Information processing device, information processing system, program, and information processing method

Abstract

An information processing device (100) comprises: an information acquisition unit (110) that acquires a plurality of pieces of information; a request acquisition unit (120) that acquires request information indicating a request from a user; a determination unit (150) that determines, on the basis of the request information acquired by the request acquisition unit (120), whether the request from the user is a specific request from among requests for causing a specific piece of information to be extracted from the plurality of pieces of information acquired by the information acquisition unit (110); an output unit (170) that, on the basis of a determination result by the determination unit (150), outputs the request information to a trained model (41) that generates an answer to the request from the user if the request from the user is not the specific request; and an extraction unit (160) that, on the basis of the determination result by the determination unit (150), extracts information corresponding to the request from the user from the plurality of pieces of information acquired by the information acquisition unit (110) if the request from the user is the specific request.

Description

Information Processing Apparatus, Information Processing System, Program, and Information Processing Method

[0001] The present disclosure relates to an information processing apparatus, an information processing system, a program, and an information processing method.

[0002] Conventionally, a design support system has been disclosed that inputs a request from a designer into a large language model and provides design support to the designer based on the information output from the large language model (see, for example, Patent Document 1). The design support system described in Patent Document 1 outputs a plurality of change candidates of editable editing elements in design information, which require the designer's judgment for their change, as a request from the designer to the large language model, so that the designer can select a desired change candidate from these change candidates.

[0003] Japanese Patent No. 7534834

[0004] Generally, there are learned models such as large language models that can output information indicating answers to a wide range of requests not limited to design support. However, generally, requests for learned models that can output information indicating answers to wide-ranging requests include those with relatively high and low accuracy of information output from the learned models. Therefore, depending on the content of the request for the learned model, there is a problem that it is difficult to improve the accuracy of the output information.

[0005] The present disclosure has been made in view of the recognition of the above problems, and an object thereof is to provide an information processing apparatus, an information processing system, a program, and an information processing method capable of improving the accuracy of information when a user obtains an answer to a request as compared with the prior art.

[0006] The information processing device relating to this disclosure is characterized by comprising: an information acquisition unit that acquires multiple pieces of information; a request acquisition unit that acquires request information that indicates a request from a user in natural language; a determination unit that determines, based on the request information acquired by the request acquisition unit, whether or not the request from the user is a specific request among requests for extracting specific information from multiple pieces of information acquired by the information acquisition unit; an output unit that, based on the determination result by the determination unit, outputs the request information to a trained model that generates a response to the request from the user based on the input of the request information if the request from the user is not a specific request; and an extraction unit that, based on the determination result by the determination unit, extracts information corresponding to the request from multiple pieces of information acquired by the information acquisition unit if the request from the user is a specific request.

[0007] According to this disclosure, it is possible to improve the accuracy of information when users obtain answers to their requests compared to before.

[0008] This is a block diagram showing the schematic configuration of an information processing system according to Embodiment 1. This is a block diagram showing an example of the hardware configuration of an information processing device according to Embodiment 1. This is a block diagram showing an example of the hardware configuration of an information processing device according to Embodiment 1. This is a flowchart showing an example of processing performed by an information processing device according to Embodiment 1. This is a scatter plot showing information extracted by an information processing device according to Embodiment 1. This is a histogram showing information extracted by an information processing device according to a modified version of Embodiment 1. This is a table showing an example of information extracted by an information processing device according to a modified version of Embodiment 1. This is a block diagram showing the schematic configuration of an information processing system according to Embodiment 2. This is a flowchart showing an example of processing performed by an information processing device according to Embodiment 2. This is a scatter plot showing information extracted by an information processing device according to Embodiment 2 when the correlation coefficient is a negative value. This is a scatter plot showing information extracted by an information processing device according to Embodiment 2 when the correlation coefficient is a positive value.

[0009] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. Embodiment 1. First, the information processing system 1 according to Embodiment 1 will be described with reference to Figure 1. Figure 1 is a block diagram showing the schematic configuration of the information processing system 1 according to Embodiment 1. The information processing system according to Embodiment 1 is a system for acquiring information indicating a request from a user and providing the user with information corresponding to the user's request. As shown in Figure 1, the information processing system 1 according to Embodiment 1 comprises a storage device 10, an input device 20, a display device 30, a model processing device 40, and an information processing device 100, which are connected wirelessly or by wire to enable communication between them. The storage device 10, the input device 20, the display device 30, the model processing device 40, and the information processing device 100 may be connected to each other via devices or communication lines not shown to enable communication of information between them.

[0010] The storage device 10 stores multiple pieces of information used by the user of the information processing system 1. For example, the storage device 10 stores information about multiple products, multiple goods, or other items as multiple pieces of information used by the user of the information processing system 1. Specifically, the storage device 10 stores multiple combinations of information, each consisting of a combination of information indicating the values ​​of two or more variables (parameters), including the value of a first variable and the value of a second variable.

[0011] For example, the storage device 10 stores information indicating the values ​​of variables related to the type, characteristics, attributes, and sales of multiple products, used by a user of the information processing system 1 who is a consumer considering purchasing a product or a product designer. For example, the storage device 10 stores the values ​​of variables related to the type, characteristics, attributes, and sales of multiple motors, used by a user of the information processing system 1 who is a motor designer. In other words, the storage device 10 is a database of multiple motors, used by a user of the information processing system 1 who is a motor designer, and stores the values ​​of columns related to the type, characteristics, attributes, and sales of each motor. Also, for example, the storage device 10 stores information indicating the values ​​of variables related to the type, characteristics, attributes, and sales of multiple LED chips, used by a user of the information processing system 1 who is an LED (Light Emitting Diode) chip designer. Also, for example, the storage device 10 stores information indicating the values ​​of variables related to the type, attributes, and sales of multiple properties, used by a user of the information processing system 1 who is a consumer looking for a property or a real estate agent.

[0012] The input device 20 receives input operations from the user and outputs information corresponding to those operations. For example, the input device 20 is composed of a keyboard, mouse, touch panel, mechanical switch, camera, microphone, etc., which accept user input operations for inputting information expressed in natural language, and outputs information expressed in natural language corresponding to the user's input operations. The information expressed in natural language corresponding to the user's input operations may be, for example, text data entered from the keyboard, text data converted into text by image recognition processing after photographing a document including drawings, or data converted into text by speech recognition processing after inputting audio signals into the microphone. The information expressed in natural language corresponding to the user's input operations may also be feature data obtained by speech recognition of audio signals, or feature data obtained by image recognition of images. The input device 20 outputs the information corresponding to the user's input operations to the information processing device 100.

[0013] The display device 30 displays information input from the input device 20, the model processing device 40, and the information processing device 100. In other words, the display device 30 visually outputs the information input from the input device 20, the model processing device 40, and the information processing device 100. For example, the display device 30 is composed of a liquid crystal display panel, an organic or inorganic EL (Electroluminescence) panel, a dot matrix display, an LED (Light Emitting Diode), etc., and displays various types of information input from the input device 20, the model processing device 40, and the information processing device 100. If the input device 20 is composed of a touch panel, the display device 30 may be configured integrally with the input device 20.

[0014] The model processing unit 40 stores a Large Language Model (LLM) 41 and, based on the input of information from the input device 20 and the information processing unit 100, generates information corresponding to the input information and outputs the generated information to the display device 30 and the information processing unit 100. For example, the model processing unit 40 is composed of a cloud server, a physical server, or other computer. For example, the LLM 41 is a trained model as a transformer model that generates a response to a user's request in natural language based on the input of information that represents the user's request in natural language. For example, the LLM 41 can be created by supervised learning using training data that includes information that represents multiple user requests in natural language and information that represents the response to each of the multiple user requests in natural language. Alternatively, the LLM 41 can be created by learning a sequence of words or sentences through self-supervised learning using a large amount of text data. Furthermore, unsupervised learning methods such as reinforcement learning and GAN (Generative Adversarial Network) may be applied to the training of LLM41. In addition, supervised and unsupervised learning may be used in combination to train LLM41. Examples of transformer models include BERT (Bidirectional Encoder Representation from Transformers) and GPT (Generative Pre-trained Transformer). Furthermore, the LLM may be a self-adaptive LLM.

