Information processing method, program, and information processing system

Abstract

An information processing method, an information processing system, and a program for efficiently obtaining a desired processing result are provided. [Solution] The method includes an input information acquisition step S100 of acquiring input information to be input to a trained model, a version identifier generation step S104 of generating a version identifier for identifying the version of the input information, a storage step S106 of storing the input information and the version identifier corresponding to the input information in a repository, and update steps S108 to S114 of generating a new version identifier in response to an update of the input information in the repository and storing the updated input information and the corresponding version identifier in the repository.

Description

[Technical field]

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

[0002] Patent Document 1 discloses a device that uses a trained model. In this device, a secure component manages parameters and version information of the trained model, and the trained model is deployed based on input information input to the trained model, and processing is performed using the trained model. This allows processing using the trained model to be performed while being protected from external attacks. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2019-101807 A Summary of the Invention [Problem to be solved by the invention]

[0004] When performing information processing using a trained model, input information may be adjusted to obtain the desired processing results from the trained model. In order to adjust the input information efficiently, it is preferable to manage the change history.

[0005] SUMMARY OF THE PRESENT EMBODIMENT In consideration of the above, an object of the present invention is to efficiently obtain a desired processing result. [Means for solving the problem]

[0006] According to an information processing method of one embodiment, an information processing method is executed by an information processing device, and includes an input information acquisition step of acquiring input information to be input to a trained model, a version identifier generation step of generating a version identifier for identifying a version of the input information, a storage step of storing the input information and the version identifier corresponding to the input information in a repository, and an update step of, when the input information in the repository is updated, generating a new version identifier in response thereto and storing the updated input information and the version identifier corresponding thereto in the repository.

[0007] According to one embodiment of an information processing method, the information processing method is executed by an information processing device and includes an input information acquisition step of acquiring input information to be input to a trained model and tag information in the input information, and a storage step of storing the input information and the tag information in a repository.

[0008] According to a program of one embodiment, an information processing device is caused to execute an information processing method including an input information acquisition step of acquiring input information to be input to a trained model, a version identifier generation step of generating a version identifier for identifying a version of the input information, a storage step of storing the input information and the version identifier corresponding to the input information in a repository, and an update step of, when the input information in the repository is updated, generating a new version identifier in response to the update and storing the updated input information and the version identifier corresponding to the input information in the repository.

[0009] According to a program of one embodiment, an information processing device is caused to execute an information processing method including an input information acquisition step of acquiring input information to be input to a trained model, an input information acquisition step of acquiring input information to be input to the trained model and tag information in the input information, and a storage step of storing the input information and the tag information in a repository.

[0010] According to one embodiment, the information processing system is executed by an information processing device, and includes an input information acquisition unit that acquires input information to be input to a trained model, a version identifier generation unit that generates a version identifier for identifying a version of the input information, a storage unit that stores the input information and the version identifier corresponding to the input information in a repository, and an update unit that, when the input information in the repository is updated, generates a new version identifier in response to the update and stores the updated input information and the version identifier corresponding to the input information in the repository.

[0011] According to one embodiment, the information processing system is executed by an information processing device, and includes an input information acquisition unit that acquires input information to be input to a trained model, an input information acquisition unit that acquires input information to be input to the trained model and tag information in the input information, and a storage unit that stores the input information and the tag information in a repository. Effect of the Invention

[0012] According to one embodiment, a desired processing result can be obtained efficiently. [Brief description of the drawings]

[0013] [Figure 1] 1 is a diagram illustrating an example of a configuration of an information processing system according to a first embodiment. [Diagram 2] FIG. 2 is a diagram illustrating an example of a hardware configuration of a server according to the first embodiment. [Diagram 3] FIG. 2 is a diagram illustrating an example of a functional configuration of a server according to the first embodiment. [Figure 4] FIG. 2 is a diagram showing an example of a processing flow of the information processing system according to the first embodiment. [Diagram 5] FIG. 11 is a diagram illustrating an example of a functional configuration of a server according to the second embodiment. [Figure 6] FIG. 11 is a diagram illustrating an example of a processing flow of an information processing system according to a second embodiment. [Figure 7] FIG. 13 is a diagram illustrating an example of a functional configuration of a server according to the third embodiment. [Figure 8] FIG. 11 is a diagram illustrating an example of a processing flow of an information processing system according to a third embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] First embodiment A first embodiment of an information processing system according to the present invention will be described below with reference to Figures 1 to 4. In each drawing, the same or equivalent components and parts are given the same reference numerals. Also, the dimensional ratios in the drawings are exaggerated for the convenience of explanation and may differ from the actual ratios.

[0015] (System Overview) First, an overview of an information processing system 10 according to the present embodiment will be described. The information processing system 10 according to the present embodiment is a system for managing input information to be input to a trained model. In the present embodiment, as an example, this input information is a prompt to be input to a large-scale language model as a trained model. This prompt is based on data such as natural language, video images, and audio. The prompt also includes so-called templates, at least a part of which can be changed or added.

[0016] (System Configuration) Fig. 1 is a diagram showing an example of a configuration of an information processing system 10 according to the present embodiment. As shown in Fig. 1, the information processing system 10 according to the present embodiment includes a server 12 as an information processing device and a user terminal 14 as an information processing device, which are communicably connected to each other via a network N. The network N is, for example, a wired LAN (Local Area Network), a wireless LAN, the Internet, a public line network, a mobile data communication network, or a combination of these.

[0017] The user terminal 14 is an example of an information processing device that is operated by a user U to input and display various information. The user terminal 14 may be a PC (Personal Computer), a smartphone, a tablet terminal, a server device, a microcomputer, a wearable device, or a combination of these.

[0018] The server 12 is an example of an information processing device that acquires information input from the user terminal 14, processes the information, and outputs the results. The server 12 may be a PC (Personal Computer), a smartphone, a tablet terminal, a server device, a microcomputer, or a combination of these. The specific configuration and operation of the server 12 will be described later.

[0019] (Hardware configuration) 2 is a block diagram showing a hardware configuration of the server 12. The server 12 includes a processor 120, a memory 122, a storage 124, a communication I / F 126, an input / output I / F 128, and a drive device 134, which are communicatively connected to each other via a bus B.

[0020] The processor 120 controls each component of the server 12 and realizes the functions of the server 12 by expanding various programs stored in the storage 124 into the memory 122 and executing them. The programs executed by the processor 120 include, but are not limited to, an OS (Operating System) and a program 220 described below. The execution of these programs by the processor 120 realizes a part of the state visualization method according to this embodiment. The processor 120 is, for example, a CPU (Central Processing Unit), an MPU (Micro Processing Unit), a GPU (Graphics Processing Unit), an ASIC (Application Specific Integrated Circuit), a DSP (Digital Signal Processor), or a combination of these.

