Programs, information processing devices, methods, and systems

Abstract

This system structurally eliminates errors and tampering caused by AI generation, thereby ensuring the consistency and reliability of the deliverables. [Solution] The program causes the processor to perform the following steps: obtain the target to be changed and the change instruction from the user; generate a first prompt including the target to be changed, the change instruction, specific information of the target to be changed, and the content of the change to the target to be changed, a system prompt for generating a structured transaction including the target to be changed, the change instruction, specific information of the target to be changed, and the content of the change to the target to be changed; input the first prompt to the generating AI and cause it to generate a structured transaction including the specific information and the content of the change included in the first prompt; verify whether the generated structured transaction conforms to a predetermined structure; execute the change to the target if the structured transaction conforms to the predetermined structure; and confirm the change to the target if the executed change to the target satisfies predetermined conditions.

Description

Technical Field

[0001] The present disclosure relates to programs, information processing apparatuses, methods, and systems.

Background Art

[0002] Patent Document 1 discloses a technique for ensuring the authenticity of image modification such as fake images. Specifically, Patent Document 1 discloses management means for managing, in the form of a blockchain, first information related to a first image and second information related to a second image modified from the first image, and determination means for determining whether the modification content of the second image is within the permitted range based on information indicating the permitted range of modification to the first image included in the first information. The management means is disclosed for a system that registers the second information in the blockchain when it is determined by the determination means that it is within the permitted range.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, Patent Document 1 mentions determination of whether the modification content of an image is within the permitted range, but does not describe a configuration for performing modification processing of an object using generative AI.

[0005] An object of the present disclosure is to structurally eliminate incorrect corrections and forgeries by generative AI and ensure the identity and reliability of the work product.

Means for Solving the Problems

[0006] To solve the above problems, a program according to one aspect of the present disclosure is a program to be executed on a computer equipped with a processor, the program to cause the processor to perform the following steps: obtain from a user a target to be changed and a change instruction relating to the content of the change to the target to be changed; generate a first prompt including the obtained target to be changed, the obtained change instruction, specific information of the target to be changed, and a system prompt for generating a structured transaction including the content of the change to the target to be changed; input the generated first prompt to a generation AI and cause it to generate the structured transaction including the specific information and the content of the change to be changed included in the first prompt; verify whether the generated structured transaction conforms to a predetermined structure; if the structured transaction conforms to the predetermined structure, execute the change to the target to be changed; and if the executed change to the target to be changed satisfies predetermined conditions, confirm the change to the target to be changed. [Effects of the Invention]

[0007] According to this disclosure, it is possible to structurally eliminate errors and tampering by generating AI, thereby ensuring the identity and reliability of the deliverables. [Brief explanation of the drawing]

[0008] [Figure 1] This figure shows a schematic configuration of an information processing system according to an embodiment of the present disclosure. [Figure 2] This figure shows the hardware configuration of a terminal device according to an embodiment of this disclosure. [Figure 3] This is a functional block diagram of a terminal device according to an embodiment of the present disclosure. [Figure 4] This figure shows the hardware configuration of the server according to the embodiment of this disclosure. [Figure 5] This is a functional block diagram of a server according to an embodiment of this disclosure. [Figure 6] This figure shows an example of the configuration of a prompt according to an embodiment of this disclosure. [Figure 7]This figure shows an example of the data structure of a user information table according to the embodiment of this disclosure. [Figure 8] This figure shows an example of the data structure of the change information table according to the embodiment of this disclosure. [Figure 9] This figure shows an example of the data structure of a transaction table according to the embodiment of this disclosure. [Figure 10] This is a flowchart showing the processing operation of the system according to the embodiment of this disclosure. [Figure 11] This figure shows a first example screen of a terminal device according to an embodiment of this disclosure. [Figure 12] This figure shows a second example screen of a terminal device according to an embodiment of the present disclosure. [Modes for carrying out the invention]

[0009] The embodiments of this disclosure will be described below with reference to the drawings. In all the drawings illustrating the embodiments, common components are denoted by the same reference numerals, and repeated explanations are omitted. The following embodiments are not intended to unduly limit the content of this disclosure as described in the claims. Not all components shown in the embodiments are necessarily essential components of this disclosure. Also, each drawing is a schematic diagram and is not necessarily a strict illustration.

[0010] Furthermore, in the following description, "processor" refers to one or more processors. A processor may be expressed, for example, as processing circuitry. At least one processor is typically a microprocessor such as a CPU (Central Processing Unit), but may be other types of processors such as a GPU (Graphics Processing Unit). At least one processor may be single-core or multi-core. Also, at least one processor may be a general-purpose processor or a purpose-specific processor.

[0011] Further, at least one processor may be a processor in a broad sense, such as a hardware circuit (e.g., FPGA (Field-Programmable Gate Array), ASIC (Application Specific Integrated Circuit)) that performs part or all of the processing.

[0012] In the following description, the expression such as "xxx table" may be used to describe information from which an output is obtained for an input. However, this information may be data of any structure or a learning model such as a neural network that generates an output for an input. Therefore, "xxx table" can be referred to as "xxx information".

[0013] In the following description, the configuration of each table is an example. One table may be divided into two or more tables, or all or part of two or more tables may be one table.

[0014] The program may be pre-installed in the information processing device shown below. For example, it may be on a recording medium readable (e.g., non-temporary) by the information processing device, and this program may be installed in the information processing device. Also, the program may be transmitted from a program distribution server to the information processing device and installed. In the following description, two or more programs may be realized as one program, or one program may be realized as two or more programs.

[0015] In the following description, identification information of various objects is used. However, the identification information may be information indicating a predetermined object, and the specific data is not limited to the embodiments. The identification information may be an identification number or an identifier including letters or symbols.

[0016] <Summary> The information processing system 1 according to this embodiment (hereinafter simply referred to as system 1) has a function of structurally preventing unintended incorrect corrections and tampering by the generative AI and ensuring the quality and reliability of the work product in text correction using the generative AI. Specifically, the server 20 causes the generative AI to generate a structured transaction based on the change target and the change instruction acquired from the user, verifies the compatibility of the transaction and the uniqueness and ambiguity of the specific information within the change target, and then executes the change. When the change result satisfies the quality assurance rule, the change is confirmed, and the data after confirmation is signaled and output via the generative AI, and the integrity of the data, such as the presence or absence of omission of units, inconsistency in order, or modification of content, is verified using identification tags. Thereby, incorrect corrections and tampering by the generative AI can be structurally eliminated, and the identity and reliability of the work product can be ensured.