[0015] The information processing device 100 includes an information acquisition unit 110, a request acquisition unit 120, a determination unit 150, an extraction unit 160, and an output unit 170.

[0016] The information acquisition unit 110 acquires various types of information from the storage device 10 and the model processing device 40. For example, the information acquisition unit 110 acquires from the storage device 10 multiple pieces of information that are used by users of the information processing system 1 and are stored in the storage device 10. Specifically, the information acquisition unit 110 acquires from the storage device 10 information about multiple products that are used by users of the information processing system 1 and are stored in the storage device 10. For example, the information acquisition unit 110 acquires from the storage device 10 multiple combination pieces of information consisting of the values ​​of two or more variables, each including the value of a first variable and the value of a second variable. In other words, the information acquisition unit 110 acquires from the storage device 10 the first combination piece of information consisting of the values ​​of the first and second variables of the first product, the second combination piece of information consisting of the values ​​of the first and second variables of the second product, ..., the nth combination piece of information consisting of the values ​​of the first and second variables of the nth product. Specifically, the information acquisition unit 110 acquires multiple combinations of information, each consisting of the values ​​of two or more variables, and each combination of values ​​of variables related to product type, characteristics, attributes, and sales.

[0017] For example, the information acquisition unit 110 acquires information indicating the type of motor as a product, such as whether each motor is a DC motor, an AC motor, or a stepping motor. Also, for example, the information acquisition unit 110 acquires information indicating the characteristics of each motor as a product, such as the efficiency, starting torque, rated torque, and rated rotational speed values ​​of each motor. Also, for example, the information acquisition unit 110 acquires information indicating the attributes of each motor as a product, such as the core length, core outer diameter, number of poles, maximum dimensions, mass, rated voltage, and rated current values ​​of each motor. Also, for example, the information acquisition unit 110 acquires information indicating the price, inventory quantity, and monthly production capacity of each motor as sales information for each motor as a product. Note that the information acquisition unit 110 only needs to be configured to acquire multiple pieces of information, and may also be configured to acquire information other than those described above from the storage device 10.

[0018] Furthermore, for example, the information acquisition unit 110 acquires information generated by the LLM 41 based on the input of information from the input device 20 and the information processing device 100 from the model processing device 40.

[0019] The request acquisition unit 120 acquires request information indicating a request from a user to the information processing system 1. For example, the request acquisition unit 120 acquires request information indicating a request from a user to the information processing system 1 in natural language, based on information from the input device 20 in response to input operations by the user. For example, the request acquisition unit 120 acquires information indicating a request for the information processing system 1 to extract specific information from multiple pieces of information acquired by the information acquisition unit 110, information indicating a request for the information processing system 1 to answer a general question, and other information indicating requests to the information processing system 1.

[0020] Specifically, the request acquisition unit 120 acquires information indicating a request for the information processing system 1 to extract specific information from multiple pieces of information acquired by the information acquisition unit 110, specifically information indicating a request for the information processing system 1 to extract specific information by searching for specific information among multiple pieces of information acquired by the information acquisition unit 110, and information indicating a request for the information processing system 1 to extract specific information by analyzing multiple pieces of information acquired by the information acquisition unit 110. More specifically, the request acquisition unit 120 acquires information indicating a request for the information processing system 1 to extract specific information from multiple pieces of information acquired by the information acquisition unit 110, specifically information indicating a request for the information processing system 1 to extract information about a specific product by searching for information about a specific product among multiple pieces of product information acquired by the information acquisition unit 110, and information indicating a request for the information processing system 1 to extract information about a specific product by analyzing information about multiple products acquired by the information acquisition unit 110. In Embodiment 1, "analysis" means, for example, a process other than a simple information search process among the processes of extracting information other than already known information from multiple pieces of information. Furthermore, in Embodiment 1, "simple information retrieval processing" refers to, for example, processing to extract information from multiple pieces of information in which a specific variable is the maximum or minimum, or processing to extract information in which a specific variable has a specific value.

[0021] For example, a request to have the information processing system 1 extract specific information by searching for specific information among multiple pieces of information acquired by the information acquisition unit 110 is a request to have the information processing system 1 extract information indicating a product by searching for information indicating a product that matches a specific condition among the information about each of the multiple products acquired by the information acquisition unit 110, specifically information of one specific variable or a combination of multiple specific variables. Also, for example, a request to have the information processing system 1 extract specific information by analyzing multiple pieces of information acquired by the information acquisition unit 110 is a request to have the information processing system 1 extract information indicating the results of statistical analysis such as calculating correlation values ​​between multiple specific variables, calculating variances of one or more specific variables, calculating standard deviations, and calculating representative values ​​such as mean values, among the combination information about each of the multiple products acquired by the information acquisition unit 110. Furthermore, for example, a request to cause the information processing system 1 to extract specific information by analyzing multiple pieces of information acquired by the information acquisition unit 110 is a request to cause the information processing system 1 to extract information indicating products that meet specific conditions, which is extracted by statistical analysis such as calculating correlation values ​​between multiple specific variables, calculating the variance of one or more pieces of specific information, calculating the standard deviation, and calculating representative values ​​such as the mean, or information on combinations of multiple variables about products that meet specific conditions, which is extracted by these statistical analyses.

[0022] Furthermore, for example, the request acquisition unit 120 acquires information indicating a request for the information processing system 1 to answer a question about general knowledge, as well as information indicating a request for the information processing system 1 to answer a question about the general meaning of a specific term. Furthermore, for example, the request acquisition unit 120 acquires information indicating a request for the information processing system 1 to answer a question about the design procedure of a specific product, as well as information indicating a request for the information processing system 1 to answer a question about the design procedure of a specific product.

[0023] The determination unit 150 determines, based on the request information acquired by the request acquisition unit 120, whether the request from the user is a specific request, which is a pre-set specific request. For example, a specific request is a request to extract specific information from multiple pieces of information acquired by the information acquisition unit 110. Specifically, a specific request is a request to extract specific information from multiple pieces of information acquired by the information acquisition unit 110 through analysis of the multiple pieces of information acquired by the information acquisition unit 110. More specifically, a specific request is a request to extract specific information from multiple pieces of information acquired by the information acquisition unit 110 through statistical analysis of the multiple pieces of information acquired by the information acquisition unit 110. Also, for example, a specific request is a request to extract specific combination information from multiple combination pieces of information acquired by the information acquisition unit 110. Specifically, a specific request is a request to extract specific combination information from multiple combination pieces of information acquired by the information acquisition unit 110 through analysis of the multiple combination pieces of information acquired by the information acquisition unit 110. More specifically, a specific request is a request to extract specific combination information from the multiple combination information acquired by the information acquisition unit 110 by performing a statistical analysis of multiple specific variables of the multiple combination information acquired by the information acquisition unit 110.

[0024] Generally, trained models capable of outputting information that provides answers to a wide range of requests, such as LLM, have difficulty with high accuracy in outputting information that provides answers to requests for which clear ethical judgments are difficult. For example, generally, such trained models have difficulty with high accuracy in outputting information that provides answers to requests that require complex analysis. For this reason, the information processing device 100 according to Embodiment 1 is configured such that the determination unit 150 determines whether the request from the user is a specific request, or in other words, whether it is appropriate to obtain an answer from a trained model. If it is appropriate to obtain an answer from a trained model, the trained model capable of outputting information that provides answers to a wide range of requests generates information that provides an answer. If it is not appropriate to obtain an answer from such a trained model, the information that provides an answer is generated by another mathematical algorithm or a dedicated trained model.

[0025] For example, the determination unit 150 determines whether the user's request is a specific request to extract specific information from multiple pieces of information acquired by the information acquisition unit 110 by determining whether the request information acquired by the request acquisition unit 120 contains specific information. Specifically, the determination unit 150 determines whether the user's request is a specific request to extract specific information from multiple pieces of information acquired by the information acquisition unit 110 by determining whether the request information acquired by the request acquisition unit 120 contains natural language information such as "Please extract it." The determination unit may also be configured to determine whether the user's request is a specific request by other means. For example, it may be configured to determine whether the user's request is a specific request by an LLM 41 or another trained model stored in a model processing device or information processing device that has learned from training data including request information indicating multiple requests from the user and information indicating whether each request is a specific request, and is capable of determining whether the input request information is a specific request.

