Information processing system, information processing method, and program

Abstract

To facilitate extracting specific information from a document.SOLUTION: An information processing system comprises: a character string extraction unit that extracts character strings from form document data, which includes the activity amount of activities that emit greenhouse gases; and a generation processing unit that provides a generator, which generates a reply based on a pre-trained language model, with the character strings and an instruction to generate activity amounts and scopes corresponding to the activity amounts, and causes the generator to generate the activity amounts and the scopes.SELECTED DRAWING: Figure 1

Description

[Technical field]

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

[0002] 2. Description of the Related Art There is known an OCR system that acquires character data from a specific area in a document by OCR processing (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2010-073201 A Summary of the Invention [Problem to be solved by the invention]

[0004] It is necessary to interpret what the characters read from the document mean.

[0005] The present invention has been made in view of the above background, and has an object to provide a technique that can easily extract specific information from a form. [Means for solving the problem]

[0006] The main invention of the present invention for solving the above problem is an information processing system comprising: a character string extraction unit that extracts a character string from report data including an activity amount of an activity that emits greenhouse gases; and a generation processing unit that instructs a generator that generates an answer based on a trained language model to generate the character string, as well as the activity amount and a scope corresponding to the activity amount, and causes the generator to generate the activity amount and the scope.

[0007] Other problems and solutions disclosed in this application will be made clear in the description of the preferred embodiments of the invention and the drawings. Effect of the Invention

[0008] According to the present invention, specific information can be easily extracted from a form. [Brief description of the drawings]

[0009] [Figure 1] FIG. 1 is a diagram illustrating an example of an overall configuration of an information processing system. [Diagram 2] 2 is a diagram illustrating an example of a hardware configuration of a management server 2. FIG. [Diagram 3] 2 illustrates an example of the software configuration of a management server 2. FIG. [Figure 4] FIG. 13 is a diagram illustrating the operation of the management server 2. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] <Summary of the Invention> The present invention will be described below with reference to the preferred embodiments thereof. [Item 1] A character string extraction unit that extracts character strings from document data including activity amounts of activities that emit greenhouse gases; a generation processing unit that instructs a generator that generates an answer based on a trained language model to generate the character string, the amount of activity, and a scope corresponding to the amount of activity, and causes the generator to generate the amount of activity and the scope; An information processing system comprising: [Item 2] The information processing system according to item 1, The generation processing unit includes: causing the generator to learn information that associates activity identification information that identifies the activity with the scope; providing instructions to the generator to generate the character string, as well as the activity amount, the scope, and the activity specification information, thereby causing the generator to generate the activity amount, the scope, and the activity specification information; An information processing system comprising: [Item 3] Item 2. An information processing system according to the present invention, A plurality of the activities form a tree structure; the generation processing unit causes the generator to learn the first activity identification information, the scope, and second activity identification information that identifies the second activity that is a parent or child in the tree structure of the first activity identified by the first activity identification information; An information processing system comprising: [Item 4] A step of extracting a character string from document data including activity data related to greenhouse gas emissions; giving an instruction to a generator that generates an answer based on a trained language model to generate the character string, as well as the type of the activity amount to be extracted and a scope corresponding to the activity amount, and causing the generator to generate the activity amount and the scope related to the type; An information processing method characterized by being executed by a computer. [Item 5] A step of extracting a character string from document data including activity data related to greenhouse gas emissions; giving an instruction to a generator that generates an answer based on a trained language model to generate the character string, as well as the type of the activity amount to be extracted and a scope corresponding to the activity amount, and causing the generator to generate the activity amount and the scope related to the type; A program for causing a computer to execute the following.

[0011] <System Overview> An information processing system according to an embodiment of the present invention will be described below. In the information processing system of this embodiment, activity amounts related to activities that emit greenhouse gases are extracted from form data (which broadly includes data on various documents related to a company's economic activities, such as invoices, receipts, and delivery notes. Form data can be text data, document data, image data, etc.), and the type of the activity (scope and / or category in the GHG Protocol) is identified. In this embodiment, an LLM (Large Language Model) such as a GPT (Generative Pretrained Transformer) is used to extract the activity amounts from the form data and identify the type of activity amount.

[0012] 1 is a diagram showing an example of the overall configuration of an information processing system. The information processing system of this embodiment is configured to include a management server 2. The management server 2 is communicably connected to a user terminal 1 via a communication network. The communication network is, for example, the Internet, and is constructed by a public telephone line network, a mobile phone line network, a wireless communication path, Ethernet (registered trademark), etc.

[0013] The user terminal 1 is a computer operated by a user. The user terminal 1 may be, for example, a smartphone, a tablet computer, or a personal computer.

[0014] The management server 2 may be a general-purpose computer such as a workstation or a personal computer, or may be logically realized by cloud computing.

[0015] <Administration Server> FIG. 2 is a diagram showing an example of the hardware configuration of the management server 2. Note that the illustrated configuration is an example, and other configurations may be used. The management server 2 includes a CPU 201, a memory 202, a storage device 203, a communication interface 204, an input device 205, and an output device 206. The storage device 203 is, for example, a hard disk drive, a solid state drive, or a flash memory that stores various data and programs. The communication interface 204 is an interface for connecting to the communication network 3, and is, for example, an adapter for connecting to Ethernet (registered trademark), a modem for connecting to a public telephone line network, a wireless communication device for wireless communication, a USB (Universal Serial Bus) connector or an RS232C connector for serial communication, etc. The input device 205 is, for example, a keyboard, a mouse, a touch panel, a button, a microphone, etc. that input data. The output device 206 is, for example, a display, a printer, a speaker, etc. that output data. Each functional unit of the management server device 2 described later is realized by the CPU 201 reading a program stored in the storage device 203 into the memory 202 and executing it, and each storage unit of the management server 2 is realized as part of the storage area provided by the memory 202 and the storage device 203.