[0017] <Overall configuration of the system> FIG. 1 is a block diagram showing an example of the overall configuration of system 1. System 1 according to this embodiment is a quality assurance system that realizes highly reliable text correction by strictly controlling the uncertainty while utilizing the function of the generative AI.

[0018] The system 1 shown in FIG. 1 includes, for example, a terminal device 10, a server 20, and a generative AI system 30. The terminal device 10, the server 20, and the generative AI system 30 are communicatively connected via, for example, a network 80.

[0019] In FIG. 1, an example in which system 1 includes one terminal device 10 is shown. However, for example, system 1 may include two or more terminal devices 10. Also, the server 20 is assumed to be composed of one device, but as another example, it may be configured as an aggregate of a plurality of devices. The way of distributing the plurality of functions required to realize the server 20 to the plurality of devices can be appropriately determined according to the processing capabilities of each device and / or the specifications required for the server 20.

[0020] The terminal device 10 is an information processing device operated by a user, such as a software or document developer, editor, or operator. The terminal device 10 can be implemented as a mobile device such as a smartphone or tablet. The terminal device 10 may also be implemented as a stationary PC (Personal Computer), laptop PC, etc. The terminal device 10 receives the user's specification of the object to be modified (e.g., source code file, configuration file, document file, etc.) and natural language modification instructions regarding the content of the modification, via a dedicated application or web browser, and transmits this information to the server 20. It also displays the final output, which is transmitted from the server 20 after the changes have been confirmed and converted into a signal.

[0021] Server 20 is a deterministic information processing device that is responsible for verifying, executing, quality-assuring, and finalizing text modifications in System 1, and operates to obtain the same input and output. Server 20 functions as a gateway to prevent the generation AI system 30 from directly accessing the target data. When Server 20 receives a change instruction from the terminal device 10, it instructs the generation AI system 30 to generate a structured transaction. Server 20 then verifies the suitability of the structured transaction returned by the generation AI system 30, and if suitability is confirmed, it provisionally executes the changes in a sandbox environment. Furthermore, after confirming that the results meet the quality assurance rules, it finalizes the changes and generates the deliverables as signaled data.

[0022] The Generative AI System 30 provides the use of Generative AI. This disclosure describes, as an example, the case in which the Generative AI System 30 provides the use of Large Language Models (LLMs), which are a type of Generative AI. Large Language Models are natural language models designed to perform multiple tasks of natural language processing. Large Language Models are an example of a trained model, and are models trained using a large number of parameters (e.g., billions to hundreds of billions) and high-level computing resources. Large Language Models are computer programs or algorithms designed to perform tasks of natural language processing. For example, in natural language processing, processes such as morphological analysis, syntactic analysis, information extraction, and text generation are performed, enabling a computer to analyze human language (i.e., natural language) and perform predetermined processing. When a prompt (instruction) is input to a Large Language Model, it generates output based on the text, image, etc., of the prompt. The prompt can be defined in natural language.

[0023] Examples of large-scale language models include the GPT® series (Generative Pre-TrAIned Transformer) developed by OPEN AI, StableLM developed by Stability AI, Llama2 developed by Meta, Palm2® and LamDA2® developed by Google. Note that other language models are also acceptable, not limited to large-scale language models. For example, BERT (Bidirectional Encoder Representations from Transformers) developed by Google may also be used. In this embodiment, the generation AI system 30 is used via an API (Application Programming Interface) provided by the server 20. Note that the generation AI system 30 is not limited to an external service; it may be operated by the provider of the server 20 itself. In this system 1, the generation AI system 30 generates structured transactions according to instructions from the server 20, but does not have direct editing rights to the target data. Furthermore, in the output of signalized data described later, it may be used as an unreliable communication channel.

[0024] A prompt is primarily a query input to a generating AI. A query can include, for example, text, strings, images, videos, audio, etc. A prompt may also include gestures. The user instructs the generating AI to process information by inputting a prompt. The user may include information in the prompt that causes the generating AI to produce a desired output.

[0025] The prompt is generated by the server 20 in a format that integrates, for example, the user's change target and change instruction (user prompt) with a pre-configured system prompt. Alternatively, the prompt may be generated by the terminal device 10 in a format that integrates, for example, the user's change target and change instruction with a system prompt.

[0026] The prompt may include information to specify the output format. Examples of such information include function definition information in a function call, JSON schema, XML format specification, and other structured data format specification. This information allows the output from the generating AI to be obtained in a predetermined structured format. For example, function definition information may include the function name, parameter names, parameter types, and parameter descriptions. Based on this output format specification information, the generating AI generates output results in the specified format.

[0027] A prompt contains an instruction. A prompt may also contain reference data. Here, an instruction is data that indicates the information processing to be performed by the generating AI. Reference data is data that the generating AI uses as a reference (learns from) when performing the information processing. The instruction and reference data may be included in the prompt in any form. For example, the instruction and reference data may be included in the prompt as text data entered in response to user operation. Another example is that the instruction may be included in the prompt as text data entered in response to user operation, and the reference data may be included in the prompt as a separate file from the text data. Furthermore, the instruction and reference data for the same instruction content do not necessarily have to be input to the generating AI as a single prompt. For example, a prompt containing the instruction may be input to the generating AI first, and then a prompt containing the reference data may be input to the generating AI.

[0028] <Terminal device configuration> Figure 2 is a block diagram showing an example of the hardware configuration of the terminal device 10 shown in Figure 1. As shown in Figure 2, the terminal device 10 comprises a control unit 101, a storage unit 102, a communication unit 103, an input unit 104, and an output unit 105. The terminal device 10 may also include a camera 106, a position sensor 107, and an acceleration sensor 108. Each block included in the terminal device 10 is electrically connected, for example, by a bus.

[0029] The control unit 101 performs various processes by executing various programs stored in the memory unit 102. The control unit 101 is, for example, a processor such as a CPU. A processor is hardware for executing instruction sets described in a program. A processor consists of an arithmetic unit, registers, peripheral circuits, etc. By operating according to the program, the control unit 101 performs the functions of an operation reception unit 131, a transmission / reception unit 132, and a presentation control unit 133. For example, the control unit 101 performs processing related to receiving the change target and change instructions, transmitting them to the server 20, and receiving and presenting signaled data received from the server 20.