[0021] The memory 122 is, for example, a Read Only Memory (ROM), a Random Access Memory (RAM), or a combination thereof. The ROM is, for example, a Programmable ROM (PROM), an Erasable Programmable ROM (EPROM), an Electrically Erasable Programmable ROM (EEPROM), or a combination thereof. The RAM is, for example, a Dynamic RAM (DRAM), a Static RAM (SRAM), a Magnetoresistive RAM (MRAM), or a combination thereof.

[0022] The storage 124 stores an OS, various programs described below, and various data. The storage 124 is, for example, a flash memory, a hard disk drive (HDD), a solid state drive (SSD), a storage class memory (SCM), or a combination of these.

[0023] The communication I / F 126 is an interface for connecting the server 12 to an external device including the user terminal 14 via the network N and controlling communication. The communication I / F 126 is, for example, an adapter compliant with Bluetooth (registered trademark), Wi-Fi (registered trademark), ZigBee (registered trademark), Ethernet (registered trademark), or optical communication (for example, Fibre Channel), but is not limited thereto.

[0024] The input / output I / F 128 is an interface for connecting an input device 132 and an output device 130 to the server 12. The input device 132 is, for example, a mouse, a keyboard, a touch panel, a microphone, a scanner, a camera, various sensors, an operation button, or a combination of these. The output device 130 is, for example, a display, a projector, a printer, a speaker, a vibrator, or a combination of these.

[0025] The drive device 134 reads and writes data from and to the disk media 136. The drive device 134 is, for example, a magnetic disk drive, an optical disk drive, a magneto-optical disk drive, or a combination of these. The disk media 136 is, for example, a compact disc (CD), a digital versatile disc (DVD), a floppy disk (FD), a magneto-optical disk (MO), a Blu-ray (registered trademark) disc (BD), or a combination of these.

[0026] In this embodiment, the program may be written into memory 122 or storage 124 during the manufacturing stage of server 12, or may be provided to server 12 via network N, or may be provided to server 12 via a non-transitory computer-readable recording medium such as disk media 136.

[0027] Moreover, the hardware configuration of the user terminal 14 is substantially the same as the hardware configuration of the server 12 described above, and therefore a detailed description thereof will be omitted.

[0028] (Functional configuration) Next, the functional configuration of the server 12 will be described. Fig. 3 is a diagram showing an example of the functional configuration of the server 12. When executing various programs, the server 12 realizes various functions using the above-mentioned hardware resources. The server 12 has a communication unit 20, a storage unit 22, and a control unit 24 as the functional configuration realized by the server 12. Each functional configuration is realized by the processor 120 reading and executing a program 220 stored in the memory 122 or the storage 124.

[0029] The communication unit 20 is realized by the communication I / F 126. The communication unit 20 transmits and receives information to and from the user terminal 14 via the network N. The communication unit 20 receives information input from the user terminal 14. The communication unit 20 also transmits information to the user terminal 14 and receives a request by the user U from the user terminal 14. Furthermore, the communication unit 20 transmits and receives information to and from an external system using an API or the like.

[0030] The storage unit 22 is realized by the memory 122 and the storage 124. In the storage unit 22, a program 220, a trained model 222, a prompt DB 224, a user information DB 226, and a processing result DB 228 are stored.

[0031] The trained model 222 is composed of at least one trained machine learning model. As described above, the trained model 222 is, for example, a large-scale language model, which is trained with a large amount of text data from articles, books, websites, etc. on the Internet. When text data called a prompt is input as input information, information processing is performed according to the prompt, and various data including the text data are generated as a result, and the data is output as output information. Note that the trained models 222 may be stored not in the storage unit 22 but in an external server or the like that can communicate with the server.

[0032] In addition, the trained model 222 generates additional information for the prompt template, which will be described later, as necessary. That is, when the contents of the prompt template are updated and the output information of the prompt template before and after the update is compared and evaluated to evaluate the prompt template itself, additional information according to the template is generated. As an example, if the prompt template is "Please translate as follows: original language:, translated language:, tone of the translated text (business-oriented, casual, easy for general users, etc.):, text to be translated:", placeholders are set for "original language", "translated language", "tone of the translated text (business-oriented, casual, easy for general users, etc.)", and "text to be translated" in this template. The placeholders are temporary spaces or marks provided for the user to input specific information later. The trained model 222 generates information corresponding to the placeholders as additional information. In the above example, for example, additional information such as "English" for the "original language", "Japanese" for the "translated language", "business-oriented" for the "tone of the translated text (business-oriented, casual, easy for general users, etc.)", and "A picture is worth a thousand words" for the "text to be translated" are generated. By adding this additional information to the template and inputting it to the trained model 222 as a prompt, output information can be obtained.

[0033] Furthermore, the trained model 222 compares how the output information changes between the prompt (based on the template) before the content is updated and the prompt (based on the template) after the content is updated, and evaluates whether the output information is in line with the intent of the prompt. In other words, it can be determined whether updating the template of the prompt will result in the desired output information. This comparative evaluation is performed by inputting the output information to the trained model 222 together with a prompt that instructs the trained model to perform the evaluation, but is not limited to this, and the evaluation may be performed by a trained model that is connected to an external communication system. This allows evaluation from different perspectives and objective evaluation.

[0034] As described above, the input information is a prompt that instructs the trained model 222 to perform processing, and is text data in a natural language. This input information is input by the user U, but is not limited to this and may be automatically generated or modified by the system.

[0035] The prompt DB 224 is a database that stores information indicating prompts input by the user U. Prompts are written in natural language, and the prompt DB 224 stores, as an example, a prompt associated with account information (described below) at the time the prompt was input. The prompt DB 224 also stores not only all prompts but also at least some of the prompts that leave room for inputting additional information. At least some of the prompts correspond to prompt templates. A template in which additional input information has been input corresponds to a prompt.

[0036] Version identifiers corresponding to prompts (including templates, the same applies below) are also stored in the prompt DB 224. The version identifier is any text data for uniquely identifying the version of the prompt, and when the contents of the prompt are updated, a new version identifier (described in detail later) is generated accordingly and is also stored in the prompt DB 224.

[0037] The user information DB 226 is a database that stores various information of the user U and the manager M who have registered to use the information processing system 10. Examples of the various information stored include user identification information for unique identification, user preference information such as interest information, user history information including past usage status, and user attribute information such as occupation, position, and authority.

[0038] The processing result DB228 is a database that stores information indicating the results of executing a process based on a prompt (also referred to as the processing execution result or the processing result). The processing result is various information including text information and image information, and the processing result DB228 stores the processing result, the prompt, and user information linked to each other. The processing result DB228 also stores at least one of evaluation and comparison of output information obtained by inputting prompts corresponding to multiple different version identifiers to the trained model 222.