[0026] Based on the determination result by the determination unit 150, the extraction unit 160 extracts information corresponding to the user's request from the multiple pieces of information acquired by the information acquisition unit 110 if the user's request is a specific request. For example, based on the determination result by the determination unit 150, if the user's request is a specific request, the extraction unit 160 extracts specific combination information corresponding to the user's request from the multiple combination information acquired by the information acquisition unit 110. Specifically, based on the determination result by the determination unit 150, if the user's request is a specific request, the extraction unit 160 extracts multiple specific combination information corresponding to the user's request from the multiple combination information acquired by the information acquisition unit 110.

[0027] For example, based on the determination result by the determination unit 150, if the user's request is a specific request, the extraction unit 160 extracts information in natural language that indicates a specific variable included in the multiple pieces of information obtained by the information acquisition unit 110, and information in natural language that indicates the extraction conditions, from the request information obtained by the request acquisition unit 120, and extracts information from the multiple pieces of information obtained by the information acquisition unit 110 in which the value of the specific variable matches the extraction conditions. Furthermore, if the user request is a specific request, the extraction unit may be configured to extract information corresponding to the user request from the multiple pieces of information acquired by the information acquisition unit 110 by other means. For example, it may be configured to extract information corresponding to the user request from the multiple pieces of information acquired by the information acquisition unit 110 using an LLM 41 or another trained model stored in the model processing device or information processing device that is trained using training data including request information indicating multiple requests from the user and information indicating variables and extraction conditions indicated by each piece of request information, and that can identify variables and extraction conditions based on the input request information. Alternatively, it may be configured to extract information corresponding to the user request from the multiple pieces of information acquired by the information acquisition unit 110 using a specific mathematical algorithm that is set in advance.

[0028] The output unit 170 outputs information from the information processing device 100 to the display device 30 and the model processing device 40. For example, if the user request is not a specific request based on the determination result by the determination unit 150, the output unit 170 outputs request information to the LLM 41 in order to cause the LLM 41 to generate a response to the user request. Also, for example, the output unit 170 outputs information indicating the result of processing by the information processing device 100 to the display device 30 so that the result of processing by the information processing device 100 is displayed on the display device 30. Specifically, the output unit 170 outputs information extracted by the extraction unit 160 to the display device 30 so that the information extracted by the extraction unit 160 is displayed on the display device. More specifically, the output unit 170 outputs multiple specific combinations of information extracted by the extraction unit 160 to the display device 30 so that multiple specific combinations of information extracted by the extraction unit 160 are displayed on the display device.

[0029] Next, the hardware configuration of the information processing device 100 will be described with reference to Figures 2 and 3. Figure 2 is a diagram showing an example of the hardware configuration of the information processing device 100, and Figure 3 is a diagram showing an example of the hardware configuration of the information processing device 100 that is different from Figure 2. For example, as shown in Figure 2, the information processing device 100 is composed of a computer having a processor 100a, a memory 100b, and an I / O port 100c, and is configured so that the processor 100a reads and executes a program stored in the memory 100b.

[0030] Furthermore, as shown in Figure 3, for example, the information processing device 100 is composed of a computer having a processing circuit 100d, which is dedicated hardware, and an I / O port 100c. The processing circuit 100d is composed of, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or a combination thereof. Each function of the information processing device 100 is realized by these processors 100a or the processing circuit 100d, which is dedicated hardware, executing a program. Note that the information processing device 100 may have hardware other than that described above. Also, the hardware configuration of the model processing device 40 is the same as that of the information processing device 100, so its description is omitted.

[0031] Next, with reference to Figures 1 and 4, the details of the processing performed by the information processing device 100 will be described. Figure 4 is a flowchart showing an example of the processing performed by the information processing device 100 according to Embodiment 1. The processing performed by the information processing device 100 shown in Figure 4 is the processing for providing information to the user in response to the user's request.

[0032] As shown in Figure 4, when the information processing device 100 starts processing, it first acquires multiple pieces of information (step ST01). For example, in this process, the information processing device 100 acquires multiple pieces of combination information from the storage device 10, which consist of combinations of values ​​of multiple variables. Specifically, in this process, the information processing device 100 acquires combination information from the storage device 10 for multiple motors, which consists of a combination of a value indicating the motor core length as the first variable and a value indicating the motor starting torque as the second variable. Note that the first and second variables are not limited to core length and starting torque, but may be various variables that constitute the combination information.

[0033] When the information processing device 100 performs the processing in step ST01, it acquires request information (step ST02). In this process, the information processing device 100 acquires request information indicating a request from the user from the input device 20, which has been operated by the user, using the request acquisition unit 120. Specifically, in this process, the information processing device 100 acquires request information from the input device 20, which has been operated by the user, who is a motor designer, using the request acquisition unit 120, which indicates a request from the user, saying in natural language, "Please extract three motors with a short core length and high starting torque." Also specifically, in this process, the information processing device 100 acquires request information from the input device 20, which has been operated by the user, who is a motor designer, using the request acquisition unit 120, which indicates a request from the user, saying in natural language, "Please tell me about the starting torque of the motor."

[0034] When the information processing device 100 performs the processing in step ST02, it determines whether the user's request is a specific request (step ST10). In this process, the information processing device 100 uses a determination unit 150 to determine whether the user's request is a specific request based on the request information obtained in step ST02. For example, in this process, the information processing device 100 uses a determination unit 150 to determine whether the user's request is a specific request to extract specific information from multiple pieces of information obtained in step ST01, based on the request information obtained in step ST02. Specifically, in this process, the information processing device 100 uses a determination unit 150 to determine whether the user's request is a specific request to extract specific information from multiple pieces of information obtained in step ST01 by determining whether the request information obtained in step ST02, which expresses the request "Please extract three motors with short core lengths and high starting torque" in natural language, contains the natural language information "Please extract".

[0035] In the process of step ST10, if the user's request is a specific request (YES in step ST10), the information processing device 100 extracts information corresponding to the user's request from multiple pieces of information (step ST13). For example, in this process, based on the request information expressed in natural language as "Please extract three motors with short core lengths and large starting torques," the information processing device 100 extracts information from multiple combinations of values ​​indicating the core length of the motors and values ​​indicating the starting torque of the motors, obtained by the information acquisition unit 110, using the extraction unit 160 to extract information on three motors that are in a combination where the core length is relatively short and the starting torque is large, and which together constitute a Pareto solution of the shortness of the core length and the magnitude of the starting torque.

[0036] Figure 5 is a scatter plot showing the information extracted by the information processing device 100 according to Embodiment 1. For example, in Figure 5, variable A represents the "shortness of the motor's core length," with the further to the right, the shorter the motor's core length. Variable B represents the "magnitude of the motor's starting torque," with the further upward, the greater the motor's starting torque. Each point represents a combination of the motor's core length and starting torque corresponding to each of the multiple pieces of information acquired by the information acquisition unit 110. For example, based on the request information obtained in step ST02, which indicates the request to "extract three motors with short core lengths and large starting torques," the information processing device 100 extracts combination information D1, combination information D2, and combination information D3, which are Pareto solutions to each other, from the multiple combination information obtained in step ST01, which consists of combinations of values ​​indicating the motor's core length and values ​​indicating the motor's starting torque as shown in Figure 5.

[0037] When the information processing device 100 performs the processing in step ST13, it outputs the extracted information to the display device 30 (step ST14). In this process, the information processing device 100 outputs the information extracted in the processing of step ST13, which corresponds to the user's request, using the output unit 170, and displays it as visual information on the display device 30, thereby providing the user with an answer to the user's request. The output unit 170 may be configured to display the combination information itself extracted in the processing of step ST13 on the display device 30, or it may be configured to display information for identifying products such as motors that correspond to the combination information extracted in the processing of step ST13 on the display device 30, or it may be configured to display a diagram, table or graph that illustrates the relationship between multiple combination information extracted in the processing of step ST13, as shown in Figure 5, on the display device 30, or it may be configured to display a pair plot diagram that includes both a scatter plot and a histogram on the display device 30.