[0030] The storage unit 102 includes a main memory and an auxiliary memory. The storage unit 102 stores various programs and various information. For example, the storage unit 102 stores an application program 120. The application program 120 includes, for example, a programming language that runs on a web browser application (not shown) stored in the storage unit 102, or it runs as a dedicated application.

[0031] The communication unit 103 performs processing such as modulation and demodulation processing for the terminal device 10 to communicate with an external device (for example, a server 20). The communication unit 103 performs transmission processing on the signal generated by the control unit 101 and transmits it to the external device. The communication unit 103 performs reception processing on the signal received from the external device and outputs it to the control unit 101.

[0032] The input unit 104 accepts instructions or information input from the user. The input unit 104 may be implemented, for example, by a touch-sensitive device that inputs instructions, etc. by touching the operating surface. If the terminal device 10 is a PC, the input unit 104 may be implemented by a reader, keyboard, mouse, etc. The input unit 104 converts the instructions, etc. input by the user into electrical signals and outputs them to the control unit 101. The input unit 104 may also include, for example, a receiving port that accepts electrical signals input from an external input device. The input unit 104 accepts input of the file or data to be changed, and natural language change instructions such as "change 30 seconds to 40 seconds," for example, via a text editor, an IDE (Integrated Development Environment) input panel, a command line interface, or a chat interface.

[0033] The output unit 105 presents information to the user. The output unit 105 is implemented, for example, by a display. The display shows various information (for example, output products that have been confirmed and signaled after being received from the server 20) in accordance with the control of the control unit 101. The display is implemented, for example, by an LCD (Liquid Crystal Display) or an organic EL (Electro-Luminescence) display. The output unit 105 may also include, for example, an output port that outputs electrical signals to an external output device. The output unit 105 includes, for example, a speaker that plays response voice from the generating AI. That is, presentation includes, for example, display on the display unit and output to a speaker or other output device.

[0034] Camera 106 is an imaging device that captures images using visible light. In other words, camera 106 is a device that receives visible light using a photodetector and outputs image data as a capture signal.

[0035] The position sensor 107 is a sensor that detects the position of the terminal device 10, and is generally a GNSS device, such as a GPS module. The position sensor 107 may detect the current position of the terminal device 10 from the position of the wireless base station to which the terminal device 10 is connected via the communication unit 103.

[0036] The acceleration sensor 108 is a sensor that detects the acceleration applied to the terminal device 10. Preferably, the acceleration sensor 108 has the function of detecting the tilt around each axis (X axis, Y axis, Z axis) of a three-dimensional coordinate system with the position of the terminal device 10 as the origin. An acceleration sensor 108 having such a function can detect the attitude of the terminal device 10 by detecting the gravitational acceleration due to the force of gravity on the Earth.

[0037] Figure 3 is a block diagram showing the functional units realized by the control unit 101. The control unit 101 comprises an operation reception unit 131, a transmission / reception unit 132, and a presentation control unit 133 as functional units. Specifically, the control unit 101 realizes each functional unit by reading the application program 120 stored in the storage unit 102 and executing the instructions contained in the application program 120.

[0038] The operation reception unit 131 processes instructions or information input from the input unit 104. Specifically, it receives the user's specification or input of the item to be changed (identification information such as file name or URI, or file data), and input of change instructions in natural language, and outputs the result to the main control logic of the control unit 101.

[0039] The transmitting / receiving unit 132 performs processing to enable the terminal device 10 to send and receive data with an external device (server 20) according to a communication protocol. Specifically, the transmitting / receiving unit 132 transmits the received change target and change instruction to the server 20 in response to instructions from the main control logic of the control unit 101. The transmitting / receiving unit 132 receives the signaled data, which is the final output, transmitted from the server 20 or the generating AI 30.

[0040] The presentation control unit 133 controls the output unit 105 to present the received signaled data to the user in accordance with instructions from the main control logic of the control unit 101. The presentation control unit 133 interprets the identification tags contained in the signaled data and displays the text data in a human-readable format. Furthermore, if an inconsistency is detected as a result of the integrity verification described later, control measures such as displaying an error message may be taken.

[0041] <Server Configuration> Figure 4 is a block diagram showing an example of the hardware configuration of the server 20 shown in Figure 1. As shown in Figure 4, the server 20 comprises a control unit 201, a storage unit 202, a communication unit 203, and an input / output IF 204. Each block included in the server 20 is electrically connected, for example, by a bus.

[0042] The control unit 201 executes various processes by running various programs (for example, application program 2024) stored in the memory unit 202. The control unit 201 is a processor such as a CPU or GPU. By operating according to the program, the control unit 201 performs the functions of a receiving control unit 2031, a transmitting control unit 2032, a presentation control unit 2033, and an information processing unit 2034.

[0043] The storage unit 202 includes main memory and auxiliary storage. The storage unit 202 stores various programs, various information, and various tables. The storage unit 202 stores an application program 2024 for executing the text modification quality assurance process according to this embodiment. The storage unit 202 stores various databases, such as a user information table 2021, a change information table 2022, and a transaction table 2023.

[0044] User Information Table 2021 is a database used to manage users of System 1, for example. User Information Table 2021 stores information such as the user's "User ID," "Name," and "Contact Information." Further details will be provided later.

[0045] The Change Information Table 2022 is a database for managing change requests received from terminal device 10. The Change Information Table 2022 stores items such as "Change Information ID," "User ID," "Item to be changed," "Change instruction," and "Receipt date and time." Further details will be described later.

[0046] Transaction Table 2023 is a database for managing the execution plan of changes generated by the generation AI system 30, as well as its verification and execution results. Transaction Table 2023 stores items such as "Transaction ID," "Change Information ID," "Specific Information," "Change Details," "Verification Status," "Decision Result," and "Data After Change Confirmation." Further details will be described later.

[0047] Application program 2024 is application software for implementing the text modification quality assurance process according to the present invention. Application program 2024 includes program code for performing a series of processes, such as receiving change instructions, generating a first prompt, verifying structured transactions, provisionally executing changes, determining quality assurance rules, confirming changes, and generating signaled data.

[0048] The communication unit 203 performs modulation and demodulation processing for the server 20 to communicate with external devices (for example, the terminal device 10 and the generation AI system 30). The communication unit 203 performs transmission processing on the signals generated by the control unit 201 and transmits them to the external devices. The communication unit 203 performs reception processing on the signals received from the external devices and outputs them to the control unit 201.

[0049] The input / output interface 204 receives instructions or information input from the administrator of server 20 and functions as an interface for presenting information to the administrator. For example, the input / output interface 204 is connected to input / output devices such as a keyboard, mouse, and display.

