a scanner unit via an operation unit such as a touch panel that functions as an input interface and inputting the imported image data as document information.

The method for obtaining document information is not limited to the above method. For example, document information can be obtained by the following method:
- Acquires multiple pieces of document information stored in the terminal device 200 by transferring them to the server device of these systems using a bulk upload system or bulk transfer system, and having the server device 100 receive them from the server device; - Acquires document information specified by the user by having the server device 100 receive it from a server device that provides a cloud storage service or electronic contract service used by the user of the terminal device 200; - Acquires document information by automatically transferring it to the server device 100 when the document information is uploaded from the terminal device 200 to the server device that provides the cloud storage service or electronic contract service used by the user of the terminal device 200, in order to use the cloud storage service or electronic contract service used by the user of the terminal device 200.

Furthermore, the server device 100 does not have to be dedicated to performing the various processes (e.g., content analysis process, etc.) described in Fig. 4, but may be provided as part of a server device that provides a cloud storage service, an electronic contract service, etc. In such a case, for example, document information is obtained by
- In order to use cloud storage services or electronic contract services used by the user of the terminal device 200, this can be done using various methods and means, such as obtaining document information from the terminal device 200 by uploading it to a server device that provides the service.

In this way, the acquisition of document information is not limited to directly receiving document information input into the terminal device 200, but can also be acquired from other server devices, etc. The method and means, the entity that inputs the document information, and the source of the document information can be any.

When the processor 111 receives document information, it generates new document ID information and stores the received document information in the document management table in association with the generated document ID information. Note that the document information may be saved, for example, in association with the case information to which the document information relates. Case information is information that indicates a case, such as case ID information. A case is, for example, a unit of work, such as the review of a contract related to certain document information. The case or case information may be set before the document information is input, or may be generated or set as a trigger when the document information is input.

The processor 111 reads the stored document information from the document management table and inputs it into the trained judgment model to determine whether the document information contains contract information (S112). As a result, the processor 111 obtains as output the judgment result from the trained judgment model as to whether the input document information contains contract information.

Here, FIG. 5 is a diagram showing a processing flow executed in the server device 100 according to an embodiment of the present disclosure. Specifically, FIG. 5 is a diagram showing an example of a processing flow for generating a trained judgment model used in S112 of FIG. 4. This processing flow is mainly performed by the processor 111 of the server device 100 reading and executing a program stored in the memory 112. The processing for generating the trained judgment model shown in FIG. 5 is performed before the determination of whether or not the contract information shown in FIG. 4 is included.

As shown in FIG. 5, the processor 111 executes a step of acquiring learning document information (learning document information) (S211). An example of learning document information is the document information illustrated in FIG. 3B, but it may also be document information that does not contain any contract information, or document information that contains only contract information. The processor 111 then executes a processing step of assigning a correct label indicating that the information is contract information to each piece of part information that makes up the document information, based on a determination by, for example, an administrator of the processing system 1 (S212). Once the labeling is complete, the processor 111 executes a step of storing the assigned correct label information as determination result information in association with the learning document information (S213).

Once the training document information and the associated correct label information have been obtained, the processor 111 executes a step of using them to perform machine learning of a determination pattern for determining whether each piece of part information that makes up the document information is a contract document (S214). As an example, this machine learning is performed by providing this set of information to a neural network that combines neurons, and repeatedly learning while adjusting the parameters of each neuron so that the output from the neural network is the same as the correct label information. Then, a step of acquiring a trained determination model is executed (S215). The acquired trained determination model may be stored in the memory 112 of the server device 100 or in another device connected to the server device 100 via a wired or wireless network.

Note that while a neural network is used as the learning machine in the above example, it is also possible to use other learning machines, such as convolutional neural networks, multi-layer hierarchical neural networks (MLP), long short term memory (LSTM), gated recurrent units (GRU), graph neural networks (GNN), and transformers; gradient boosting decision trees (GBDTs) such as Light Gradient Boosting Machine (LightGBM), XGBoost, and CatBoost; ridge regression, logistic regression, support vector regression (SVR), nearest neighbor methods, decision trees, regression trees, and random forests.