[0039] The control unit 24 is realized by the processor 120 reading and executing the program 220 from the memory 122 (see FIG. 2) and cooperating with other hardware configurations. The control unit 24 includes an information acquisition unit 242, an information processing unit 244, and an output unit 246.

[0040] The information acquisition unit 242 acquires various information required for processing in the information processing unit 244. Specifically, the information acquisition unit 242 acquires prompts stored in a storage unit including the memory unit 22 and a database connected to an external communication-enabled device, and prompts input by the user U. Note that the various prompts acquired are not limited to text data, and may utilize information from various sensors such as voice input, gestures, hardware operations, eye tracking, and biosensors.

[0041] The information acquisition unit 242 also acquires various information for executing a search for a prompt stored in the storage unit 22 or an external database connected to enable communication. That is, the information acquisition unit 242 acquires search execution instruction information and search condition information as various information from the user terminal 14 or an external system connected to enable communication using an API or the like. This search condition information is based on metadata such as keywords and tags in natural language including the contents of the prompt, version identifiers, branch identifiers, branch names (described in detail below), creators, and creation dates and times. That is, the information acquisition unit 242 corresponds to the "input information acquisition unit" recited in claims 15 and 16. The search condition information can also be rephrased as search keywords, search terms, search phrases, queries, and search input information.

[0042] Based on the prompt, the information processing unit 244 performs information processing using the trained model 222. The trained model 222 outputs an answer or a generated result corresponding to the prompt based on knowledge accumulated therein.

[0043] Furthermore, the information processing unit 244 constantly monitors the acquired prompt and determines whether the contents of the prompt match an existing prompt stored in the storage unit. For example, this determination is performed by generating a hash value or a checksum of the prompt, comparing it with existing prompt data, and treating it as a new prompt if a matching prompt does not exist. In addition, the match determination may be performed by keyword matching, natural language processing using the trained model 222, using feedback information from the user or the system, or the like. If an identical prompt does not exist, the information processing unit 244 generates a unique version identifier for the prompt as a new prompt. This version identifier is used to identify the prompt, and for example, an identifier that guarantees uniqueness is generated using the current date and time or a random value. Furthermore, the information processing unit 244 stores the new prompt together with the generated version identifier in the storage unit. Note that the information processing unit 244 is also capable of generating a version identifier for output information from the trained model 222.

[0044] Furthermore, the information processing unit 244 generates additional information corresponding to a placeholder in a template of a prompt as necessary. Specifically, the information processing unit 244 extracts a placeholder of the template and generates additional information to be inserted into the placeholder. The additional information includes dynamically generated data (for example, the latest information acquired from outside) and static data (fixed information), and is generated according to a data format (for example, "NAME", "DATE", "TIME", etc.) specified for the placeholder. Note that a template and an instruction to generate additional information corresponding to the placeholder may be input as a prompt to the trained model 222, and the additional information may be obtained using the output information. Furthermore, the information processing unit 244 may request the user U to input additional information as necessary. This is particularly effective in an interactive system in which the system and the user U converse in a chat format. Then, by inserting the generated additional information into the placeholder, a more specific and customized prompt is formed.

[0045] Furthermore, the information processing unit 244 compares and evaluates output results based on prompts of different versions. This comparative evaluation makes it possible to clarify the differences in performance and usefulness between the versions.

[0046] Furthermore, the information processing unit 244 generates and executes a query for searching for prompts in the storage unit based on execution instruction information and search condition information acquired from the user U or the system. The search results are provided to the user U through the user terminal 14 as a list of corresponding prompts or metadata. Furthermore, the information processing unit 244 creates a branch in response to an operation by the user U for a prompt corresponding to a specific version identifier, and the branch information can also be stored in the prompt DB 224. That is, by referring to a specific (original) version identifier, new input information based on the branch creation operation by the user U is created, and a new version identifier for it is generated. At least one of a branch identifier and a branch name is generated as branch information by associating the generated version identifier with the original version identifier. This information is stored in the prompt DB 224. That is, the information processing unit 244 corresponds to the "version identifier generating unit", the "storage executing unit", and the "update unit" recited in claims 15 and 16.

[0047] As described above, the information processing unit 244 manages input information, output information, and version identifiers such as prompts and templates, and stores the input information and version identifiers in the prompt DB 224 and the output information in the processing result DB 228. Therefore, the prompt DB 224, the processing result DB 228, and the information processing unit 244 work together to function as a repository.

[0048] The output unit 246 controls the information processing unit 244 so that the processing result is output to the user terminal 14 .

[0049] (Processing Executed by Information Processing System 10) Next, the operation of the information processing system 10 will be described. Fig. 4 is a flow chart showing an example of the flow of processing by the information processing system 10. The processor 120 reads out the program 220 stored in the storage 124, expands it in the memory 122, and executes it, thereby performing processing. Although not shown, when the processor 120 receives operation information for terminating the operation of the information processing system 10, or information on the end of an operation from the user terminal 14 during the determination processing being executed (these are simply referred to as "termination operation"), the processor 120 ends the processing based on the program 220 being processed.

[0050] The processor 120 judges whether or not input information has been acquired (step S100). If input information has not been acquired (step S100: NO), the processor 120 shifts the process to step S130 described later. On the other hand, if input information has been acquired (step S100: YES), the processor 120 judges whether the content of the input information matches existing input information stored in the storage unit (step S102). If they match, that is, if the input information is not new (step S102: YES), the processor 120 shifts the process to step S108 described later. On the other hand, if they do not match, that is, if the input information is new (step S102: NO), the processor 120 generates a version identifier for the input information (step S104) and stores the input information and the version identifier in the storage unit (step S106). The process of step S100 described above corresponds to the "input information acquisition step" recited in claim 1. Also, the process of step S104 corresponds to the "version identifier generation step" recited in claim 1. Furthermore, the process of step S106 corresponds to the "storing step" recited in claim 1.

[0051] The processor 120 monitors whether the input information has been updated (step S108) and determines whether the input information has been updated (step S110). If there has been no update (step S110: NO), the processor 120 proceeds to step S130, which will be described later. On the other hand, if there has been an update (step S110: YES), the processor 120 generates a version identifier for the updated input information (step S112) and stores the updated input information and the version identifier in the storage unit (step S114). The processes from step S108 to step S114 correspond to the "update step" recited in claim 1.

[0052] Processor 120 determines whether or not additional information is required for the input information (i.e., whether or not the input information is a template having a placeholder) (step S116). If additional information is not required (step S116: NO), processor 120 proceeds to step S120, which will be described later. On the other hand, if additional information is required (step S116: YES), processor 120 generates additional information (step S118).