[0050] Figure 5 is a block diagram showing the functional units realized by the control unit 201. The control unit 201 comprises a receiving control unit 2031, a transmitting control unit 2032, a presentation control unit 2033, and an information processing unit 2034 as functional units. Specifically, the control unit 201 realizes each functional unit by reading the application program 2024 stored in the storage unit 202 and executing the instructions contained in the application program 2024.

[0051] The receiving control unit 2031 processes data or information transmitted from an external device via the communication unit 203. Specifically, the receiving control unit 2031 receives, for example, the data to be changed and the change instructions transmitted from the terminal device 10. The receiving control unit 2031 also receives text data, including structured transactions, transmitted from the generation AI system 30.

[0052] The transmission control unit 2032 performs processing via the communication unit 203 to enable the server 20 to send and receive data with external devices according to a communication protocol. Specifically, the transmission control unit 2032 transmits, for example, the first prompt data and the final signaled data to the generation AI system 30 in response to instructions from the information processing unit 2034. If verification fails, it also sends feedback including the error information to the generation AI system 30.

[0053] The presentation control unit 2033 controls the transmission control unit 2032 to present the execution results of the information processing unit 2034 (for example, the verification results of structured transactions created by the generation AI system 30, or signaled data which is a confirmed deliverable) to the terminal device 10. The presentation control unit 2033 also controls a display device connected to the input / output IF 204, for example, to present various information to the administrator of the server 20.

[0054] The information processing unit 2034 performs a series of processes related to quality assurance for text corrections. The information processing unit 2034 verifies the conformity of the structured transactions generated by the generation AI system 30, and if conformity is guaranteed, it executes the changes indicated by the text corrections.

[0055] Specifically, the information processing unit 2034 first obtains the object to be changed and the change instruction from the terminal device 10. The object to be changed includes, for example, text data such as source code files, configuration files, or document files. The change instruction is an instruction regarding the content of the change to the object to be changed, and includes, for example, natural language text instructing the change of a specific setting value. Next, the information processing unit 2034 generates a first prompt that includes context information related to the obtained object to be changed, the change instruction, and a system prompt which is a set of instructions for generating a structured transaction, and sends it to the generation AI system 30 via the transmission control unit 2032. The information processing unit 2034 receives a response text containing a structured transaction from the generation AI system 30. The structured transaction has a structure that includes identification information (Target) for uniquely identifying the location of the change within the object to be changed, and change content (Operation) that indicates the specific content of the change. The identification information includes, for example, a string pattern or line number range that is assumed to be uniquely identifiable within the target data. The change content includes, for example, operation types such as replacement, insertion, and deletion.

[0056] The information processing unit 2034 physically extracts (cuts out) only the structured transaction portion from the natural language text containing noise generated by the generation AI system 30. Next, the information processing unit 2034 verifies whether the extracted structured transaction conforms to a predetermined structure. This verification of conformity to the predetermined structure includes, for example, verifying the existence of required keys and the appropriateness of the data types of each value, and verifying the uniqueness within the data to be modified using specific information contained in the structured transaction. Specifically, if a required key is missing or a data type mismatch is found during the conformity verification process, it is determined that the transaction does not conform to the predetermined structure. Furthermore, if there are multiple matching candidates for the specific information, if there are no matching candidates, or if an approximate match is detected that matches except for differences in whitespace or line breaks, it is determined that the transaction does not conform to the predetermined structure. If it is determined that the transaction does not conform, the information processing unit 2034 stops the execution of the change to eliminate the inference made by the generation AI and feeds back the reason for the mismatch to the generation AI system 30. At this time, for example, the information processing unit 2034, based on the feedback error information, instructs the generating AI system 30 to expand (or concretize) the search conditions by including surrounding strings (preceding or succeeding lines, or blocks, etc.) in the specific information, and to regenerate a new structured transaction so that it can be uniquely identified within the target data. This enables an automated feedback loop in which the generating AI system 30 reconsiders the search conditions and resubmits a new structured transaction.

[0057] If all validations are passed, the information processing unit 2034 provisionally executes the changes to be modified in accordance with structured transactions in a logically isolated sandbox environment. At this time, areas other than uniquely identified parts are retained as immutable zones. Next, the information processing unit 2034 determines whether the modified data obtained as a result of the provisional execution satisfies predetermined quality assurance rules. The quality assurance rules include, for example, specific prohibited elements, required elements, regular expression patterns, or syntax rules. If a rule violation is detected, the change is not finalized, and the violation details are fed back to the generation AI system 30 to prompt replanning.

[0058] If the quality assurance rules are met, the information processing unit 2034 confirms the changes to be modified. Specifically, it reflects (commits) the results of the trial execution to the actual data and saves the modified state to the storage unit 202. After confirming the changes, the information processing unit 2034 divides the confirmed data into predetermined units (e.g., row units or semantic block units), assigns an identification tag to each unit, and generates signaled data. The identification tag includes at least unit identification information (e.g., a sequential number) and a verification value (e.g., a hash value) to verify the integrity of the content. In particular, by making the verification value a chain hash value calculated based on the verification value of the preceding unit, it is guaranteed that multiple units as a whole have indivisible order and integrity.

[0059] Finally, the information processing unit 2034 outputs the generated signaled data via the generation AI system 30, which acts as a communication relay unit, and performs integrity verification using identification tags based on the signaled data immediately before presentation. The information processing unit 2034 physically extracts only the output portion with the identification tag from the text output from the generation AI system 30, and verifies whether there are any missing units, inconsistencies in order, or alterations to the content (editing such as summarization or omission) by referring to the identification tag. If the verification fails, the output is discarded and a retransmission request is sent to the generation AI system 30. If the verification is successful, it is presented to the terminal device 10 as a valid output. When sending a retransmission request, the information processing unit 2034 may also send error information indicating that an incongruity such as summarization or omission was detected in the previous output, and a command to force the output of the data as is without modification. This enables the generation AI system 30 to accurately re-output the signaled data without modification based on the error information and command.

[0060] Figure 6 shows an example of the configuration of the first prompt P used in this embodiment. As shown in Figure 6, prompt P is text data and consists of a structured transaction generation instruction P1, an output format P2, and a user prompt P3. Of these, the structured transaction generation instruction P1 and the output format P2 correspond to the system prompt, which is a set of instructions for generating a structured transaction, and the user prompt P3 corresponds to the item to be changed and the change instruction obtained from the user.