In addition, Figure 5 describes a trained judgment model generated by training pairs of training document information and correct labels. However, it is also possible to use a trained judgment model that utilizes deep learning, such as a generative trained judgment model. Among such generative trained judgment models, a trained judgment model known as a large-scale language model (LLM) is particularly preferred. More preferred examples of such trained judgment models include BERT (Bidirectional Encoder Representations from Transformers) and GPT (Generative Pre-trained Transformer), with ChatGPT or GPT-4 being particularly preferred. In this way, a general-purpose trained model can be used as the trained judgment model, rather than one generated solely for the judgment process.

The trained judgment model exemplified above may be stored as a program in memory 112 within server device 100 and processed by processor 111 executing the program, or it may be stored as a program in memory within another processing device communicatively connected to server device 100 and processed by a processor within that processing device. In the latter case, processor 111 of server device 100 sends input information in the form of a question to the processing device and obtains answer information as output from that processing device.

Furthermore, the trained judgment model does not have to be formed from a single trained judgment model; multiple trained judgment models can be combined and used. As an example, trained judgment models with the same training document information but different learning device structures can be used, or one trained judgment model can be a large-scale language model trained based on general-purpose training information, while the other trained judgment model can be a trained judgment model trained based on training document information including contract information, etc. When multiple trained judgment models are used in this way, output can be obtained by ensembling the answer information from each trained judgment model. Furthermore, the training document information input to the trained judgment model and the document information used in the judgment process can be converted into a numerical representation (e.g., vector values) that is easy to process with the trained judgment model by appropriate conversion processing such as embedding.

Furthermore, while Figure 5 illustrates a case in which each process is performed by the processor 111 of the server device 100, the process may also be generated by, for example, a model generation device connected to the server device 100 via a wired or wireless network.

Returning again to FIG. 4, when processor 111 obtains as output from the trained judgment model the judgment result as to whether or not contract information is included in the input document information, it determines whether or not the judgment result indicates that contract information is included (S113). If the judgment result indicates that contract information is included, processor 111 assigns classification information called "contract document" indicating that contract information is included, and stores the classification information in the document management table in association with the document ID information (S114). On the other hand, if the judgment result indicates that contract information is not included, processor 111 assigns classification information called "non-contract document" indicating that contract information is not included, and stores the classification information in association with the document ID in the document management table (S119).

As shown in FIG. 3B, the document information may include parts information S1-S4 and S8-S17, which are contract information, and parts information S5-S7 and S18-S20, which are non-contract information. In such cases, it is preferable for processor 111 to extract each of the parts information S1-S4 and S8-S17 determined to be contract information. Specifically, processor 111 divides the data files constituting document information B1 into data files constituting parts information S1-S4, data files constituting parts information S5-S7, data files constituting parts information S8-S17, and parts information S18-S20. Processor 111 then generates new document ID information B1-1 for parts information S1-S4, and stores the document information (parts information S1-S4) and classification information indicating a "contract document" in association with the document ID information. Similarly, the processor 111 generates new document ID information B1-2 for parts information S8 to S17 and stores the document information (parts information S8 to S17) and classification information indicating "contract document" in association with the document ID information. Similarly, the processor 111 generates new document ID information B1-3 for parts information S5 to S7 and stores the document information (parts information S5 to S7) and classification information indicating "non-contract document" in association with the document ID information. Similarly, the processor 111 generates new document ID information B1-4 for parts information S18 to S20 and stores the document information (parts information S18 to S20) and classification information indicating "non-contract document" in association with the document ID information. Then, in subsequent processing, each newly extracted document information is processed separately. This enables documents such as contracts to be handled more appropriately.

Note that this type of processing is an example of a case where contract information and non-contract information are mixed in the document information. In other words, when there is a mixture, the processor 111 may determine that the document information as a whole contains contract information and assign the label "contract document" to indicate that it contains contract information. In this case, the entire document information may be processed as a single entity in subsequent processing as well.

In addition, while Figure 4 shows only two classifications, contract documents and non-contract documents, as classification information, it is possible to provide even more detailed classifications by, for example, further subdividing the labels used when generating the trained judgment model and then training the model. That is, if contract information is included, processor 111 may store, as classification information, classifications based on the type of contract, such as outsourcing contracts, non-disclosure agreements, and purchase orders. Furthermore, if contract information is not included, processor 111 may store, as classification information, classifications based on the type of non-contract document, such as employee invention regulations and work rules.