[0053] The processor 120 inputs input information to the trained model 222 (step S120), and acquires output information from the trained model 222 (step S122). Then, the processor 120 generates a version identifier for the acquired output information (step S124).

[0054] The processor 120 compares the output information corresponding to the different versions of the input information (step S126). For the different versions, the latest version may be compared with the immediately previous version, or any different versions may be compared. The processor 120 evaluates the output information and thus the input information based on the comparison result (step S128), and stores the evaluation result in the storage unit in association with the input information (step S130).

[0055] The processor 120 judges whether or not it has acquired search execution instruction information and search condition information (step S132). If it has not been acquired (step S132: NO), the processor 120 ends the processing based on the program 220. On the other hand, if it has been acquired (step S132: YES), the processor 120 executes the search (step S134), outputs the search result to be displayed on the user terminal 14 (step S136), and ends the processing based on the program 220. The processing from step S132 to step S136 corresponds to the "acquisition step" of claim 2. The processing from step S132 to step S134 corresponds to the "search step" of claim 11. Although not shown, when the user U performs a branch creation operation based on the search result, a branch identifier and a branch name are generated according to the operation and stored in the prompt DB 224.

[0056] (Effects of the First Embodiment) According to the information processing system 10 of this embodiment, the following steps are executed: an input information acquisition step of acquiring input information to be input to the trained model 222; a version identifier generation step of generating a version identifier for identifying the version of the input information; a storage step of storing the input information and the version identifier corresponding to the input information in the repository; and an update step of generating a new version identifier in response to an update of the input information in the repository and storing the updated input information and the version identifier corresponding to the input information in the repository. Therefore, a unique version identifier is generated by the version identifier generation step for each input information acquired by the input information acquisition step. This makes it possible to identify different versions of input information, and the change history of each version is clearly managed. Furthermore, the storage step stores the acquired input information and its version identifier in the repository. This makes it possible to distinguish and track past versions and the current version, since the input information is saved for each version. This history management makes it easy to identify which version of input information was used for the trained model. Furthermore, each time the input information in the repository is updated, a new version identifier is generated by the update step, and is stored in the repository together with the updated input information. As a result, the adjustment history of the input information is systematically managed, so that the input information can be changed and adjusted efficiently. Based on this history, it is possible to refer to the past changes and quickly generate an optimal prompt. Based on the input information and its version history managed in this way, it is possible to select an appropriate version of the input information and supply it to the trained model 222. As a result, it is possible to efficiently obtain the desired processing result. In addition, by comparing and evaluating the processing results based on the version-managed information, the generation and adjustment of optimal prompts is promoted. This makes it possible to maximize the effect of information processing by the trained model.

[0057] Furthermore, the steps executed by the information processing system 10 include an acquisition step of acquiring input information corresponding to a specified version identifier from the repository when the specified version identifier is specified, so that input information of the specified version can be accurately acquired from the repository by a user or another system designating a desired version identifier. This makes it possible to easily track and refer to the history and changes of a specific version.

[0058] Furthermore, since the information processing system 10 performs at least one of comparing and evaluating the results of inputting a plurality of pieces of input information with different version identifiers into the trained model 222, the performance of each prompt version is compared and evaluated by inputting a plurality of pieces of input information with different version identifiers into the trained model. As a result, it is possible to identify which version generates the most accurate result and obtain feedback for improving the accuracy of the model. In addition, by comparing a plurality of input information versions, it is possible to find the optimal prompt form and standardize the prompt. As a result, it is possible to efficiently generate the optimal prompt and quickly achieve the desired processing result. Furthermore, in the information processing system 10, it is possible to evaluate the diversity of data and the robustness of the model through a comparison of the results using different versions of input information. It is possible to confirm the operation of the trained model 222 under various conditions and obtain clues for improving the fault tolerance and reliability of the entire system. Furthermore, by comparing the histories of multiple versions, it is possible to perform a detailed analysis of how the input information of each version affected the processing result. As a result, it is possible to identify the improvements and the effects of the changes in each version and take further improvement measures. In addition, by comparing the results using the input information for each version identifier, it is possible to use the results as material for tuning and re-learning the trained model 222. This improves the performance and realizes stable operation of the trained model 222. Furthermore, using the results of the comparative evaluation, it is possible to select and provide the most useful prompt version for the user U. This is expected to optimize the user experience and improve satisfaction. Furthermore, by automating the comparative evaluation between versions, a series of processes from generating prompts to evaluation, feedback, and adaptation can be efficiently executed. This allows the system to always select and utilize the latest optimal prompt. This feedback loop allows the quality of the prompts to be continuously improved, and the performance of the trained model 222 to be optimized.

[0059] In addition, the storage step in the information processing system 10 clarifies the relationship between the input information and the output information, since the output information from the trained model based on the acquired input information is linked to the input information and stored in the repository. This makes it easy to track and understand which input information has brought about which output information. Furthermore, since the input information and the corresponding output information are stored together, they can be quickly and efficiently referenced and reused as needed. This makes it easy to perform new analysis and improvements based on past processing results. Furthermore, by storing the output information linked to the input information in the repository, searches based on specific output results can be easily performed. This makes it possible to quickly obtain input information that matches specific conditions and its output results.

[0060] Furthermore, the information processing system 10 generates additional information necessary to obtain output information from the trained model 222 according to the input information, and acquires the input information to which the additional information has been added. Therefore, even if the input information is in the form of a template having a placeholder, it is possible to determine whether the template can obtain the desired processing result by adding the generated additional information to the placeholder of the template and inputting the template to the trained model 222. In other words, the input information including the template can be efficiently evaluated and improved. Based on the evaluation results for each version of the input information, the improvement points of the new input information become clear, and a continuous improvement cycle can be formed.

[0061] Furthermore, a version identifier corresponding to the output information is generated and stored in the repository, making each output information uniquely identifiable, thereby making it easy to track which output information was generated for which input information.

[0062] In addition, since the method includes an output step of outputting at least one of the input information and the output information in a viewable manner, it becomes possible to compare and evaluate the output results of the trained model for the input information. This makes it possible to select optimal prompts and improve the response accuracy of the trained model and the entire system.

[0063] Furthermore, by storing branch information corresponding to the input information in the repository, it is possible to evaluate multiple different revisions based on each version identifier and to deploy the input information based on various scenarios, thereby more efficiently achieving the desired processing results.

[0064] Furthermore, since the method includes a search step in which the input information is searched for using search condition information in natural language, the user U can input his / her own questions or requests in the original form, and can proceed with the search without stress. This can improve the satisfaction of the user U.