[0061] (Instruction to generate structured transaction P1) The structured transaction generation instruction P1 includes a set of instructions that specify the role, purpose, and constraints on the generation AI system 30 when performing this task. For example, it may include a role such as, "You are an expert who interprets the user's natural language modification instructions and converts them into machine-executable structured transactions," or constraint information that takes the system's verification logic into account, such as, "Vague specific information will be rejected in the subsequent deterministic verification process, so exclude guesswork by the generation AI and generate specific information that is as accurate and unique as possible." It may also include an instruction stating, "No natural language responses, apologies, or self-assessments are required," in order to strictly control the output.

[0062] (Output format P2) Output format P2 includes a schema definition of the structured transaction to be generated, and information defining the meaning of each component. Specifically, it includes information specifying a structured data format such as JSON or YAML, which has required keys such as "Target" for uniquely identifying the changed parts within the target data, and "Operation" indicating the specific details of the changes. It may also include detailed specification definitions such as the expected data type of the value corresponding to each key, or the assumption that the Target information can be uniquely identified within the target data.

[0063] (User prompt P3) User prompt P3 is user input information that the generating AI system 30 refers to when performing information processing, and may include information corresponding to the instruction statement or reference data described above. User prompt P3 includes context information of the data to be changed or a part thereof, obtained by the information processing unit 2034 from the change information table 2022, or file data obtained from the terminal device 10, and text data of the user's natural language change instruction. For example, it may include specific data such as "Target to change: 'TIMEOUT_SEC = 30'" and "Change instruction: 'Please set the timeout to 40 seconds'".

[0064] <Data structure> In this embodiment, the main data structures managed by the storage unit 202 of the server 20 will be described with reference to Figures 7, 8, and 9. Note that the data structures described are examples and do not exclude data not described.

[0065] Figure 7 shows an example of the data structure of User Information Table 2021. As shown in Figure 7, User Information Table 2021 is a table that has columns such as Name and Contact Information, with User ID as the key. User Information Table 2021 is a database for managing users of System 1, for example. Various information about the same user is stored in a single record.

[0066] The "User ID" field stores an identifier to uniquely identify a user of System 1. The "Name" field stores, for example, the user's name. The "Contact Information" field stores, for example, information about the user's contact details (e.g., phone number, email address, etc.). For example, User Information Table 2021 is generated or updated based on the information entered when a user registers to use System 1.

[0067] Figure 8 shows an example of the data structure of the change information table 2022. As shown in Figure 8, the change information table 2022 is a table that has columns such as user ID, item to be changed, change instruction text, and reception date and time, with the change information ID as the key. This table is used to manage change requests from users.

[0068] The "Change Information ID" field stores an identifier to uniquely identify each change request. The "User ID" field is an identifier to identify the user who requested the change, and corresponds to the User ID in Figure 7. The "Item to be Changed" field stores identification information such as the file name, database record identifier, URI, or file data (imported source code or document data, etc.) that will be changed. The "Change Instruction" field stores the change instruction entered by the user in natural language (e.g., "Change the timeout to 40 seconds"). The "Received Date and Time" field stores the date and time when this change request was received.

[0069] Figure 9 shows an example of the data structure of transaction table 2023. As shown in Figure 9, transaction table 2023 is a table that has columns such as change information ID, specific information, change details, verification status, judgment result, and data after change confirmation, with transaction ID as the key. This table is used to manage the execution plan of changes generated by the generation AI system 30, as well as its verification and execution results.

[0070] The "Transaction ID" field stores an identifier that uniquely identifies each structured transaction. The "Change Information ID" field is an identifier that indicates which change request this transaction corresponds to, and corresponds to the Change Information ID in Figure 8. The "Specific Information" field stores information (e.g., string pattern, row number range, etc.) generated by the generation AI system 30 to uniquely identify the changed parts within the data to be changed. The "Change Details" field stores the details of the specific change operation (e.g., replacement string, insertion / deletion operation instructions, etc.).

[0071] The "Verification Status" field stores the verification result of the transaction's conformance to the predetermined structure (e.g., "Verification in progress," "Success," "Non-conformity (Missing required key)," "Non-conformity (Multiple matches)," "Non-conformity (Approximate match detected)," etc.). The "Judgment Result" field stores the judgment result of the quality assurance rule on the data after the trial execution (e.g., "Pass," "Violation (Prohibited element detected)," etc.). The "Data after Change Confirmation" field stores the signaled data (data including identification tags and chained hash values ​​for each partition unit, or the result of its integrity verification) which is the output when the change is confirmed.

[0072] <Operation> Figure 10 is a flowchart showing an example of the operation of System 1 according to this embodiment. The operation flow shown in Figure 10 includes a series of processes to ensure the quality of text correction using the generation AI.

[0073] In step S11, the terminal device 10 sends the target to be changed and the change instruction to the server 20 based on the user's operation. Specifically, the transmitting / receiving unit 132 of the terminal device 10 receives input of the identification information of the file to be changed or the file data itself (for example, "TIMEOUT_SEC = 30") and a change instruction in natural language such as "change the timeout setting from 30 seconds to 40 seconds", and sends it to the server 20.

[0074] In step S12, the receiving control unit 2031 of the server 20 acquires the items to be changed and the change instructions transmitted from the terminal device 10. The information processing unit 2034 of the server 20 records this acquired information in the change information table 2022, assigning a unique change information ID to it.

[0075] In step S13, the server 20 generates a first prompt. Specifically, the information processing unit 2034 of the server 20 takes the data to be changed, which includes the description "TIMEOUT_SEC = 30" obtained in step S12, and the change instruction text "Change from 30 to 40", as the user prompt P3, and integrates it with the system prompt (structured transaction generation instruction P1 and output format P2) stored in the memory unit 202 to generate the first prompt. The information processing unit 2034 inputs the generated first prompt to the generation AI system 30.

[0076] In step S14, the generating AI system 30 generates a structured transaction based on the first prompt sent from the server 20. The information processing unit 2034 of the server 20 receives a response text containing the structured transaction from the generating AI system 30. This response text contains natural language noise and a structured transaction, for example, "Understood. The following is the transaction: {"target":"TIMEOUT_SEC = 30","operation":"TIMEOUT_SEC = 40"}". At this time, the information processing unit 2034 physically extracts only the part of the structured transaction that contains the information ""target":"TIMEOUT_SEC = 30"" and ""operation":"TIMEOUT_SEC = 40"" from the response text which contains natural language noise.