In addition, in Figure 4, a determination is made as to whether or not contract information is included in the trained determination model, but this determination may also be made by further processing the output from the trained determination model by the processor 111. Furthermore, the output information for the above-mentioned subdivided contract types and non-contract documents may simply be rounded down to contract information or non-contract documents.

Next, if processor 111 determines in S113 and S114 that the document information contains contract information and is assigned the classification of "contract document," it executes a template determination process to determine which party provided the document information as a template (S115). As an example of the template determination process, processor 111 stores contract information (template information) of a template contract prepared by each user of a service provided by processing system 1 or each organization to which the user belongs. Processor 111 then determines whether the template was prepared by itself by calculating the similarity with the template information (e.g., cosine similarity, Jaccard coefficient, etc.). If processor 111 determines that the document information is its own template based on the similarity with the template (S116), it assigns a label "own template" to the document information and stores the assigned label in association with the document ID information of the document information as template information (S117).

Next, if it is determined in S113 and S114 that the document information contains contract information and is classified as a "contract document," the processor 111 executes a content analysis process on the document information (S118). The content analysis process may be any process that analyzes the content of the contract indicated by the contract information.
- Processing to extract attribute information from contract information - Processing to identify related contract information from contract information - Processing to analyze risks in a contract from contract information - Combinations of these processing are examples.

Here, FIG. 6 is a diagram showing a processing flow executed in the server device 100 according to an embodiment of the present disclosure. Specifically, FIG. 6 is a diagram showing an example of the processing flow of the "process of extracting attribute information from contract information" and the "process of identifying related contract information from contract information" from the content analysis processing executed in S118 of FIG. 4. This processing flow is mainly performed by the processor 111 of the server device 100 reading and executing a program stored in the memory 112.

As shown in FIG. 6, processor 111 reads files constituting document information including contract information (S311) and extracts attribute information contained in the document information (S312). This extraction process is performed by processor 111 executing a document analysis program on the files constituting the document information. As an example, processor 111 first extracts text information by analyzing the files constituting the document information. Note that this text information may be text information extracted and stored during the document type determination process. Processor 111 then performs linguistic analysis on the extracted text information to extract information on multiple pre-set items, such as the contract title, names of the parties to the contract, transaction amount, contract start date, contract end date, whether or not the contract is automatically renewed, contract period, contract termination notice deadline, and contract conclusion date.

When information is extracted for each item, the processor 111 associates each piece of extracted information with the document ID information and stores it in the attribute information (S313). Note that for items that could not be extracted, null data is stored.

Next, processor 111 generates a search term based on the extracted attribute information (S314). The search term is information used to search for candidates for other contracts (related contracts) related to the contract indicated by the contract information read in S311, and corresponds to a so-called search query. Processor 111 sets, for example, the names of the parties to the contract, the contract title, the contract conclusion date, etc., from the attribute information extracted in S312, as search terms.

Next, the processor 111 searches all document information stored in the document management table using the search term generated in S314 (S315). Specifically, the processor 111 calculates the relevance of each piece of document information to the search term, and identifies document information for which the calculated relevance is equal to or greater than a predetermined value. The processor 111 then extracts, from the identified document information, the document ID information of document information for which "contract document" is stored as classification information, as candidates for related contracts. The processor 111 transmits each document ID information and each document information as candidates for related contracts to the terminal device 200 via the communication interface 113 (S316).

When the terminal device 200 receives the related contract information candidates, the processor processes the terminal device 200 and displays a list of related contract candidates via the output interface. The processor of the terminal device 200 then accepts user input via the input interface, selects a related contract, and transmits the selection results to the server device via the communication interface. When the processor 111 of the server device 100 receives the selection results via the communication interface 113 (S317), it stores the received selection results as related contract information in association with the document ID information (S318).

This completes the processing flow for performing the "processing to extract attribute information from contract information" and the "processing to identify related contract information from contract information" out of the content analysis processing executed in S118 of Figure 4. Note that in Figure 6, both the "processing to extract attribute information from contract information" and the "processing to identify related contract information from contract information" are executed, but it goes without saying that only one of these processes may be executed, or other content analysis processes may be executed in addition.