[0038] In the process of step ST10, if the user's request is not a specific request (NO in step ST10), the information processing device 100 outputs the request information to the LLM 41 (step ST17). In other words, in this process, the information processing device 100 inputs the request information obtained in the process of step ST02 to the LLM 41, causing the LLM 41 to generate a response to the user's request. For example, in this process, based on the request information that expresses the request "Please tell me about the motor's starting torque" in natural language, the information processing device 100 outputs the request information to the LLM 41 via the output unit 170.

[0039] When the information processing device 100 performs the processing in step ST17, it acquires the information generated by the LLM 41 (step ST18). In this process, the information processing device 100 inputs the request information acquired in the processing of step ST02 into the LLM 41, and the information generated by the LLM 41 is acquired by the information acquisition unit 110. For example, if in the processing of step ST17, request information expressing the request "Please tell me about the motor's starting torque" in natural language is input into the LLM 41, then in the processing of step ST18, the information processing device 100 acquires the information generated by the LLM 41, which expresses the explanation of the motor's starting torque in natural language, using the information acquisition unit 110.

[0040] When the information processing device 100 performs the processing in step ST18, it outputs the acquired information to the display device 30 (step ST19). In this process, the information processing device 100 outputs the information from the LLM 41 acquired in the processing of step ST18 using the output unit 170 and displays it as visual information on the display device 30, thereby providing the user with a response to the user's request. If the user's request is a specific request, the information processing device 100 may be configured to display on the display device 30, together with the information acquired from the LLM 41 in the processing of step ST18, information corresponding to the user's request extracted by the extraction unit 160 from multiple pieces of information in the processing of step ST13.

[0041] The information processing device 100 terminates processing after performing the processing in step ST14 or step ST19. For example, the information processing device 100 is configured to perform the processing from step ST01 to step ST19 described above each time the user makes a specific input operation to the input device 20 to start using the information processing system 1.

[0042] As described above, the information processing apparatus 100 according to Embodiment 1 includes an information acquisition unit 110 that acquires a plurality of pieces of information, a request acquisition unit 120 that acquires request information indicating a request from a user in natural language, and a determination unit 150 that determines whether a request from the user is a specific request among requests for extracting specific information from the plurality of pieces of information acquired by the information acquisition unit 110 based on the request information acquired by the request acquisition unit 120. When the request from the user is not a specific request based on the determination result by the determination unit 150, an output unit 170 that outputs the request information to the LLM 41 that generates an answer to the request from the user based on the input of the request information, and when the request from the user is a specific request based on the determination result by the determination unit 150, an extraction unit 160 that extracts information corresponding to the request from the user from the plurality of pieces of information acquired by the information acquisition unit 110.

[0043] Further, in the information processing apparatus 100 according to Embodiment 1, when the request from the user is a specific request based on the determination result by the determination unit 150, the extraction unit 160 is configured to extract a plurality of specific combination information that are Pareto solutions corresponding to the request from the user among the plurality of combination information acquired by the information acquisition unit 110.

[0044] With such a configuration, even when the information processing apparatus 100 uses a learned model with insufficient information accuracy when extracting specific information from a plurality of pieces of information, for requests that the learned model is not good at, answers can be generated by other algorithms, so the information accuracy when the user obtains an answer to the request can be improved compared to the conventional case.

[0045] In Embodiment 1, the information processing apparatus 100 is configured to output the information extracted by the extraction unit 160 by the output unit 170 and display the scatter diagram shown in FIG. 5 on the display device 30, but is not limited thereto. The information processing apparatus may be configured to display the information extracted by the extraction unit on the display device 30, and may be configured to display information other than the scatter diagram of the plurality of pieces of information on the display device 30.

[0046] FIG. 6 is a histogram showing information extracted by the information processing apparatus according to a modification of Embodiment 1. For example, the information processing apparatus may be configured to extract, from a plurality of information acquired by the information acquisition unit 110, information indicating a histogram regarding a specific variable by the extraction unit, and cause the display device 30 to display the extracted information as the histogram shown in FIG. 6.

[0047] FIG. 7 is a table showing an example of information extracted by the information processing apparatus according to a modification of Embodiment 。For example, the information processing apparatus may be configured to extract, by the extraction unit, a correlation coefficient between a plurality of variables from a plurality of combination information each consisting of a combination of values indicating the plurality of variables acquired by the information acquisition unit 110, and cause the display device 30 to display the extracted correlation coefficient as the table shown in FIG. 7. For example, in FIG. 7, it shows that the correlation coefficient between variable A and variable B is -0.8, the correlation coefficient between variable A and variable C is -0.8, and the correlation coefficient between variable B and variable C is 0.8.

[0048] Embodiment 2. Next, referring to FIGS. 8 to 11, the information processing system 1A according to Embodiment 2 will be described. The information processing system 1A according to Embodiment 2 has a different configuration regarding the functions of the information processing apparatus from the information processing system 1 according to Embodiment 1, but the other configurations are the same. For the same configurations as those in Embodiment 1, the same names and reference numerals as those in Embodiment 1 are given and the description thereof is omitted

[0049] FIG. 8 is a block diagram showing a schematic configuration of the information processing system 1A according to Embodiment 2. As shown in FIG. 8, the information processing system 1A according to Embodiment 2 includes a storage device 10, an input device 20, a display device 30, a model processing device 40, and an information processing device 100A, which are connected wirelessly or wired so as to be able to communicate with each other. Note that the storage device 10, the input device 20, the display device 30, the model processing device 40, and the information processing device 100A may be communicably connected to each other via a device or communication line not shown in the figure.

[0050] The information processing device 100A includes an information acquisition unit 110, a request acquisition unit 120, a prompt generation unit 130, a calculation unit 140, a determination unit 150A, an extraction unit 160A, and an output unit 170A.

[0051] The prompt generation unit 130 generates a type prompt to cause the LLM 41 to generate type information indicating which of a plurality of pre-set types the user's request belongs to, based on the request information acquired by the request acquisition unit 120. For example, the prompt generation unit 130 generates a type prompt to cause the LLM 41 to generate type information indicating whether the user's request belongs to a specific type corresponding to a specific request or a non-specific type that does not correspond to a specific request, based on the request information acquired by the request acquisition unit 120, which expresses the request "Please extract three motors with short core lengths and high starting torque" in natural language, and generates a type prompt to cause the LLM 41 to generate type information indicating whether the user's request belongs to a specific type corresponding to a specific request for extracting specific combination information from multiple combination information through analysis of multiple combination information acquired by the information acquisition unit 110, or a non-specific type that does not correspond to a specific request. Specifically, the prompt generation unit 130 generates a type prompt to cause the LLM 41 to generate type information indicating whether the user's request corresponds to a specific type that extracts a specific combination of information from multiple combinations of information obtained by the information acquisition unit 110 through analysis of multiple combinations of information, or to a non-specific type that does not correspond to a specific request.

[0052] The prompt generation unit 130 may be configured to generate a type prompt that generates type information for making a final determination of the type of user request by inputting a prompt to the LLM 41 only once, or it may be configured to generate a type prompt that generates type information for making a final determination of the type of user request by inputting a prompt to the LLM 41 multiple times. For example, the prompt generation unit 130 may be configured to generate a first type prompt that causes the LLM 41 to generate first type information that causes the LLM 41 to generate first type information that causes the LLM 41 to generate first type information that causes the LLM 41 to generate first type information that causes the LLM 41 to generate second

[0053] Furthermore, the prompt generation unit 130 generates a variable prompt to cause the LLM 41 to generate variable information indicating which two or more variables correspond to the user's request, based on the request information obtained by the request acquisition unit 110, which is a request for the information processing system 1A to respond regarding two or more variables. In other words, based on the request information obtained by the request acquisition unit 120, which expresses the request "Please extract three motors with short core lengths and high starting torque" in natural language, the prompt generation unit 130 generates a variable prompt to cause the LLM 41 to generate variable information indicating that the two or more variables corresponding to the user's request are "core length" and "starting torque," if the request is a request to cause the information processing system 1A to respond for two or more variables.