[0065] In order to search the input information using search condition information in natural language, pre-structured metadata or the like may be added to the input information. As a specific example, context information such as "intention," "theme," and "classification" is added to the input information as metadata, either manually by a user U or automatically by the trained model 222, and then vectorized. Then, the search condition information input in the search step is also vectorized to obtain input information with a high degree of similarity between the vectors. This can further improve the search accuracy.

[0066] Second Embodiment Next, an information processing system 500 according to a second embodiment of the present invention will be described with reference to Figures 5 and 6. The information processing system 500 according to the second embodiment has a basic configuration similar to that of the first embodiment, and is characterized in that tag information in input information is acquired and stored in a storage unit. Note that the same components as those in the first embodiment are denoted by the same reference numerals and their description will be omitted.

[0067] (Functional configuration) The functional configuration of the server 70 in the information processing system 500 will be described. Fig. 5 is a diagram showing an example of the functional configuration of the server 70. When executing various programs, the server 70 realizes various functions using hardware resources similar to those of the server 12 of the first embodiment. The server 70 has a communication unit 20, a storage unit 72, and a control unit 74 as the functional configuration realized by the server 70. Each functional configuration is realized by the processor 120 reading and executing a program 720 stored in the memory 122 or the storage 124.

[0068] The storage unit 72 is realized by the memory 122 and the storage 124. In the storage unit 72, a program 720, a trained model 222, a prompt DB 724, a user information DB 226, and a processing result DB 228 are stored.

[0069] The prompt DB 724 stores information indicating the prompt input by the user U, a template, and a version identifier, similar to the prompt DB 224 in the first embodiment. The prompt DB 724 also stores tag information, which will be described later.

[0070] The control unit 74 is realized by the processor 120 reading and executing the program 720 from the memory 122 (see FIG. 2) and cooperating with other hardware configurations. The control unit 74 includes an information acquisition unit 742, an information processing unit 744, and an output unit 246.

[0071] The information acquisition unit 742, like the information acquisition unit 242 in the first embodiment, acquires various information required for processing in the information processing unit 744 and various information for searching for prompts. The information acquisition unit 742 also acquires tag information. The tag information acquired by the information acquisition unit 742 is tag information directly input by the user U. For example, when the input information is related to technology, the information acquisition unit 742 acquires tag information directly input by the user U, such as "technology". The format in which the user U directly inputs tag information includes direct input of keywords, selection of predefined information, and the like. The information acquisition unit 742 also acquires tag information generated by the information processing unit 744 (details will be described later) in addition to the tag information directly input by the user U. That is, the information acquisition unit 742 corresponds to the "input information acquisition unit" and "tag information acquisition unit" of claim 16.

[0072] The information processing unit 744, like the information processing unit 244 in the first embodiment, processes information using the trained model 222 based on the prompt, constantly monitors the acquired prompt, generates additional information corresponding to the placeholder in the template of the prompt, compares output results based on different versions of the prompt to perform evaluation, and searches for the prompt in the storage unit. The information processing unit 744 also generates tag information and stores it in the storage unit. The tag information generated by the information processing unit 744 includes information generated by extracting keywords from the input information, information generated using the trained model 222, and information generated by using an external tagging service through API integration or the like. Specific examples of tag information include the following categories. First, tags based on the subject or theme of the input information include "technology," "business," "health," "education," and "entertainment." Furthermore, tags reflecting the intention or purpose of the user U in the input information include "question," "suggestion," "problem solving," "information search," and "feedback." Next, tags based on the emotion or tone of the input information include "positive," "negative," "neutral," "question," and "gratitude." In addition, tags based on the specific use of the input information include "marketing," "product reviews," "customer support," "educational materials," and "project management." In addition, tags based on specific keywords and phrases included in the input information can be used. Examples include "machine learning," "API," "healthcare," "blockchain," and "leadership." In addition, tags based on the user demographic or market that the input information targets can be used, such as "for beginners," "for experts," "for students," "for businesses," and "for the general public." In addition, tags when the input information is related to a specific region or location can be used, such as "Japan," "America," "Tokyo," "New York," and "regional information." In addition, tags when the input information is related to a specific time or period can be used, such as "2023," "this week," "next month," "annual events," and "history." These tag information is generated and managed appropriately according to the content and purpose of the input information. Specifically, as an example, "Please explain the latest AI technology related to technology.For input information such as "Technology," "AI," "Education," and "For experts" are generated as tags. In this way, by using tag information, it becomes easier to search and classify input information, and the efficiency of the information processing system is improved. In other words, the information processing unit 744 corresponds to the "storage execution unit" recited in claim 16.

[0073] (Processing performed by information processing system 500) Next, the operation of the information processing system 500 will be described. Fig. 6 is a flowchart showing an example of the flow of processing by the information processing system 500. The processor 120 reads out the program 720 stored in the storage 124, expands it in the memory 122, and executes it, thereby performing processing. Note that the same processes as those in the first embodiment are denoted by the same reference numerals, and the description thereof will be omitted.

[0074] If the input information is new at step S102 (step S102: NO), the processor 120 judges whether or not tag information has been input by the user U through the user terminal 14 (step S200). If the information has been input (step S200: YES), the processor 120 shifts the process to step S104, and stores the tag information in the storage unit at step S106. On the other hand, if the information has not been input (step S200: NO), the processor 120 generates tag information based on the input information (step S202), and then shifts the process to step S104, and stores the tag information in the storage unit at step S106. The process of step S202 corresponds to the "tag generation step" recited in claim 10.

[0075] (Effects of the second embodiment) The information processing system 500 according to this embodiment has the same configuration as the first embodiment, except that tag information in the input information is acquired and stored in the storage unit, and the like, so that the same action and effect as the first embodiment can be obtained. In addition, the information processing system 500 has an input information acquisition step of acquiring input information to be input to a trained model and tag information in the input information, and a storage step of storing the input information and the tag information in a repository, so that by using the tag information stored in the repository, it is possible to quickly search and filter input information that matches a specific condition. This allows the required data to be acquired efficiently, and analysis and processing can be performed smoothly. In addition, by utilizing the tag information, the user can set appropriate tags based on his / her own input information, and easily search and refer to information later. This improves the ease of use of the system and improves the user experience.

[0076] Furthermore, since the information processing system 500 includes a tag generation step of generating tag information based on input information, manual tagging work is significantly reduced, enabling efficient data processing. This reduces the burden on the user U and the administrator, and improves the operational efficiency of the entire system. Furthermore, automatic tag generation enables consistent tagging. This unifies tagging standards even for different users and time periods, improving the consistency and accuracy of data.