FIG. 7 is a diagram showing a processing flow executed by the server device 100 according to an embodiment of the present disclosure. Specifically, FIG. 7 is a diagram showing an example of the processing flow for "processing for analyzing risks in a contract from contract information" in the content analysis processing executed in S118 of FIG. 4. This processing flow is mainly performed by the processor 111 of the server device 100 reading and executing a program stored in the memory 112.

As shown in FIG. 7, processor 111 reads a file constituting document information including contract information (S411) and inputs the read document information into the trained risk analysis model (S412). Specifically, processor 111 uses language analysis to break down the text information extracted from the target document information into individual clauses constituting the contract. Then, processor 111 inputs the text information constituting each clause into the trained risk analysis model for each of the broken down clauses. Processor 111 obtains the risk analysis results for each clause as output from the trained risk analysis model.

A trained risk analysis model can be generated, for example, as follows. First, a processor generates training text information by extracting text information for each individual clause that makes up the training contract. The processor also generates label information by labeling the generated training contract document information with information indicating the level of risk (e.g., "high," "medium," and "low"). The processor then provides the set of training text information and label information to a neural network that combines neurons, and repeats learning while adjusting each parameter so that the output from the neural network is the same as the label information. In this way, a trained risk analysis model is obtained.

As with the trained judgment model shown in Figure 5, such trained risk analysis models can use learners other than neural networks, and can also utilize generative trained analysis models such as large-scale language models. Furthermore, as with the trained judgment model shown in Figure 5, trained risk analysis models can also combine multiple trained risk analysis models.

Furthermore, the trained risk analysis model may not be generated by the server device 100, but may be generated or trained by a model generation device connected to the server device 100 via a wired or wireless network.

The trained risk analysis model may be stored as a program in the memory 112 of the server device 100 and processed by the processor 111 executing the program, or it may be stored as a program in the memory of another processing device communicatively connected to the server device 100 and processed by the processor within that processing device.

The server device 100 stores various risk-related information, such as risk content and how to deal with the risk, in association with text information that is the level of risk, the title of the clause, and the content of the clause. Therefore, when the processor 111 obtains analysis results indicating the level of risk from the trained risk analysis model, it obtains the risk content and how to deal with the risk based on the analysis results and the titles and text information of the individual clauses input to the trained analysis model. The processor 111 then stores the analysis results, risk content, how to deal with the risk, etc. as risk information (S413).

This completes the processing flow for performing the "process for analyzing risks in a contract from contract information" of the content analysis process executed in S118 of Figure 4. Note that Figure 7 describes the case where only the "process for analyzing risks in a contract from contract information" is executed, but the processes shown in Figure 6 may also be executed in addition.

Returning to Figure 4, once the various content analysis processes are completed, the processor 111 generates output information for the document judgment screen based on the attribute information, related contract information, risk information, etc. obtained through the content analysis processes. The processor 111 then transmits the generated output information for the document judgment screen to the terminal device 200 via the communication interface 113 (S120).

Note that in S113 of Figure 4, if the input document information does not include contract information, "non-contract document" is stored as the classification information, and the execution of the template determination process and content analysis process performed on document information classified as a "contract document" is restricted. Specifically, the processor 111 skips the execution of the template determination process and content analysis process on document information classified as a "non-contract document," omits part of the process, or executes a process different from the process content. This ends the processing flow.

As such, the template determination process and content analysis process shown in Figures 6 and 7 generally impose a heavy processing burden on the server device 100. Performing such processes on all document information would result in unnecessary processing load, so as shown in Figure 4, the template determination process and content analysis process are performed only on document information that includes contract information. This reduces the processing burden on the server device and enables more efficient processing of the document information received in S111 of Figure 4.

8A to 8D are diagrams illustrating examples of screens output on the terminal device 200 according to an embodiment of the present disclosure. Specifically, FIGS. 8A to 8D are diagrams illustrating examples of document judgment screens output on the terminal device 200 in S120 of FIG. 4. That is, FIGS. 8A to 8D illustrate examples of document judgment screens output via an output interface on the terminal device 200 that has received output information for the document judgment screen from the server device 100.