[0054] The information acquisition unit 110 according to Embodiment 2 acquires type information generated by the LLM 41 as a result of inputting a type prompt generated by the prompt generation unit 130 to the LLM 41. In addition, the information acquisition unit 110 according to Embodiment 2 acquires variable information generated by the LLM 41 as a result of inputting a variable prompt generated by the prompt generation unit 130 to the LLM 41.

[0055] The calculation unit 140 calculates the correlation coefficient between the values ​​of two or more variables based on a plurality of combination information obtained by the information acquisition unit 110, each consisting of a combination of the values ​​of two or more variables. For example, the calculation unit 140 calculates the correlation coefficient between the values ​​of a first variable and the values ​​of a second variable based on a plurality of combination information obtained by the information acquisition unit 110, each consisting of a combination of the values ​​of a first variable and the values ​​of a second variable. The method for calculating the correlation coefficient can be, for example, Pearson's correlation coefficient or Spearman's correlation coefficient.

[0056] The determination unit 150A determines whether a user request is a specific request to extract specific information from multiple pieces of information acquired by the information acquisition unit 110, based on an analysis of the multiple pieces of information acquired by the information acquisition unit 110. For example, the determination unit 150A determines whether a user request is a specific request to extract specific information from multiple pieces of information acquired by the information acquisition unit 110, based on a statistical analysis of the multiple pieces of information acquired by the information acquisition unit 110. Alternatively, for example, the determination unit 150A determines whether a user request is a specific request to extract multiple specific combinations of information from multiple combinations of information acquired by the information acquisition unit 110. Furthermore, the determination unit 150A determines whether a user request is a specific request based on the type information generated by the LLM 41.

[0057] Furthermore, the determination unit 150A may be configured to determine whether a request from the user is a specific request based on the correlation coefficient calculated by the calculation unit 140. For example, the determination unit 150A may be configured to determine that a request from the user is not a specific request if the value of the correlation coefficient calculated by the calculation unit 140 is positive. Also, for example, the determination unit 150A may be configured to determine that a request from the user is not a specific request if it is a request to extract specific combination information from a plurality of combination information acquired by the information acquisition unit 110, and the value of the correlation coefficient calculated by the calculation unit 140 is negative or the correlation coefficient is a value indicating no correlation.

[0058] Based on the determination result by the judgment unit 150A, if the user's request is a specific request, the extraction unit 160A extracts information corresponding to the user's request from among the multiple pieces of information acquired by the information acquisition unit 110. For example, based on the determination result by the judgment unit 150A, if the user's request is a specific request, the extraction unit 160A extracts multiple specific combinations of information corresponding to the user's request from among the multiple combinations of information acquired by the information acquisition unit 110. In other words, based on the determination result by the judgment unit 150A, if the user's request is a specific request, the extraction unit 160A extracts multiple specific combinations of information consisting of combinations of values ​​of two or more variables corresponding to the user's request from among the multiple combinations of information acquired by the information acquisition unit 110. Also, for example, based on the determination result by the judgment unit 150A, if the user's request is a specific request, the extraction unit 160A extracts information corresponding to the user's request from among the multiple pieces of information acquired by the information acquisition unit 110 by analyzing the multiple pieces of information.

[0059] Furthermore, for example, based on the determination result by the determination unit 150A and the variable information generated by the LLM 41, if the user request is a specific request, the extraction unit 160A extracts multiple specific combination information from the multiple combination information acquired by the information acquisition unit 110, each consisting of a combination of values ​​of two or more variables corresponding to the user request. Furthermore, for example, based on the determination result by the determination unit 150A, if the user request is a specific request, the extraction unit 160A extracts multiple specific combination information from the multiple combination information acquired by the information acquisition unit 110 that are Pareto solutions corresponding to the user request. Furthermore, for example, based on the calculation result by the calculation unit 140, if the user request is a specific request for the information processing system 1A to extract combination information consisting of a combination of values ​​of a first variable and a second variable, and the correlation coefficient between these two variables is not positive, the extraction unit 160A extracts multiple specific combination information from the multiple combination information acquired by the information acquisition unit 110 that are Pareto solutions corresponding to the user request.

[0060] Furthermore, the information processing device may be configured to cause the LLM 41 to extract specific information corresponding to the user's request from among the multiple pieces of information acquired by the information acquisition unit 110, even if the user's request is a specific request, based on the determination result by the determination unit. For example, if the specific request is a request to cause the information processing device to extract specific combination information from among the multiple combination pieces of information acquired by the information acquisition unit 110, the information processing device may be configured to cause the LLM 41 to extract specific combination information corresponding to the user's request by inputting the request information and the multiple combination pieces of information into the LLM 41, even if the user's request is a specific request, as long as the correlation coefficient value related to the combination information calculated by the calculation unit is positive, sufficient information accuracy can be obtained even when the LLM 41 generates a response to the request.

[0061] The output unit 170A outputs the type prompt generated by the prompt generation unit 130 to the LLM 41 in order to cause the LLM 41 to generate type information. Furthermore, if the correlation coefficient calculated by the calculation unit 140 is positive, the output unit 170A outputs the multiple combination information acquired by the information acquisition unit 110 and the request information acquired by the request acquisition unit 120 to the LLM 41 in order to cause the LLM 41 to generate information indicating the combination information corresponding to the user's request from among the multiple combination information acquired by the information acquisition unit 110. The output unit 170 also outputs the variable prompt generated by the prompt generation unit 130 to the LLM 41.

[0062] The hardware configuration of the information processing device 100A is the same as that of the information processing device 100 according to Embodiment 1, so its description will be omitted.

[0063] Next, with reference to Figures 8 to 11, the details of the processing performed by the information processing device 100A will be described. Figure 9 is a flowchart showing an example of the processing performed by the information processing device 100A according to Embodiment 2. Note that some of the processing performed by the information processing device 100A according to Embodiment 2 is the same as the processing performed by the information processing device 100 according to Embodiment 1, so the same processing as in Embodiment 1 is denoted by the same reference numerals as in Embodiment 1 and its description is omitted.

[0064] As shown in Figure 9, when the information processing device 100A starts processing, it first acquires multiple pieces of information (step ST01). After performing the processing in step ST01, the information processing device 100A acquires request information (step ST02). For example, in this process, the information processing device 100A acquires request information from the input device 20, which has been operated by the user, a motor designer, using the request acquisition unit 120. This request information indicates a request from the user, stating in natural language, "Please extract three motors with short core lengths and high starting torque." More specifically, in this process, the information processing device 100A acquires request information from the input device 20, which has been operated by the user, a motor designer, using the request acquisition unit 120. This request information indicates a request from the user, stating in natural language, "Please tell me about the starting torque of the motors."

[0065] When the information processing device 100A performs the processing in step ST02, it generates a first type prompt (step ST03). In this process, the information processing device 100A generates a first type prompt to cause the LLM 41 to generate first type information indicating which of a plurality of pre-set first types the user's request falls under. For example, in this process, the information processing device 100A generates a first type prompt to cause the LLM 41 to generate first type information indicating which of a plurality of pre-set first types, "analysis," "search," and "question," the user's request falls under. Specifically, in this process, the information processing device 100A generates a prompt in natural language that causes the LLM 41 to generate first-type information indicating which of the pre-set first-types, "analysis," "search," and "question," the user's request falls under. The prompt reads the following message from the user and determines whether the user's intent requires data analysis, is a search that can be performed with filtering operations only, or is simply a question. Please answer with the result of your determination as either 'analysis,' 'search,' or 'question.' "Please extract three motors with a short core length and high starting torque." Specifically, in this process, the information processing device 100A generates a prompt in natural language that causes the LLM 41 to generate first-type information indicating which of the pre-set first-types, "analysis," "search," and "question," the user's request falls under. The prompt reads the following message from the user and determines whether the user's intent requires data analysis, is a search that can be performed with filtering operations alone, or is simply a question. Please answer with the result of your determination as either 'analysis,' 'search,' or 'question.' "Please tell me about the motor's starting torque."

[0066] When the information processing device 100A performs the processing in step ST03, it outputs a first type prompt to the LLM 41 (step ST04). In other words, in this process, the information processing device 100A inputs the first type prompt generated in the processing of step ST03 to the LLM 41, causing the LLM 41 to generate first type information, which is the answer to the first type prompt.

[0067] When the information processing device 100A performs the processing in step ST04, it obtains first type information from the LLM 41 (step ST05). In this process, the information processing device 100A obtains first type information, which indicates which of the multiple first types the user request falls under, generated by the LLM 41 as a result of the input of the first type prompt to the LLM 41 in the processing of step ST04.

[0068] When the information processing device 100A performs the processing in step ST05, it determines whether the user's request may be a specific request (step ST06). In this process, the information processing device 100A determines whether the user's request may be a specific request based on the first type information obtained in the processing in step ST05. For example, if a specific request is a request to extract specific combination information from multiple combination information obtained by the information acquisition unit 110, and the correlation coefficient calculated by the calculation unit 140 is negative or the correlation coefficient is a value indicating no correlation, even if the LLM 41 generates first type information indicating that the user's request falls under "analysis" among "analysis," "search," and "question," it cannot be immediately determined that the user's request is a specific request. Therefore, when the LLM41 generates first type information indicating that a user request falls under "analysis" among "analysis," "search," and "question," the information processing device 100A is configured to acquire second type information, which is further type information for determining whether or not the request is a specific request, as the request may be a specific request.

[0069] In the process of step ST06, if the user's request is likely to be a specific request (YES in step ST06), the information processing device 100A generates a second type prompt and a variable prompt (step ST07). In this process, based on the fact that in the process of step ST06, the LLM 41 has generated first type information indicating that the user's request falls under "analysis" among "analysis," "search," and "question," the information processing device 100A generates a second type prompt and a variable prompt, which are prompts to cause the LLM 41 to generate second type information and variable information, which are further type information for determining whether the request is a specific request.

[0070] For example, in this process, the information processing device 100A generates a second-type prompt and a variable prompt to cause the LLM 41 to generate second-type information indicating whether correlation analysis is necessary, and variable information for identifying multiple variables to be used in the correlation analysis. In other words, in this process, the information processing device 100A generates a second-type prompt and a variable prompt to cause the LLM 41 to generate second-type information indicating whether correlation analysis is required, and variable information for identifying multiple variables to be the subject of the correlation analysis. Specifically, the information processing device 100A generates a prompt in natural language as the second-type prompt and variable prompt, which reads: "Read the following message from the user and determine whether the user's intent requires correlation analysis. Also, check whether the user has mentioned multiple variables, and if the user has mentioned multiple variables, determine that correlation analysis is necessary. Also, if correlation analysis is necessary, please provide information for identifying the variables to be used in the correlation analysis. "Please extract three motors with a short core length and high starting torque." Specifically, the information processing device 100A generates a prompt in natural language as a second-type prompt and a variable prompt, which reads: "Read the following message from the user and determine whether the user's intent requires correlation analysis. Also, check whether the user is referring to multiple variables, and if the user is referring to multiple variables, determine that correlation analysis is required. If correlation analysis is required, also provide information to identify the variables to be used in the correlation analysis. 'Please tell me about the motor's starting torque.'"

[0071] When the information processing device 100A performs the processing in step ST07, it outputs a second type prompt and a variable prompt to the LLM 41 (step ST08). In other words, in this process, the information processing device 100A inputs the second type prompt and the variable prompt generated in the processing of step ST07 to the LLM 41, causing the LLM 41 to generate second type information and variable information, which are the answers to the second type prompt and the variable prompt.

[0072] When the information processing device 100A performs the processing in step ST08, it acquires second-type information and variable information from the LLM 41 (step ST09). In this process, the information processing device 100A acquires second-type information, which indicates which of the multiple second-types the user request falls under, and variable information, which indicates which two or more variables correspond to the user request, generated by the LLM 41 as a result of the second-type prompt and variable prompt being input to the LLM 41 in the processing of step ST08. In other words, in this process, the information processing device 100A acquires second-type information, which indicates whether the user request falls under a request requiring correlation analysis or a request that does not require correlation analysis, generated by the LLM 41 as a result of the second-type prompt and variable prompt being input to the LLM 41 in the processing of step ST08, and variable information, which, if the user request requires correlation analysis, indicates which two or more variables are used for correlation analysis.

[0073] Specifically, in this process, the information processing device 100A acquires Type 2 information, indicated in natural language as "correlation analysis required," and variable information, indicated in natural language as "iron core length" and "starting torque." Also, specifically, in this process, the information processing device 100A acquires Type 2 information, indicated in natural language as "correlation analysis not required," and does not acquire variable information.

[0074] When the information processing device 100A performs the processing in step ST09, it determines whether the user's request is a specific request (step ST20). In this process, the information processing device 100A determines whether the user's request is a specific request based on the second type information and variable information obtained in the processing of step ST09. For example, in this process, the information processing device 100A determines whether the user's request is a specific request based on the second type information and variable information obtained in the processing of step ST09. Specifically, in this process, the information processing device 100A determines whether the user's request is a specific request that requires correlation analysis of multiple pieces of information obtained by the information acquisition unit 110, based on the second type information and variable information obtained in the processing of step ST09.

[0075] In the process of step ST20, if the user's request is a specific request (YES in step ST20), the information processing device 100A calculates a correlation coefficient (step ST11). In this process, the information processing device 100A calculates a correlation coefficient between multiple variables identified by the variable information, based on the multiple pieces of information obtained in the process of step ST01 and the variable information obtained in the process of step ST09.

[0076] When the information processing device 100A performs the processing in step ST11, it determines whether the calculated correlation coefficient is a positive value (step ST12). In this process, the information processing device 100A determines whether it is appropriate to obtain a response to the user's request from the LLM41 by having the determination unit 150A determine whether the value of the correlation coefficient calculated in the processing of step ST11 is positive or not.

[0077] In step ST12, if the calculated correlation coefficient is not a positive value (NO in step ST12), the information processing device 100A extracts one or more combinations of information corresponding to the user's request (step ST23). In this process, based on the fact that the value of the correlation coefficient calculated in step ST11 was not positive, the information processing device 100A determines that it is not appropriate to obtain a response to the user's request from the LLM41, and extracts information indicating a response to the user's request from the multiple pieces of information obtained by the information acquisition unit 110 in step ST01 using the extraction unit 160A. For example, in this process, the information processing device 100A extracts multiple combinations of information that are Pareto solutions to each other as one or more combinations of information corresponding to the user's request using the extraction unit 160A.

[0078] Figure 10 is a scatter plot showing the information extracted by the information processing device 100A according to Embodiment 2, where the correlation coefficient is negative. Similar to Figure 5, in Figure 10, variable A represents the "shortness of the motor's core length," with the rightward movement indicating a shorter core length. Variable B represents the "magnitude of the motor's starting torque," with the upward movement indicating a larger starting torque. Each point represents a combination of motor core length and starting torque corresponding to each of the multiple pieces of information acquired by the information acquisition unit 110. Furthermore, Figure 10 shows that the starting torque tends to be larger as the core length decreases and smaller as the core length increases, and the correlation coefficient between "shortness of the core length" and "magnitude of the starting torque" is negative. For example, in the process of step ST23, the information processing device 100A extracts combination information D4, combination information D5, and combination information D6, which are Pareto solutions of each other, from a plurality of combination pieces of information obtained by the information acquisition unit 110, which consist of combinations of values ​​indicating the core length of the motors and values ​​indicating the starting torque of the motors shown in Figure 10.

[0079] When the information processing device 100A performs the processing in step ST23, it outputs the extracted information to the display device (step ST14). In this process, the information processing device 100A outputs the information extracted in the processing of step ST23, which corresponds to the user's request, using the output unit 170A, and displays it as visual information on the display device 30, thereby providing the user with a response to the user's request. The output unit 170A may be configured to display the combination information itself extracted in the processing of step ST23 on the display device 30, or it may be configured to display information for identifying a product such as a motor corresponding to the combination information extracted in the processing of step ST23, such as a product identification number, on the display device 30, or it may be configured to display a diagram, table or graph that illustrates the relationship between multiple combination information extracted in the processing of step ST23, as shown in Figure 10, on the display device 30, or it may be configured to display a pair plot diagram that includes both a scatter plot and a histogram on the display device 30.

[0080] In step ST06, if the user's request is not a specific request (NO in step ST06), in step ST20, if the user's request is not a specific request (NO in step ST20), or in step ST12, if the calculated correlation coefficient is positive (YES in step ST12), the information processing device 100A determines whether the user's request is a request relating to multiple combinations of information (step ST15). In this process, the information processing device 100A uses a determination unit 150A to determine whether the user's request is a request that requires the generation of a response based on multiple combinations of information obtained in step ST01, based on either the request information obtained in step ST02 or the variable information obtained in step ST09.

[0081] In step ST15, if the user's request concerns multiple combinations of information (YES in step ST15), the information processing device 100A outputs the multiple combinations of information obtained in step ST01 and the request information obtained in step ST02 to the LLM 41 (step ST16). In this process, the information processing device 100A inputs the multiple combinations of information obtained in step ST01 and the request information obtained in step ST02 to the LLM 41, causing the LLM 41 to generate a response corresponding to the user's request regarding multiple combinations of information.

[0082] In step ST15, if the user's request is not a request for multiple combinations of information (NO in step ST15), the information processing device 100A outputs the request information to the LLM 41 (step ST17). After performing the processing in step ST17, the information processing device 100A obtains the information generated by the LLM 41 (step ST18).

[0083] Figure 11 is a scatter plot showing the information extracted by the information processing device 100A according to Embodiment 2, where the correlation coefficient is positive. For example, in Figure 11, variable A represents the "shortness of the motor's core length," with the further to the right the motor's core length is, the shorter it is. Variable B represents the "smallness of the motor's maximum dimensions," with the further upward the motor's maximum dimensions are, the smaller they are. Each point represents a combination of the motor's core length and maximum dimensions corresponding to each of the multiple pieces of information acquired by the information acquisition unit 110. Furthermore, in Figure 11, it can be seen that there is a tendency for the maximum dimensions to be smaller as the core length is shorter, and for the maximum dimensions to be larger as the core length is longer, and the correlation coefficient between "shortness of the core length" and "smallness of the maximum dimensions" is positive. For example, in step ST18, the information processing device 100A obtains motor combination information extracted by the LLM 41 based on the request information obtained in step ST02, which states, "Please extract motors with short core lengths and small maximum dimensions," from a plurality of combination information consisting of combinations of values ​​indicating the core length of the motor and values ​​indicating the maximum dimensions of the motor, obtained in step ST01. In this way, even if the LLM 41 generates the answer to the user's request when the correlation coefficient between the variables in the combination information is positive, the accuracy of the information obtained by the user is not likely to decrease. Note that even if the correlation coefficient between the variables in the combination information is positive, the information processing device may be configured to extract the answer to the user's request by the extraction unit 160A.

[0084] When the information processing device 100A performs the processing in step ST18, it outputs the acquired information to the display device 30 (step ST19). In this process, the information processing device 100A may be configured to display the information generated by the LLM 41 itself on the display device 30, or it may be configured to display a diagram, table, or graph that illustrates the relationships between multiple combinations of information extracted in the processing of step ST23 as shown in Figure 5 on the display device 30, or it may be configured to display a pair plot diagram that includes both a scatter plot and a histogram on the display device 30.

[0085] The information processing device 100A terminates processing after performing the processing in step ST14, step ST16, or step ST19. For example, the information processing device 100A is configured to perform the processing from step ST01 to step ST19 described above each time a specific input operation to start using the information processing system 1 is performed by a user to the input device 20.

[0086] As described above, the information processing device 100A according to Embodiment 2 includes an information acquisition unit 110 that acquires a plurality of pieces of information, a request acquisition unit 120 that acquires request information that indicates a request from a user in natural language, a determination unit 150A that determines whether the request from the user is a specific request among requests for extracting specific information from a plurality of pieces of information acquired by the information acquisition unit 110 based on the request information acquired by the request acquisition unit 120, an output unit 170A that outputs the request information to the LLM 41 which generates a response to the request from the user based on the input of the request information if the request from the user is not a specific request based on the determination result of the determination unit 150, an extraction unit 160A that extracts information corresponding to the request from the user from a plurality of pieces of information acquired by the information acquisition unit 110 if the request from the user is a specific request based on the determination result of the determination unit 150, and a prompt generation unit 130 that generates a type prompt to cause the LLM 41 to generate type information that indicates which of a plurality of pre-set types the request from the user belongs to, based on the request information.

[0087] With this configuration, the information processing device 100A can improve the accuracy of determining whether a user request is a specific request for which it is inappropriate to have the LLM 41 generate a response, based on the type information, thereby improving the accuracy of the information obtained by the user when obtaining a response to a request compared to the conventional method.

[0088] Furthermore, the information processing device 100A according to Embodiment 2 includes an information acquisition unit 110 that acquires a plurality of combination information consisting of combinations of values ​​of two or more variables from a plurality of variables including a first variable and a second variable, a request acquisition unit 120 that acquires request information indicating a request from a user in natural language, a determination unit 150A that determines whether the request from the user is a specific request among requests to extract specific information from a plurality of pieces of information acquired by the information acquisition unit 110 based on the request information acquired by the request acquisition unit 120, and if the request from the user is not a specific request based on the determination result by the determination unit 150, the request information The system includes: an output unit 170A that outputs request information to an LLM 41 that generates a response to a user request based on the input of information; an extraction unit 160A that, based on the determination result by the determination unit 150, extracts a plurality of specific combination information consisting of combinations of values ​​of two or more variables corresponding to the user request from a plurality of combination information acquired by the information acquisition unit 110, if the user request is a specific request; and a prompt generation unit 130 that generates a variable prompt to cause the LLM 41 to generate variable information indicating which two or more variables correspond to the user request based on the request information.

[0089] For example, in the information processing device 100A according to Embodiment 2, the information acquisition unit 110 acquires variable information generated by the LLM 41 as a result of inputting a variable prompt generated by the prompt generation unit 130 to the LLM 41, and the extraction unit 160A is configured to extract, based on the determination result by the determination unit 150A and the variable information generated by the LLM 41, a plurality of specific combination information consisting of combinations of two or more variable values ​​corresponding to the user's request from the plurality of combination information acquired by the information acquisition unit 110, if the user's request is a specific request.

[0090] With this configuration, the information processing device 100A can suppress the extraction of variable combination information that differs from the user's request when extracting information that corresponds to a user's request based on variable information, thereby improving the accuracy of the information obtained by the user when obtaining a response to their request compared to conventional methods.

[0091] Furthermore, the information processing device 100A according to Embodiment 2 includes a calculation unit 140 that calculates a correlation coefficient between the value of a first variable and the value of a second variable based on a plurality of combination pieces of information acquired by the information acquisition unit 110, and the determination unit 150A is configured to determine whether or not a request from a user is a specific request based on the correlation coefficient calculated by the calculation unit 140.

[0092] For example, in the information processing device 100A according to Embodiment 2, the output unit 170A is configured to output the multiple combination information and request information acquired by the information acquisition unit 110 to the LLM 41 in order to cause the LLM 41 to generate information indicating the combination information corresponding to the user's request from among the multiple combination information acquired by the information acquisition unit 110 when the value of the correlation coefficient calculated by the calculation unit 140 is positive.

[0093] With this configuration, the output of information that provides answers to requests requiring complex analysis—which is generally not handled well by pre-trained models such as large-scale language models—is performed by algorithms other than the pre-trained model itself. This improves the accuracy of the information users receive when obtaining answers to their requests compared to conventional methods.

[0094] In any of the embodiments described above, the trained model stored in the model processing device only needs to have the functions of the LLM41 described above, and the trained model used in the information processing system is not limited to an LLM. For example, the trained model stored in the model processing unit may be a trained model other than an LLM that learns from training data including information expressed in natural language and information expressing the answer to that information in natural language, and generates an answer to that input in natural language based on the input of information expressed in natural language; or it may be a trained model generated by an algorithm other than a transformer, such as deep learning, that learns from training data including information expressed in natural language and information expressing the answer to that information in natural language, and generates an answer to that input in natural language based on the input of information expressed in natural language; or it may be a trained model other than an LLM that learns from training data including information other than natural language, such as information expressed in a specific computer language, or information based on pre-set rules and non-natural language information expressing the answer to that information, and generates an answer to that input as non-natural language information based on the input of non-natural language information; or, when the request acquisition unit acquires request information corresponding to the selection operation by having the user select one of the pre-set options, it may be a trained model that generates an answer to the user's request based on the input of such request information. The trained model stored in the model processing unit may be, for example, a Vision and Language Model (VLM), which is a trained model that fuses natural language and images.

[0095] Furthermore, in any of the embodiments described above, the information processing device may have some or all of the functions of components other than the information processing device in the information processing system, or some of the components of the information processing device may be provided in other devices that are communicably connected to the information processing device.

[0096] Furthermore, this disclosure allows for free combination of each embodiment, modification of any component of each embodiment, or omission of any component in each embodiment.

[0097] The information processing device relating to this disclosure can be used in a system for acquiring information indicating a user's request and providing the user with information that corresponds to the user's request.

[0098] 1 Information processing system, 1A Information processing system, 10 Storage device, 20 Input device, 30 Display device, 40 Model processing device, 41 Large-scale language model (LLM, trained model), 100 Information processing device, 100A Information processing device, 100a Processor, 100b Memory, 100c I / O port, 100d Processing circuit, 110 Information acquisition unit, 120 Request acquisition unit, 130 Prompt generation unit, 140 Calculation unit, 150 Judgment unit, 150A Judgment unit, 160 Extraction unit, 160A Extraction unit, 170 Output unit, 170A Output unit, A Variable, B Variable, C Variable, D1 Combination information, D2 Combination information, D3 Combination information, D4 Combination information, D5 Combination information, D6 Combination information.

Claims

1. An information processing device comprising: an information acquisition unit that acquires multiple pieces of information; a request acquisition unit that acquires request information indicating a request from a user in natural language; a determination unit that determines, based on the request information, whether the request from the user is a specific request among requests for extracting specific information from the multiple pieces of information acquired by the information acquisition unit; an output unit that, based on the determination result by the determination unit, outputs the request information to a trained model that generates a response to the request from the user based on the input of the request information if the request from the user is not a specific request; and an extraction unit that, based on the determination result by the determination unit, extracts information corresponding to the request from the user from the multiple pieces of information acquired by the information acquisition unit if the request from the user is a specific request.

2. The information processing apparatus according to claim 1, wherein the determination unit determines whether the request from the user is a specific request for extracting specific information by analyzing the plurality of pieces of information acquired by the information acquisition unit, and the extraction unit, based on the determination result by the determination unit, extracts information corresponding to the request from the user by analyzing the plurality of pieces of information acquired by the information acquisition unit if the request from the user was a specific request.

3. The information processing apparatus according to claim 1 or 2, characterized in that the information acquisition unit acquires a plurality of combination information consisting of a combination of the value of a first variable and the value of a second variable, the determination unit determines whether the request from the user is a specific request for extracting a plurality of specific combination information from the plurality of combination information acquired by the information acquisition unit, and the extraction unit, based on the determination result by the determination unit, extracts the plurality of specific combination information corresponding to the request from the user from the plurality of combination information acquired by the information acquisition unit if the request from the user was a specific request.

4. The information processing apparatus according to any one of claims 1 to 3, comprising: a prompt generation unit that generates a type prompt to cause the trained model to generate type information indicating which of a plurality of pre-set types the request from the user falls under based on the request information; an output unit that outputs the type prompt generated by the prompt generation unit to the trained model; an information acquisition unit that acquires the type information generated by the trained model as a result of inputting the type prompt generated by the prompt generation unit to the trained model; and a determination unit that determines whether the request from the user is a specific request based on the type information generated by the trained model.

5. The information processing apparatus according to claim 3, further comprising a calculation unit that calculates a correlation coefficient between the value of the first variable and the value of the second variable based on the plurality of combination information acquired by the information acquisition unit, wherein the determination unit determines whether the request from the user is a specific request based on the correlation coefficient calculated by the calculation unit.

6. The information processing apparatus according to claim 5, wherein, if the value of the correlation coefficient calculated by the calculation unit is positive, the output unit outputs the multiple combination information acquired by the information acquisition unit and the request information to the trained model in order to cause the trained model to generate information indicating the combination information corresponding to the request from the user among the multiple combination information acquired by the information acquisition unit.

7. The information processing apparatus according to claim 3, wherein the information acquisition unit acquires a plurality of combination information consisting of combinations of values ​​of two or more variables from a plurality of variables including the first variable and the second variable, and the extraction unit, based on the determination result by the determination unit, extracts a plurality of specific combination information consisting of combinations of values ​​of two or more variables corresponding to the request from the user from the plurality of combination information acquired by the information acquisition unit, if the request from the user is a specific request.

8. The information processing apparatus according to claim 7, comprising: a prompt generation unit that generates a variable prompt for the trained model to generate variable information indicating which of the two or more variables correspond to the request from the user, based on the request information; an output unit that outputs the variable prompt generated by the prompt generation unit to the trained model; an information acquisition unit that acquires the variable information generated by the trained model as a result of inputting the variable prompt generated by the prompt generation unit to the trained model; and an extraction unit that, based on the determination result by the determination unit and the variable information generated by the trained model, extracts a plurality of specific combination pieces of information from the plurality of combination pieces of information acquired by the information acquisition unit, each consisting of a combination of values ​​of the two or more variables corresponding to the request from the user.

9. The information processing apparatus according to claim 3, characterized in that, based on the determination result by the determination unit, the extraction unit extracts a plurality of specific combinations of information from the plurality of combinations of information acquired by the information acquisition unit, which are Pareto solutions corresponding to the user's request.

10. The information processing apparatus according to any one of claims 1 to 9, characterized in that the output unit outputs the plurality of specific combinations of information extracted by the extraction unit to the display device so as to display the plurality of specific combinations of information on the display device.

11. The information processing apparatus according to any one of claims 1 to 10, characterized in that the trained model is a transformer model that generates a response to the request from the user based on the input of the request information which indicates the request from the user in natural language.

12. An information processing system comprising: an information processing device according to claim 10; a model processing device for storing the learned model and performing processing using the learned model; and a display device.

13. A program to cause a computer to function as: an information acquisition unit that acquires multiple pieces of information; a request acquisition unit that acquires request information indicating a request from a user in natural language; a determination unit that determines, based on the request information, whether the request from the user is a specific request among requests for extracting specific information from the multiple pieces of information acquired by the information acquisition unit; an output unit that, based on the determination result by the determination unit, outputs the request information to a trained model that generates a response to the request from the user based on the input of the request information if the request from the user is not a specific request; and an extraction unit that, based on the determination result by the determination unit, extracts information corresponding to the request from the user from the multiple pieces of information acquired by the information acquisition unit if the request from the user is a specific request.

14. An information processing method performed by an apparatus comprising an information acquisition unit, a request acquisition unit, a determination unit, an output unit, and an extraction unit, the method comprising: the step of the information acquisition unit acquiring a plurality of pieces of information; the step of the request acquisition unit acquiring request information that indicates a request from a user in natural language; the step of the determination unit determining, based on the request information, whether the request from the user is a specific request among the requests for extracting specific information from the plurality of pieces of information acquired by the information acquisition unit; the step of the output unit outputting the request information to a trained model that generates a response to the request from the user based on the input of the request information, if the request from the user is not a specific request, based on the determination result by the determination unit; and the step of the extraction unit extracting information corresponding to the request from the user from the plurality of pieces of information acquired by the information acquisition unit, if the request from the user is a specific request, based on the determination result by the determination unit.

Images