[0077] In the above-described embodiment, the configuration includes a tag generation step in which tag information is generated based on input information, but the configuration is not limited to this and may be such that tag information is not generated and only tag information input directly by user U or tag information from other systems, etc. is acquired.

[0078] Third embodiment Next, an information processing system 600 according to a third embodiment of the present invention will be described with reference to Figures 7 and 8. The information processing system 600 according to the third embodiment has a basic configuration similar to that of the first embodiment, and is characterized in that it acquires feedback information for input information. Note that the same components as those in the first embodiment are denoted by the same reference numerals and their description will be omitted.

[0079] (Functional configuration) The functional configuration of the server 80 in the information processing system 600 will be described. Fig. 7 is a diagram showing an example of the functional configuration of the server 80. When executing various programs, the server 80 realizes various functions using hardware resources similar to those of the server 12 of the first embodiment. The server 80 has a communication unit 20, a storage unit 82, and a control unit 84 as the functional configuration realized by the server 80. Each functional configuration is realized by the processor 120 reading and executing a program 820 stored in the memory 122 or the storage 124.

[0080] The storage unit 82 is realized by the memory 122 and the storage 124. In the storage unit 82, a program 820, a trained model 222, a prompt DB 824, a user information DB 226, and a processing result DB 228 are stored.

[0081] The prompt DB 824 stores information indicating the prompt input by the user U, a template, and a version identifier, similar to the prompt DB 224 in the first embodiment. The prompt DB 824 also stores feedback information, which will be described later.

[0082] The control unit 84 is realized by the processor 120 reading and executing the program 720 from the memory 122 (see FIG. 2) and cooperating with other hardware configurations. The control unit 84 includes an information acquisition unit 842, an information processing unit 844, and an output unit 246.

[0083] The information acquisition unit 842, like the information acquisition unit 242 in the first embodiment, acquires various information required for processing in the information processing unit 844 and various information for searching for prompts. The information acquisition unit 842 also acquires feedback information. That is, the information acquisition unit 842 corresponds to the "feedback information acquisition unit" of claim 19. The feedback information acquired by the information acquisition unit 842 is feedback information on the input information directly input by the user U, and an example of the feedback information is a quantitative evaluation of satisfaction and accuracy. A specific example of the evaluation is an evaluation score ranging from 1 (very dissatisfied) to 5 (very satisfied). Furthermore, the feedback information also includes comments expressing specific opinions and impressions on the input information and the processing results thereof. For example, comments such as "the explanation was easy to understand" and "I would like more detailed information" are included. Next, examples of feedback expressing simple positive and negative feedback include thumbs up (such as "like") or thumbs down. In addition, it is possible to additionally attach tag information as feedback, and tag information such as "useful", "highly relevant", and "specific" may be used. Furthermore, the time when the user used the input information and the results is also included in the feedback information. This allows the usefulness and involvement of the information to be measured. The usage time is recorded in the form of, for example, 5 minutes, 10 minutes, or 30 minutes. Furthermore, the result selected by the user U from the multiple results provided can also be used as feedback information. Also, reports of errors and bugs related to the input information and results are important, and error reports such as "link is broken" and "information is incorrect" are also included in the feedback information. In addition, the log of what actions the user U took is also feedback information. The action log includes information such as "clicked on a link" and "downloaded". Furthermore, the satisfaction index indicating the satisfaction with a specific item is also included in the feedback information. For example, satisfaction with "response speed", "accuracy of information", and "ease of use" are included. In addition to this, requests for additional information or follow-up from the user U are also feedback information. For example, follow-up requests such as "please tell me the next step" and "I would like similar information" are included.

[0084] The information processing unit 844, like the information processing unit 244 of the first embodiment, processes information using the trained model 222 based on the prompt, constantly monitors the acquired prompt, generates additional information corresponding to the placeholder in the template of the prompt, compares output results based on different versions of the prompt to perform evaluation, and searches for the prompt in the storage unit. In addition, the information processing unit 844 generates at least one of a correction proposal and a correction idea for the input information as necessary based on the feedback information. That is, as a specific example, if the input information includes a prompt "Please explain in detail the latest trends in AI technology," and the feedback information includes a comment "The processing took a long time. It may be better to reduce the number of specific examples," a correction proposal of "Simplify the detailed explanation and reduce the number of specific examples" is generated. This correction proposal can be "Please give an overview of the latest trends in AI technology."

[0085] Next, an example will be described in which an evaluation score is provided as feedback information. When the input information includes a prompt saying "Please explain in detail the latest trends in AI technology," and the feedback information includes an evaluation score of 2, it is estimated from the score that the user U is dissatisfied, and an example of a correction suggestion is generated, saying "Simplify the content of the prompt to make it easier to understand." This correction suggestion can be "Please give an overview of the latest trends in AI technology."

[0086] Next, an example is given when thumbs up / thumbs down information is provided as feedback information. When the input information is a prompt "Please tell me the basics of marketing strategy," and thumbs down information is provided as feedback information, it is inferred from the feedback information that the user U is dissatisfied, and an example of a correction proposal is generated as "Make the content of the prompt more specific and add practical elements." This correction proposal can be "Please tell me the basics of marketing strategy and some specific examples."

[0087] Next, an example is given when tag information is provided as feedback information. When the input information is a prompt saying "Please explain in detail the optimization method of deep learning algorithms," and the feedback information is tagged with the tag "difficult," an example of a correction suggestion is generated, which is "Simplify difficult terms and add easy-to-understand explanations." This correction suggestion can be, "Please explain the optimization method of deep learning algorithms in a way that is easy for beginners to understand."

[0088] Next, an example is given in which the usage time is provided as feedback information. The input information is a prompt saying "Please explain why this problem could not be solved." The feedback information is "usage time (stay time) 2 minutes," and if this is judged to be short in this situation, it is estimated that a more detailed explanation is required, and an example of a correction suggestion is generated: "Add a more detailed explanation to encourage users to stay longer." This correction suggestion can be, "Could you please tell me more about how this problem occurred?"

[0089] Next, an example of a case where a selection of a result is provided as feedback information. When the input information is a prompt "Please tell me about the theory of economics." and a simple explanation is selected as feedback information, "Simplify the content for beginners" is generated as an example of a correction proposal. This correction proposal can be "Please explain the basic theory of economics in an easy-to-understand manner for beginners." Note that the above example is a case of a prompt, but the same applies to the case of a template. That is, when the prompt template is "Please explain the latest information about XXX (XXX corresponds to a placeholder) in detail. Please include five specific examples." and the feedback information is a comment "The processing took a long time. It may be better to reduce the number of specific examples.", an example of a correction proposal is generated as "Reduce the number of specific examples and emphasize the overview." This correction proposal can be "Please explain the latest information about XXX. Please include two specific examples." Note that in this embodiment, a configuration is adopted in which a correction proposal and a correction proposal are generated separately, but this is not limited to this, and at least one of the correction proposal and the correction proposal may be generated.

[0090] (Processing performed by information processing system 600) Next, the operation of the information processing system 600 will be described. Fig. 7 is a flowchart showing an example of the flow of processing by the information processing system 600. The processor 120 reads out a program 720 stored in the storage 124, expands it in the memory 122, and executes it to perform processing. Note that the same processes as those in the first embodiment are denoted by the same reference numerals, and the description thereof will be omitted.

[0091] After the process of step S136, the processor 120 judges whether or not feedback information has been acquired (step S300). If feedback information has not been acquired (step S300: NO), the processor 120 ends the process based on the program 820. On the other hand, if feedback information has been acquired (step S300: YES), the processor 120 analyzes the feedback information and judges whether or not at least one of a correction suggestion and a correction plan for the input information is necessary (step S302). If a correction suggestion or the like is not necessary (step S302: NO), the processor 120 ends the process based on the program 820. On the other hand, if a correction suggestion or the like is necessary (step S302: YES), the processor 120 generates at least one of a correction suggestion and a correction plan (step S304) and proceeds to step S300. The above-mentioned step S300 corresponds to the "feedback information acquisition step" recited in claim 11. Also, the above-mentioned step S302 corresponds to the "correction recommendation step" recited in claim 12.

[0092] (Effects of the Third Embodiment) The information processing system 600 according to this embodiment has the same configuration as the first embodiment, except for the fact that feedback information on input information is acquired, and therefore has the same operational effects as the first embodiment. Furthermore, since the information processing system 600 acquires feedback information on input information, it can identify which input information requires the most corrections and which input information is most effective. This makes it possible to most effectively allocate resources in updating input information and increase the efficiency of development and improvement.

[0093] Furthermore, the information processing system 600 executes a correction recommendation step of generating at least one of a correction suggestion and a correction proposal for the input information based on the feedback information, so that the user U does not need to find and correct the problem by himself / herself. This reduces the burden on the user U, and enables the user U to use the system comfortably. Furthermore, the input information can be continuously improved by the synergistic effect of the feedback information and the correction recommendation step.

[0094] In the above-described embodiment, the feedback information is analyzed and at least one of the correction proposal and the correction plan of the input information is generated as necessary. However, the present invention is not limited to this. When the feedback information is analyzed and the feedback information is good, the recommendation level of the input information corresponding to the feedback information may be increased or set as a standard for evaluation, and other operations may be performed other than the generation of the correction proposal and the correction plan. In addition, for input information for which good feedback information has been obtained, the output information corresponding to this input information may be controlled to be output at a predetermined timing. As an example, this predetermined timing is when input information identical or similar to the input information for which good feedback information has been obtained is input at a different timing. In other words, since the output information can be returned without processing the information again in the trained model 222, it is possible to quickly output output information that is estimated to be highly satisfying for the user U (because the user U has obtained good feedback information) while suppressing the calculation cost.

[0095] (Variation 1) Furthermore, if we look at the above-mentioned information processing system from a different perspective, the problem (purpose) that the information processing system of this embodiment aims to solve can also be seen as "efficiently obtaining the desired processing results even when the trained model is switched."

[0096] If the problem is understood as above, the invention as a means for solving the problem is, for example, as follows. "An information processing method executed by an information processing device, An input information acquisition step for acquiring input information to be input to the trained model; a version identifier generating step of generating a version identifier for identifying a version of the input information; storing the input information and the version identifier corresponding to the input information in a repository; An output information acquisition step of acquiring output information which is a result of inputting the input information corresponding to a predetermined version identifier into the trained model and a trained model different from the trained model, respectively; an evaluation step of comparing and evaluating each of the output information; an adjustment step of changing the input information when the result of the evaluation is a predetermined result; "An information processing method including the above."

[0097] According to the above configuration, input information corresponding to a predetermined version identifier is input to each of the different trained models, and the output information is compared and evaluated. Then, an adjustment step is performed to change the input information when the evaluation result in the evaluation step is a predetermined result, i.e., dissimilar, and thus the output information can be changed based on the changed input information. That is, it is possible to make the output information of one trained model similar to the output information of the other trained model. This makes it possible to suppress changes in the output information even when the trained model is switched, so that the desired result can be obtained efficiently. Note that the evaluation of the output information of one trained model and the output information of the other trained model can be performed by inputting the respective output information and a prompt for an instruction to compare and evaluate them into the large-scale language model. Also, in the adjustment step, the input information can be changed by inputting a prompt for an instruction to adjust the input information based on the evaluation result into the large-scale language model.

[0098] (Variation 2) Furthermore, if we look at the above-mentioned information processing system from a different perspective, the problem (purpose) that the information processing system of this embodiment aims to solve can also be seen as "reflecting the updated content in linked external systems, etc."

[0099] If the problem is understood as above, the invention as a means for solving the problem is, for example, as follows. "An information processing method executed by an information processing device, An input information acquisition step for acquiring input information to be input to the trained model; a version identifier generating step of generating a version identifier for identifying a version of the input information; storing the input information and the version identifier corresponding to the input information in a repository; an updating step of generating a new version identifier corresponding to an update of the input information in the repository and storing the updated input information and the corresponding version identifier in the repository when the input information in the repository is updated; and an output step of outputting at least one of the input information and the output information in a usable manner."

[0100] According to the above configuration, since at least one of the input information and the output information is output in a usable manner in the output step, even if the input information is updated in the update step, the updated result can be output. As a result, the updated input information can be used in an external system linked to the system that executes this information processing method, so that immediate response to changes can be realized. In other words, usability can be improved.

[0101] <Additional Notes> The present embodiment includes the following disclosure.

[0102] (Appendix 1) An information processing method executed by an information processing device, An input information acquisition step for acquiring input information to be input to the trained model; a version identifier generating step of generating a version identifier for identifying a version of the input information; storing the input information and the version identifier corresponding to the input information in a repository; an updating step of generating a new version identifier corresponding to an update of the input information in the repository and storing the updated input information and the corresponding version identifier in the repository when the input information in the repository is updated; An information processing method comprising:

[0103] (Appendix 2) and an acquisition step of acquiring, when a predetermined version identifier is designated, the input information corresponding to the version identifier from the repository. 2. An information processing method as described in Appendix 1.

[0104] (Appendix 3) At least one of comparing and evaluating the results of inputting a plurality of pieces of input information having different version identifiers into the trained model, 10. The information processing method according to claim 1 or 2.

[0105] (Appendix 4) storing branch information corresponding to the input information in the repository; An information processing method according to any one of Supplementary Note 1 to Supplementary Note 3.