Figure 8A shows a document determination screen 10 that displays attribute information extracted in the content analysis process of Figure 6 after being classified as a contract document in S113 of Figure 4 or the like. The document determination screen 10 includes at least a classification information display area 11, a document information display area 12, a related information display area 13, and a template area 14. The classification information display area 11 is an area that displays classification information resulting from a classification based on the determination result of whether or not the document contains contract information in S113 of Figure 4 or the like. In the example of Figure 8A, "Contract" is displayed in the classification information display area 11, indicating that the document information displayed in the document information display area 12 contains contract information. Note that when a user's operation input to the downward-pointing triangle icon on the right side of the classification information display area 11 is accepted via the input interface, it is possible to switch to other document information with other classification information.

The document information display area 12 is an area for displaying the document information 15 received in S111 of FIG. 4 in any of the following formats: document data, presentation data, image data, print layout data, or text data extracted from files in these formats. The example in FIG. 8A shows an example of display in image data format, but the document information may be displayed in another data format, depending on the user's preferences. Referring to FIG. 8A, the document information 15 displayed in the document information display area 12 includes the contract title "Service Outsourcing Agreement," as well as text such as "X Corporation enters into..." and "Article 1 (Purpose)...." In other words, by displaying the specific contents of the document information 15 in the document information display area 12, it is possible to confirm the document information 15 by comparing it with information indicating the results of subsequent processing.

The related information display area 13 is an area for displaying at least one of various pieces of information (e.g., attribute information, related contract information, or risk information shown in FIG. 3A) obtained through various processes performed on document information. The related information display area 13 includes an attribute icon 16 and a related contract icon 17 at its top. That is, when a user's operation input for either the attribute icon 16 or the related contract icon 17 is accepted via the input interface, information corresponding to the icon to which the operation input was made is displayed in the related information display area 13. In FIG. 8A, the attribute icon 16 is displayed in a manner distinguishable from the related contract icon 17, indicating that an operation input for the attribute icon 16 has been accepted. Therefore, the related information display area 13 displays the attribute information extracted and stored by the processes of S312 and S313 in FIG. 6. Specifically, the related information display area 13 displays the following attribute information: the contract title, the names of the parties, the transaction amount, whether the contract can be changed, the contract start date, the contract end date, whether automatic renewal is enabled, the contract period, and the contract termination notice deadline. The related information display area 13 also includes a confirmed icon 18 and a correction icon 19. When a user input is accepted for each of these icons, it is possible to store a "confirmed" flag for the stored attribute information or to execute a correction process to correct the content.

The template area 14 is an area for displaying template information obtained in S115-S117 of Figure 4 to identify which party provided the template for the contract. In the example of Figure 8A, a check mark has been entered in the checkbox next to the "Company template" label in the template area 14. This indicates that information indicating that the template for document information 15 displayed in the document information display area 12 is a template prepared by the user has been stored. In other words, the user can confirm that document information 15 was created based on their own template.

Next, Figure 8B shows the document determination screen 10 for a document that has been classified as a contract document in S113 of Figure 4 or the like and that displays related contract information identified in the content analysis process of Figure 6. Note that, of the document determination screen 10 shown in Figure 8B, the classification information display area 11, document information display area 12, and template area 14 are the same as those in Figure 8A, and therefore their description will be omitted.

In Figure 8B, the related contract icon 17 in the related information display area 13 is displayed so as to be distinguishable from the attribute icon 16, indicating that operational input for the related contract icon 17 has been accepted. Therefore, the related information display area 13 displays the related contract information identified by the processing of S314 to S318 in Figure 6. Specifically, the related information display area 13 displays the attribute information for the memorandum of change to the outsourcing terms and the non-disclosure agreement as related contract information. The related information display area 13 also includes an add icon 22 and a modify icon 23. When user operational input for each of these icons is accepted, the information stored as related contract information can be added to or modified as desired by the user.

Next, Figure 8C shows a document determination screen 30 for a document that has been classified as a contract document in S113 of Figure 4 or the like and that displays risk information analyzed in the content analysis process of Figure 7. Of the document determination screen 30 shown in Figure 8C, the classification information display area 31, document information display area 32, and template area 34 are similar to the classification information display area 11, document information display area 12, and template area 14 of Figure 8A, and therefore their description will be omitted.