[0077] In step S15, the server 20 verifies whether the structured transaction extracted in step S14 conforms to a predetermined structure. Verification of conformity to the predetermined structure includes, for example, verifying the existence of required keys and the appropriate data types of each value, and verifying the uniqueness within the data to be modified using specific information contained in the structured transaction. Specifically, the information processing unit 2034 first verifies whether the extracted structured transaction conforms to a predetermined schema defined in the output format P2 (for example, the existence of the required keys "target" and "operation" and the data types of each value). Here, for example, in data type verification, the information processing unit 2034 determines whether the actual value output from the generating AI system 30 is composed of a different data type, such as a numeric type, when, for example, a string type is specified as the value of "target". Next, the information processing unit 2034 searches within the data to be modified using specific information within the structured transaction (for example, the string pattern "TIMEOUT_SEC = 30") and verifies the uniqueness of whether the specific information can be uniquely identified within the target data. In the conformance verification process, if a required key is missing or a data type is incompatible, or if there are multiple matching candidates for specific information, no matching candidates exist, or an approximate match is detected that matches except for differences in whitespace or line breaks, the information processing unit 2034 determines that the data does not conform to the predetermined structure in order to eliminate the inference made by the generation AI system 30, and terminates the process. If it is determined that the data does not conform, the information processing unit 2034 feeds back error information, including the reason for the nonconformity (for example, "The target cannot be uniquely identified because there are multiple matching candidates"), to the generation AI system 30, and returns to step S14 to regenerate a new structured transaction.

[0078] In step S16, if the server 20 meets the requirements of the verification in step S15, it executes the changes. Specifically, the information processing unit 2034 of the server 20 performs a provisional execution process in a logically isolated sandbox environment, replacing the location identified by the specific information "TIMEOUT_SEC = 30" with "TIMEOUT_SEC = 40". At this time, the areas other than the identified location are retained as immutable areas.

[0079] In step S17, the server 20 verifies whether the data obtained by executing the changes in step S16 satisfies predetermined conditions. Specifically, the information processing unit 2034 of the server 20 applies quality assurance rules, including, for example, prohibited elements, required elements, regular expression patterns, or syntax rules, to the modified data and confirms that there are no rule violations. If it is determined that the predetermined conditions are met (the verification is passed), the information processing unit 2034 proceeds to step S18, formally reflects the provisionally executed changes in the original target data to confirm the changes, and saves the modified state to the storage unit 202. On the other hand, if it is determined that the predetermined conditions are not met (a rule violation is detected), the information processing unit 2034 does not confirm the changes, but feeds back structured error information indicating the violation to the generating AI system 30, and returns to step S14 to regenerate a new structured transaction.

[0080] In step S18, the server 20 generates signaled data based on the data confirmed in step S17. The information processing unit 2034 of the server 20 divides the confirmed data into predetermined units (e.g., rows) and assigns an identification tag to each unit that includes unique identification information (e.g., a sequential number such as "003") and a verification value (e.g., a hash value such as "7b8e"). Here, the verification value is generated as a chain hash value calculated based on the verification values ​​of the units preceding the unit in question. This ensures that the divided units as a whole have an indivisible order and completeness.

[0081] In step S19, the generating AI system 30 transmits the signaled data. Specifically, the generating AI system 30 receives the signaled data transmitted from the transmission control unit 2032 of the server 20 and transmits the signaled data to the terminal device 10 as a communication relay unit. The generating AI system 30 performs the process of transferring the signaled data, which includes identification tags and verification values ​​(chained hash values) for each division unit generated by the information processing unit 2034 of the server 20, to the terminal device 10 via a logical verification channel constructed within its output text stream.

[0082] In step S20, the terminal device 10 presents the signaled data. Specifically, the presentation control unit 133 of the terminal device 10 displays the received signaled data on an output unit 105 such as a display.

[0083] In step S21, the terminal device 10 determines whether or not the identification tags included in the signaled data have been changed (integrity verification). Specifically, the presentation control unit 133 of the terminal device 10 physically extracts units of signaled data enclosed by the identification tags from the text immediately preceding the presentation in step S20. Next, the presentation control unit 133 checks each extracted unit for any omissions or inconsistencies in order using the identification information included in the identification tags, and for any content modifications using the chained hash value. Here, content modification refers to unintended summarization, omission, or other editing performed by the generation AI system 30 when outputting as a relay unit for communication. For example, if unintended summarization, omission, or other editing occurs, the chained hash value will lose consistency, making it possible to immediately detect the output of incomplete data by the generation AI system 30. If a change (inconsistency) in the identification tags is detected (if modification has occurred), the presentation control unit 133 discards the presented output as unreliable and returns to step S19 to request the generation AI system 30 to resend the signaled data. On the other hand, if no change in the identification tag is detected (i.e., there is no modification), the terminal device 10 confirms the data as a valid output and terminates the series of processes. It should be noted that the entity performing the integrity verification using the identification tag and the determination in the event of failure is not limited to the terminal device 10, but may also be the information processing unit 2034 of the server 20. For example, it is also possible to configure the server 20 to receive identification information of packets from the terminal device 10 and then determine whether the data is missing or modified.

[0084] In this way, the text correction quality assurance process according to this embodiment is completed.

[0085] <Screen example> In this embodiment, an example of a screen displayed on the terminal device 10 will be described.

[0086] Figure 11 shows an example of a first screen of a terminal device according to the embodiment of this disclosure. Figure 11 shows the input screen for change instructions presented to the terminal device 10 in step S11. This screen can be implemented, for example, as a text editor, an input panel for an IDE, or a chat interface.

[0087] As shown in Figure 11, the first example screen includes, for example, a first area 1411 and a second area 1412. The first area 1411 is an area where input instructions are presented to the user. Specifically, the first area 1411 displays instruction messages to prompt the user to take action, such as "Please enter the file you wish to change and the changes you wish to make." The second area 1412 is a text input area for the user to enter change instructions in natural language. When the user enters text such as "Change the value of TIMEOUT_SEC from 30 seconds to 40 seconds" into this area and presses the send button, a change request is sent from the terminal device 10 to the server 20 according to step S11.

[0088] Figure 12 is a diagram showing a second example screen of a terminal device according to the embodiment of this disclosure. Figure 12 shows the screen on which the final output is presented to the terminal device 10 after processing in step S20.

[0089] As shown in Figure 12, the second example screen includes, for example, a first area 1421, a second area 1422, a third area 1423, and a fourth area 1424. The first area 1421 is an area where input instructions from the server 20 are presented, similar to the first area 1411 in Figure 11. The second area 1422 is an area where the text of the change target and change instruction entered by the user in step S11 is presented as history. The third area 1423 is an area where the data (deliverable) that has been confirmed and signaled by the server 20 is presented. The fourth area 1424 is a text input area for the user to enter additional change instructions in natural language, similar to the second area 1412 in Figure 11.

[0090] As described above, the signaled data presented in the third area 1423 has a structure in which each line is surrounded by identification tags ('[SIG:…]' and '[END_SIG]'), and is assigned unique identification information and a verification value which is a chained hash value. For example, in the line '[SIG:7b8e|003] TIMEOUT_SEC = 40 [END_SIG]' in the third area 1423, "7b8e" in the identification tag is a hash value (verification value), indicating that the content of the line has been correctly reflected based on the user's modification instructions displayed in the second area 1422, and that the integrity of the data is guaranteed. If the presentation control unit 133 of the terminal device 10 determines in the integrity verification of step S21 that there has been no modification of the content, it may present only the formatted text with the identification tags removed in the third area 1423. On the other hand, if an inconsistency is detected, the presentation control unit 133 displays a warning message near the first region 1421 or the third region 1423 and performs control such as prompting a retransmission request via the fourth region 1424.

[0091] <Variation> In the above embodiment, a client-server configuration in which the terminal device 10, server 20, and generation AI system 30 are connected via a network 80 is shown as an example, but the disclosure is not limited thereto. For example, all functions according to this embodiment may operate on a single information processing device (e.g., a high-performance personal computer) in a standalone configuration. In this case, the functions of the terminal device 10 and server 20 are integrated into the same device, and the generation AI model used operates within the device or in a local network environment.

[0092] In the above embodiment, an example was shown in which the server 20 performs part of the processing from steps S12 to S18 and step S21, but the disclosure is not limited thereto. For example, if the terminal device 10 has sufficient processing capacity, the control unit 101 of the terminal device 10 may perform all or part of the processing as the execution unit. By adopting such an edge computing configuration, communication delays in the network 80 can be reduced, or security can be improved by not sending highly confidential file data to an external server.

[0093] In the above embodiment, the generative AI system 30 is shown as an external API service, but a private generative AI model built or operated independently within the organization may also be used. This further reduces the risk of information leakage and makes it possible to apply tuning that is specific to a particular programming language or internal regulations.

[0094] In the above embodiment, an example was described in which the server 20 deterministically verifies the suitability of the structured transaction and executes the changes to the target. However, the embodiment is not limited to this. For example, if the generation AI system 30 has an agent function that can execute external tools, the generation AI system 30 itself may directly identify the target and perform the change processing, and the server 20 or terminal device 10 may acquire the execution log and subsequently verify and determine whether predetermined conditions (such as quality assurance rules) are met.

[0095] Furthermore, the validation and modification of structured transactions may be performed by an application or script running on the terminal device 10 instead of the server 20. In this case, it becomes possible to perform quality assurance validation directly on the client side while balancing the load across the entire system without going through the server 20.

[0096] In the above embodiment, JSON was given as an example of the structured transaction format, but YAML, XML, or any other arbitrary structured data format defined by the user may also be used.

[0097] <Summary> As described above, according to this embodiment, the server 20 obtains the target to be changed and the change instruction from the user, generates a first prompt including these and a system prompt, inputs it to the generation AI system 30, and causes it to generate a structured transaction including specific information of the target to be changed and the change content. The server then verifies whether the generated structured transaction conforms to a predetermined structure, executes the change if it conforms, and confirms the change if the executed change satisfies predetermined conditions such as quality assurance rules. This eliminates the probabilistic and unreliable nature of the generation AI and allows only high-quality text corrections that have passed a deterministic verification process, thereby structurally eliminating erroneous corrections and tampering by the generation AI and ensuring the identity and reliability of the deliverables.

[0098] Furthermore, according to this embodiment, in the verification step, if there are multiple candidates that match specific information within the target to be changed, or if there are no matching candidates, the server 20 will stop executing the change. This prevents problems such as the generating AI correcting unintended parts based on ambiguous instructions, or the process failing because it cannot find the parts to be corrected.

[0099] Furthermore, according to this embodiment, in the verification step, if the server 20 detects a candidate within the target to be changed that matches specific information after excluding differences in spaces or line breaks, it stops the execution of the change. This eliminates the risk of accidentally changing unintended similar parts due to slight differences in format, and enables a more rigorous and secure change process.

[0100] Furthermore, according to this embodiment, the server 20 uses the predetermined condition that the changes to the target of the change satisfy quality assurance rules including prohibited elements, required elements, regular expression patterns, or syntax rules. This makes it possible to systematically enforce the organization's coding conventions, security policies, compliance requirements, etc., and prevents compliance violations caused by the content generated by the generation AI.

[0101] Furthermore, according to this embodiment, when the server 20 confirms a change, it divides the confirmed data into predetermined units, assigns an identification tag containing identification information and verification values ​​to each unit, and generates signaled data. This allows information for verifying the integrity of the final output itself to be embedded, thereby establishing a foundation for guaranteeing the integrity of the data in subsequent processing and along the transmission path.

[0102] Furthermore, according to this embodiment, the server 20 outputs the generated signal data via the generation AI system 30, and based on the output data, it verifies whether there are any missing units, inconsistent order, or alterations to the content using identification tags. This allows monitoring and detection of whether the generation AI performs any editing such as summarizing, omitting, or formatting at the final output stage, and ensures the identity between the verified output and the final output presented to the user.

[0103] Furthermore, according to this embodiment, in the step of verifying whether there are any missing units, inconsistent order, or alterations to the content using identification tags, the server 20 (or terminal device 10) extracts signaled data from the text output via the generation AI system 30. Then, based on the extracted signaled data, it verifies whether there are any missing units, inconsistent order, or alterations to the content using identification tags. As a result, even if the generation AI system 30 adds extraneous natural language around the signaled data, it is possible to accurately extract and verify only the necessary information, thereby improving the robustness of the system.

[0104] Furthermore, according to this embodiment, in the step of verifying whether the structured transaction conforms to a predetermined structure, the server 20 extracts the structured transaction from the text generated by the generation AI system 30. Then, it verifies whether the extracted structured transaction conforms to the predetermined structure. As a result, even if the generation AI system 30 outputs text in a format other than the one instructed, the server 20 can accurately extract the intended instruction portion from it and correctly execute the subsequent verification process, thereby increasing the generation AI system 30's resilience to unpredictable output.