[0016] 3 is a diagram showing an example of the software configuration of the management server 2. The management server 2 includes a generator 210, a character string extraction unit 211, and a generation processing unit 212.

[0017] The generator 210 generates an answer based on the trained language model. The generator 210 may be a trained language model trained by machine learning and a function for generating a character string using the model. In this embodiment, the trained language model is assumed to be GPT. Note that the trained language model may not be managed by the management server 2, and an external server may be provided with the trained model, and an answer to an instruction may be generated using the trained language model by calling an API provided by the external server.

[0018] The character string extraction unit 211 extracts character strings from the form data. It is assumed that the form data includes the amount of activity that emits greenhouse gases. When the form data is text data, the character string extraction unit 211 can read out the contents of the text data. When the form data is image data, the character string extraction unit 211 can extract character strings drawn in the image data by a known OCR process. When the form data is binary data such as a word processor document or a spreadsheet document, the character string extraction unit 211 can extract character string data from the binary data by a known method.

[0019] The generation processing unit 212 causes the generator 210 to generate an activity amount and a scope (and / or a category). The generation processing unit 212 gives the generator 210 an instruction to generate the character string extracted by the character string extraction unit 211, a specified activity amount, and an instruction to generate a scope (and / or a category) corresponding to the activity amount, thereby causing the generator 210 to generate the activity amount and the scope (and / or a category).

[0020] The generation processing unit 212 can cause the generator to learn information that associates activity identification information that identifies an activity with a scope (and / or category). The generation processing unit 212 may perform fine tuning to cause a pre-trained language model to learn data that associates activity identification information with a scope and / or category, or may include data that associates activity identification information with a scope and / or category in an instruction (prompt) to be given to the generator 210. The generation processing unit 212 can cause the generator 210 to generate an activity amount, a scope, and activity identification information related to the activity instructed to the generator 210 by giving a character string extracted from the form data and an instruction to generate the scope and activity identification information.

[0021] In addition, when a plurality of activities form a tree structure (when major classifications, minor classifications, details, etc. of the activities are defined), the tree structure may be trained. In this case, the generation processing unit 212 can train the generator 210 to learn, for example, the first activity specification information, the scope, and the second activity specification information that specifies the second activity that is a parent or child in the tree structure of the first activity specified by the first activity specification information. In this learning, the trained language model may be updated in advance by fine tuning, or the prompt may include information that specifies the above tree structure (information that indicates the parent-child relationship between the first and second activity specification information). In this way, the generation processing unit 212 can generate multiple hierarchical levels of activity specification information from character strings included in the form data.

[0022] Alternatively, the management server 2 may be provided with a vector storage unit that records information indicating an activity and information obtained by vectorizing the information indicating the activity, and the generation processing unit 212 may vectorize the character string extracted by the character string extraction unit 211, read information indicating an activity corresponding to a vector close to the vector of the character string from the vector storage unit, and include the information in the prompt to be provided to the generator 210. The specialized information (information indicating an activity) stored in the vector storage unit may be provided to the generator 210 as learning data to generate the amount and scope (and / or category) of activity.

[0023] <Operation> FIG. 4 is a diagram illustrating the operation of the management server 2. As shown in FIG.

[0024] The management server 2 receives report data from the user terminal 1, extracts character strings from the received report data (S301), provides learning data including combinations of activity identification information and scope, the character strings extracted from the report data, and instructions to generate activity amount, scope, and activity identification information to the generator 210 (S302), and acquires and outputs the activity amount, scope, and activity identification information generated by the generator 210 (S303).

[0025] As described above, according to the information processing system of this embodiment, it is possible to efficiently generate the amount of activity related to an activity that emits greenhouse gases and the scope and / or category of the activity from the form data.

[0026] Although the present embodiment has been described above, the above embodiment is intended to facilitate understanding of the present invention, and is not intended to limit the present invention. The present invention may be modified or improved without departing from the spirit of the present invention, and equivalents thereof are also included in the present invention. [Explanation of symbols]

[0027] 1 User terminal 2 Management Server

Claims

1. An information processing system characterized in that it provides a generator that generates answers based on a trained language model with instructions to generate a string extracted from report data including the amount of activity of greenhouse gas emitting activities, as well as the type of activity to be extracted and the category corresponding to the activity, and causes the generator to generate the type and the category.

2. The information processing system according to claim 1, The generation processing unit provides the generator with the instruction, including information that associates activity identification information identifying the activity with the category. An information processing system characterized by the following.

3. The information processing system according to claim 2, Multiple of the aforementioned activities constitute a tree structure, The generation processing unit provides the generator with the instruction including the first activity identification information, the category, and the second activity identification information that identifies the second activity which is the parent or child of the first activity identified by the first activity identification information in the tree structure. An information processing system characterized by the following.

4. An information processing method characterized in that a computer gives instructions to a generator that generates answers based on a trained language model to generate a string extracted from report data including the amount of activity of greenhouse gas emitting activities, as well as the type of the amount of activity to be extracted and the category corresponding to the amount of activity, and the computer performs the step of causing the generator to generate the type and the category.

5. A program that causes a computer to perform the steps of giving instructions to a generator that generates answers based on a trained language model to generate a string extracted from report data including the amount of activity of greenhouse gas emitting activities, as well as the type of activity to be extracted and the category corresponding to the activity, and causing the generator to generate the type and the category.

Images