In FIG. 8C, the related information display area 33 includes risk information. Specifically, the related information display area 33 includes an overall result area 36 that displays the results of a risk analysis of the entire contract indicated in the document information 35, and clause risk display areas 37 and 38 that display the results of a risk analysis of individual clauses. The clause risk display area 37 displays clauses that have been determined to pose a "high" risk as a result of analyzing the risk of individual clauses included in the document information 35. The clause risk display area 38 displays clauses that have been determined to pose a "low" risk as a result of analyzing the risk of individual clauses included in the document information 35. In other words, the clause risk display area 37 and the clause risk display area 38 have outer frames that vary in display style depending on the level of risk. This allows the user to visually recognize the level of risk more intuitively.

In addition, each of the clause risk display areas 37 and 38 includes an indication of the degree of risk (for example, an indication such as "Importance: High" or "Importance: Low"), an indication of the content of the risk (for example, an indication such as "<Additional> Is the breakdown of the commission fee specified?" or "<Omission> Is there a provision regarding late payment penalties?"), and an indication of how to deal with the risk (for example, an indication such as "Example of response: Add the following as necessary..."). In other words, each of the clause risk display areas displays the content stored as risk information in the document management table.

The overall results area 36 contains aggregated information on the results of risk analysis for each clause included in the document information 35. In other words, the overall results area 36 contains the aggregated results of the number of clauses analyzed as having a "high" risk (importance), the number of clauses analyzed as having a "medium" risk, and the number of clauses analyzed as having a "low" risk.

Note that Figures 8A and 8B illustrate the case where each piece of information is switched and displayed by operating the attribute icon 16 and the related contract icon 17. Similarly, for the risk information shown in Figure 8C, a new risk icon may be provided so that each piece of information can be switched and displayed.

Next, Figure 8D shows the document determination screen 50, which is displayed when the document is classified as a non-contract document in S113 or the like of Figure 4, from among the document determination screens. The document determination screen 50 includes at least a classification information display area 51, a document information display area 52, and a related information display area 53. The classification information display area 51 is an area that displays classification information resulting from the classification based on the determination result of whether or not the document contains contract information in S113 or the like of Figure 4. In the example of Figure 8D, "Non-contract" is displayed in the classification information display area 51, indicating that the document information 54 displayed in the document information display area 52 is non-contract information. Note that when a user's operation input to the downward-pointing triangle icon on the right side of the classification information display area 51 is accepted via the input interface, it is possible to switch to other document information assigned with other classification information.

The document information display area 52 is an area for displaying the document information 54 received in S111 of FIG. 4 in any of the following formats: document data format, presentation data format, image data format, print layout data format, or text data format obtained by extracting text information from files in these formats. The example in FIG. 8D shows an example of display in image data format, but the document may be displayed in another data format, depending on the user's preferences. Referring to FIG. 8D, the document information 54 displayed in the document information display area 52 includes the document title "Work Rules" and text such as "Article 1 (Purpose)...". In other words, by displaying the specific contents of the document information 54 in the document information display area 52, it is possible to check the document information 54 against information indicating the results of subsequent processing performed on the document.

The related information display area 53 is an area for displaying various information (e.g., attribute information) obtained through various processes performed on the document information 54. The related information display area 53 displays attribute information extracted through attribute analysis processes performed on the document information 54, for example. Specifically, the related information display area 53 displays the document title and its implementation date as attribute information. The related information display area 53 also includes a confirmed icon 55 and a correction icon 56. When a user input is accepted for each of these icons, it is possible to store a "confirmed" flag for the stored attribute information or to perform a correction process to correct its contents.

As such, Figure 8D displays different information from the attribute information, related contract information, and risk information displayed for document information classified as a contract document in Figures 8A to 8C. This is because, for document information classified as a non-contract document, the execution of the processes for generating the attribute information, related contract information, and risk information in Figures 6 and 7 was skipped, or at least part of each process was only executed.

As described above, this embodiment provides a processing device, processing program, and processing method that can process document information more efficiently.

8. Other
4 to 7 illustrate the use of a trained judgment model or a trained risk analysis model in the processes related to each judgment or analysis. However, using such a trained model is merely an example, and judgments and analyses may naturally be performed using methods that do not use a trained model. For example, a large number of example contract documents may be stored in advance, and the processor 111 may determine whether the input document information is a contract document or a non-contract document by calculating the similarity between the input document information and each example document or by recognizing the title of the input document information. Furthermore, if the document information is assigned classification information other than classification information indicating whether it is a contract document or a non-contract document (e.g., a classification indicating whether two or more parties are listed or whether an agreement or similar wording is included) or attribute information, the processor may determine whether the document information is a contract document or a non-contract document based on the other classification information or attribute information.