[0106] (Appendix 5) The storing step associates output information from the trained model based on the acquired input information with the input information and stores the output information in the repository. An information processing method according to any one of Supplementary Note 1 to Supplementary Note 4.

[0107] (Appendix 6) Generate additional information necessary to obtain the output information from the trained model according to the input information, and acquire the input information to which the additional information has been added. 6. An information processing method as described in Appendix 5.

[0108] (Appendix 7) generating and storing the version identifier corresponding to the output information in the repository; 6. An information processing method as described in Appendix 5.

[0109] (Appendix 8) and an output step of outputting at least one of the input information and the output information in a viewable manner. 6. An information processing method as described in Appendix 5.

[0110] (Appendix 9) An information processing method executed by an information processing device, An input information acquisition step of acquiring input information to be input to the trained model and tag information in the input information; storing the input information and the tag information in a repository; An information processing method comprising:

[0111] (Appendix 10) and generating the tag information based on the input information. 10. An information processing method as described in Supplementary Note 9.

[0112] (Appendix 11) A search step of searching the input information using search condition information in a natural language. An information processing method according to any one of Supplementary Notes 1 to 10.

[0113] (Appendix 12) An information processing method executed by an information processing device, An input information acquisition step for acquiring input information to be input to the trained model; storing the input information in a repository; a feedback information acquisition step of acquiring feedback information corresponding to the input information in the repository; An information processing method comprising:

[0114] (Appendix 13) A correction recommendation step of generating at least one of a correction proposal and a correction plan for the input information based on the feedback information; 13. The information processing method according to claim 12.

[0115] (Appendix 14) In the information processing device, An input information acquisition step for acquiring input information to be input to the trained model; a version identifier generating step of generating a version identifier for identifying a version of the input information; storing the input information and the version identifier corresponding to the input information in a repository; an updating step of generating a new version identifier corresponding to an update of the input information in the repository and storing the updated input information and the corresponding version identifier in the repository when the input information in the repository is updated; A program for executing an information processing method including the steps of:

[0116] (Appendix 15) In the information processing device, An input information acquisition step for acquiring input information to be input to the trained model; An input information acquisition step of acquiring input information to be input to the trained model and tag information in the input information; storing the input information and the tag information in a repository; A program for executing an information processing method including the steps of:

[0117] (Appendix 16) In the information processing device, An input information acquisition step for acquiring input information to be input to the trained model; storing the input information in a repository; a feedback information acquisition step of acquiring feedback information corresponding to the input information in the repository; A program for executing an information processing method including the steps of:

[0118] (Appendix 17) An information processing system executed by an information processing device, An input information acquisition unit that acquires input information to be input to the trained model; a version identifier generating unit for generating a version identifier for identifying a version of the input information; a storage execution unit that stores the input information and the version identifier corresponding to the input information in a repository; an update unit that, when the input information in the repository is updated, generates a new version identifier in response to the update and stores the updated input information and the corresponding version identifier in the repository; An information processing system having the above configuration.

[0119] (Appendix 18) An information processing system executed by an information processing device, An input information acquisition unit that acquires input information to be input to the trained model; A tag information acquisition unit that acquires input information to be input to the trained model and tag information in the input information; a storage execution unit that stores the input information and the tag information in a repository; An information processing system having the above configuration.

[0120] (Appendix 19) An information processing system executed by an information processing device, An input information acquisition unit that acquires input information to be input to the trained model; a storage execution unit that stores the input information in a repository; a feedback information acquisition unit that acquires feedback information corresponding to the input information in the repository; An information processing system having the above configuration.

[0121] The embodiments disclosed herein are illustrative in all respects and should not be considered as limiting. The scope of the present invention is indicated by the claims, not by the above meaning, and is intended to include all modifications within the meaning and scope of the claims. The present invention is not limited to the above-mentioned embodiments, and various modifications are possible within the scope of the claims, and embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention. [Explanation of symbols]

[0122] 10. Information Processing Systems 12 Server (information processing device) 14 User terminal (information processing device) 220 Programs 222 trained models 242 Information acquisition unit (input information acquisition unit) 244 Information processing unit (version identifier generation unit, storage execution unit, update unit) 500 Information Processing Systems 600 Information Processing Systems 742 Information acquisition unit (input information acquisition unit, tag information acquisition unit) 744 Information processing unit (version identifier generation unit, storage execution unit, update unit) 842 Information acquisition unit (input information acquisition unit) 844 Information processing unit (version identifier generation unit, storage execution unit, update unit)

Claims

1. An information processing method executed by an information processing device, An input information acquisition step for acquiring input information to be input to the trained model; a version identifier generating step of generating a version identifier for identifying a version of the input information; storing the input information and the version identifier corresponding to the input information in a repository; an updating step of generating a new version identifier corresponding to an update of the input information in the repository and storing the updated input information and the corresponding version identifier in the repository when the input information in the repository is updated; An information processing method comprising:

2. and an acquisition step of acquiring, when a predetermined version identifier is designated, the input information corresponding to the version identifier from the repository. The information processing method according to claim 1 .

3. At least one of comparing and evaluating the results of inputting a plurality of pieces of input information with different version identifiers into the trained model, The information processing method according to claim 1 .

4. storing branch information corresponding to the input information in the repository; The information processing method according to claim 1 .

5. The storing step associates output information from the trained model based on the acquired input information with the input information and stores the output information in the repository. The information processing method according to claim 1 .

6. Generate additional information necessary to obtain the output information from the trained model according to the input information, and acquire the input information to which the additional information has been added. The information processing method according to claim 5.

7. generating and storing the version identifier corresponding to the output information in the repository; The information processing method according to claim 5.

8. and an output step of outputting at least one of the input information and the output information in a viewable manner. The information processing method according to claim 5.

9. A search step of searching the input information using search condition information in a natural language. The information processing method according to claim 1 .

10. In the information processing device, An input information acquisition step for acquiring input information to be input to the trained model; a version identifier generating step of generating a version identifier for identifying a version of the input information; storing the input information and the version identifier corresponding to the input information in a repository; an updating step of generating a new version identifier corresponding to an update of the input information in the repository and storing the updated input information and the corresponding version identifier in the repository when the input information in the repository is updated; A program for executing an information processing method including the steps of:

11. An information processing system executed by an information processing device, An input information acquisition unit that acquires input information to be input to the trained model; a version identifier generating unit for generating a version identifier for identifying a version of the input information; a storage execution unit that stores the input information and the version identifier corresponding to the input information in a repository; an update unit that, when the input information in the repository is updated, generates a new version identifier in response to the update and stores the updated input information and the corresponding version identifier in the repository; An information processing system having the above configuration.

Images