[0105] Although several embodiments of this disclosure have been described above, these embodiments can be implemented in a variety of other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. For example, configurations and processes in one embodiment may be combined with configurations and processes in another embodiment, or a modification of one embodiment may be applied to another embodiment. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims and their equivalents.

[0106] (Note) The details described in each of the above embodiments are noted below.

[0107] (Note 1) A program to be executed on a computer equipped with a processor, The aforementioned processor, The steps include obtaining from the user the item to be changed and the change instructions regarding the changes to the item to be changed, A step of generating a first prompt that includes the acquired target for modification, the acquired change instruction, and a system prompt for generating a structured transaction that includes specific information about the target for modification and the content of the change to the target for modification. The steps include inputting the generated first prompt into the generating AI and causing it to generate the structured transaction including the specific information and the changes contained in the first prompt, A step of verifying whether the generated structured transaction conforms to a predetermined structure, The steps include: executing the changes to the target of the change when the structured transaction conforms to the predetermined structure; If the changes to the target of the modification that have been executed satisfy predetermined conditions, the step of confirming the changes to the target of the modification, A program that executes the command.

[0108] (Note 2) In the verification step described above, if there are multiple candidates that match the specified information within the target of the change, or if there are no matching candidates, the execution of the change is stopped. The program described in (Note 1).

[0109] (Note 3) In the verification step described above, if a candidate matching the specified information is detected within the target of modification, excluding differences in spaces or line breaks, the execution of the modification is stopped. The program described in either (Appendix 1) or (Appendix 2).

[0110] (Note 4) The aforementioned predetermined condition is that the change to the subject of the change satisfies quality assurance rules, including prohibited elements, required elements, regular expression patterns, or syntax rules. The program described in any of (Appendix 1) to (Appendix 3).

[0111] (Note 5) When the aforementioned changes are confirmed, the following steps are taken to generate signal data: Divide the confirmed data into predetermined units, assign an identification tag containing identification information and verification values ​​to each unit, The program described in any of (Appendix 1) to (Appendix 4) to be executed by the aforementioned processor.

[0112] (Note 6) The steps include: outputting the generated signal data via the generation AI; A step of verifying, based on the output signal data, whether there are any missing units, inconsistencies in order, or alterations in content using the identification tag, The program described in (Appendix 5) that causes the aforementioned processor to execute.

[0113] (Note 7) In the step of verifying whether there are any missing units, inconsistent order, or alterations to the content using the identification tag, the signaled data is extracted from the text output via the generating AI, and the identification tag is used to verify whether there are any missing units, inconsistent order, or alterations to the content based on the extracted signaled data. The program described in (Appendix 6).

[0114] (Note 8) In the step of verifying whether the structured transaction conforms to the predetermined structure, the structured transaction is extracted from the text generated by the generation AI, and the extracted structured transaction is verified to conform to the predetermined structure. The program described in any of (Appendix 1) through (Appendix 7).

[0115] (Note 9) A method to be performed on a computer equipped with a processor, wherein the processor performs all steps performed in any of the inventions described in (Appendix 1) to (Appendix 8).

[0116] (Note 10) An information processing apparatus comprising a processor, wherein the processor performs all steps performed in any of the inventions described in (Appendix 1) to (Appendix 8).

[0117] (Note 11) A system comprising means for performing all steps performed in any of the inventions described in (Appendix 1) to (Appendix 8). [Explanation of Symbols]

[0118] 1... System 10…Terminal device 101... Control Unit 102...Storage section 103... Communications Department 104...Input section 105...Output section 20... Server 201... Control Unit 202...Storage section 203... Communications Department 204… Input / Output Interface 30…Generating AI system 80…Network

Claims

1. A program to be executed on a computer equipped with a processor, The aforementioned processor, The steps include obtaining from the user the item to be changed and the change instructions regarding the changes to the item to be changed, A step of generating a first prompt that includes the acquired target for modification, the acquired change instruction, and a system prompt for generating a structured transaction that includes specific information about the target for modification and the content of the change to the target for modification. The steps include inputting the generated first prompt into the generating AI and causing it to generate the structured transaction including the specific information and the changes included in the first prompt, A step of verifying whether the generated structured transaction conforms to a predetermined structure, The steps include: executing the changes to the target of the change when the structured transaction conforms to the predetermined structure; If the changes to the target of the modification that have been executed satisfy predetermined conditions, the step of confirming the changes to the target of the modification, A program that executes the command.

2. In the step of verifying whether the structured transaction conforms to the predetermined structure, if there are multiple candidates that match the specific information within the target of modification, or if there are no matching candidates, the execution of the modification is stopped. The program according to claim 1.

3. In the step of verifying whether the structured transaction conforms to the predetermined structure, if a candidate matching the specific information is detected within the target of modification, excluding differences in spaces or line breaks, the execution of the modification is stopped. The program according to claim 1.

4. The aforementioned predetermined condition is that the change to the subject of the change satisfies quality assurance rules, including prohibited elements, required elements, regular expression patterns, or syntax rules. The program according to claim 1.

5. When the aforementioned changes are confirmed, the following steps are taken to generate signal data: Divide the confirmed data into predetermined units, assign an identification tag containing identification information and verification values ​​to each unit, The program according to claim 1, which is to be executed by the processor.

6. The steps include: outputting the generated signal data via the generating AI; A step of verifying, based on the output signal data, whether there are any missing units, inconsistencies in order, or alterations in content using the identification tag, The program according to claim 5, which causes the processor to execute the following.

7. In the step of verifying whether there are any missing units, inconsistent order, or alterations to the content using the identification tag, the signaled data is extracted from the text output via the generating AI, and the identification tag is used to verify whether there are any missing units, inconsistent order, or alterations to the content based on the extracted signaled data. The program according to claim 6.

8. In the step of verifying whether the structured transaction conforms to the predetermined structure, the structured transaction is extracted from the text generated by the generation AI, and the extracted structured transaction is verified to conform to the predetermined structure. The program according to claim 1.

9. A method performed on a computer having a processor, wherein the processor performs all steps performed in any of the inventions according to claims 1 to 8.

10. An information processing apparatus comprising a processor, wherein the processor performs all steps performed in the invention according to any one of claims 1 to 8.

11. A system comprising means for performing all steps performed in the invention according to any one of claims 1 to 8.

Images