In Figures 6 and 7, the content analysis processes described are "processing to extract attribute information from contract information," "processing to identify related contract information from contract information," and "processing to analyze risks in a contract from contract information." However, other content analysis processes may also be performed. Examples of such content analysis processes include a dashboard generation process that displays the results of tallying the attribute information of each document information stored in the document management table, and a document information editing process.

It is also possible to associate contract documents with non-contract documents. For example, if a contract document and a non-contract document are generated by splitting a single file, or if multiple files uploaded in a single folder contain contract documents and non-contract documents, the non-contract documents are likely to contain information that supplements the contract document. For business efficiency, it is preferable to store such multiple files in an associated manner. Therefore, it is preferable to associate files determined to be contract documents with documents determined to be non-contract documents, or to store them in an associated manner for a single case, or to prompt the user to associate them. This association of contract documents with non-contract documents is an example of content analysis processing.

The embodiments and variations of this disclosure are presented as examples and are not intended to limit the scope of this disclosure. This embodiment and variations may be implemented in a variety of other forms, and various omissions, substitutions, and modifications may be made without departing from the spirit of this disclosure. These embodiments and variations are within the scope and spirit of the invention, and are included in the invention and its equivalents as set forth in the claims.

The processes and procedures described in this disclosure can be realized not only by those explicitly described in the embodiments, but also by software, hardware, or a combination of these. Specifically, the processes and procedures described in this disclosure can be realized by implementing logic corresponding to the processes in media such as integrated circuits, volatile memory, non-volatile memory, magnetic disks, and optical storage. Furthermore, the processes and procedures described in this disclosure can be implemented as computer programs and executed by various computers, including processing devices and server devices.

Even if the processes and procedures described in this disclosure are described as being performed by a single device, software, component, or module, such processes or procedures may be performed by multiple devices, multiple software, multiple components, and/or multiple modules. Furthermore, even if the various pieces of information described in this disclosure are described as being stored in a single memory or storage unit, such information may be stored in multiple memories provided in a single device or in multiple memories distributed across multiple devices. Furthermore, the software and hardware elements described in this disclosure may be realized by integrating them into fewer components or by decomposing them into more components.

1 Processing system 100 Server device 200 Terminal device

Claims (12)

A processing device comprising at least one processor,
the at least one processor:
Obtain the input document information,
determining whether the acquired document information includes contract information;
classifying the acquired document information according to the result of the determination;
a processing unit configured to perform processing for: The processing device described in claim 1, wherein the determination of whether the document information includes the contract information is performed by inputting the document information into a trained determination model. The processing device described in claim 2, wherein the trained judgment model is obtained by learning based on training document information and training judgment result information indicating whether the training document information contains contract information. The processing device described in claim 1, wherein the document information includes information on multiple parts divided into predetermined units. The at least one processor:
determining whether or not each of the plurality of pieces of part information includes contract information;
If the result of the determination is that at least one piece of part information among the plurality of pieces of part information includes contract information, it is determined that the document information includes contract information.
The processing device of claim 4 configured to perform processing for: The at least one processor:
determining whether or not each of the plurality of pieces of part information includes contract information;
If the result of the determination is that at least one piece of part information among the plurality of pieces of part information includes contract information, classification information indicating that the at least one piece of part information includes contract information is assigned to the at least one piece of part information.
The processing device of claim 4 configured to perform processing for: The processing device according to claim 1, wherein if the acquired document information includes the contract information, a content analysis process is executed to analyze the contents of the contract information. The processing device according to claim 7, wherein the content analysis process is a process for acquiring attribute information of the contract information based on the content of the contract information. The processing device described in claim 7, wherein the content analysis process is a process for determining the risk of the contract indicated by the contract information based on the content of the contract information. The processing device according to claim 7, wherein execution of the content analysis process is restricted if the acquired document information does not include the contract information. When executed by at least one processor,
Obtain the input document information,
determining whether the acquired document information includes contract information;
classifying the acquired document information according to the result of the determination;
A processing program that causes the at least one processor to function in such a manner. A processing method executed by at least one processor, comprising:
obtaining input document information;
determining whether the acquired document information includes contract information;
classifying the acquired document information according to the result of the determination;
A processing method comprising: