Information processing apparatus, information processing method, and program

Abstract

To propose a prediction model considering special circumstances.SOLUTION: Prompt acquisition means acquires a prompt described in a natural language and including a request for information concerning a specification of a prediction task and special circumstances that may affect the prediction task. Special situation acquisition means interprets the prompt using a language model and acquires information on the special situation that can affect the designated prediction task. Model information acquisition means acquires model information related to a model that can be used when a special situation is applicable. Output means outputs information on the special situation and an answer including the acquired model information.SELECTED DRAWING: Figure 4

Description

Technical Field

[0001] The present disclosure relates to predictions using machine learning models.

Background Art

[0002] Various predictions are made using machine learning models. When a prediction error occurs during the operation of the model, the model designer or administrator will investigate the occurrence of events that did not exist during the learning of the model and the presence or absence of changes in social trends, and consider countermeasures. However, such investigations have been time-consuming and costly. Patent Document 1 discloses a method for predicting fluctuations in power demand in consideration of the influence degree of events.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] One object of the present invention is to propose a prediction model in consideration of special circumstances.

Means for Solving the Problems

[0005] In one aspect of the present disclosure, an information processing apparatus includes: prompt acquisition means for acquiring a prompt described in a natural language and including a specification of a prediction task and a request for information on special circumstances that can affect the prediction task; special circumstance acquisition means for interpreting the prompt using a language model and acquiring information on special circumstances that can affect the specified prediction task; model information acquisition means for acquiring model information on a model that can be used when the special circumstances are met; Output means for outputting an answer including information on the special circumstances and the acquired model information.

[0006] In another aspect of the present disclosure, the information processing method is executed by a computer, obtains a prompt described in natural language and including a specification of a prediction task and a request for information on special circumstances that can affect the prediction task, interprets the prompt using a language model to obtain information on special circumstances that can affect the specified prediction task, obtains model information on a model that can be used when the special circumstances apply, and outputs an answer including the information on the special circumstances and the acquired model information.

[0007] In still another aspect of the present disclosure, a program obtains a prompt described in natural language and including a specification of a prediction task and a request for information on special circumstances that can affect the prediction task, interprets the prompt using a language model to obtain information on special circumstances that can affect the specified prediction task, obtains model information on a model that can be used when the special circumstances apply, and causes a computer to execute a process of outputting an answer including the information on the special circumstances and the acquired model information.

Advantages of the Invention

[0008] According to the present disclosure, it becomes possible to propose a prediction model in consideration of special circumstances.

Brief Description of the Drawings

[0009]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

[0010] Hereinafter, preferred embodiments of the present disclosure will be described with reference to the drawings. <First Embodiment> [Overall Configuration] FIG. 1 shows the overall configuration of a machine learning model management system (hereinafter simply referred to as the "management system") to which the information processing apparatus according to the present disclosure is applied. When a user such as an administrator of a machine learning model (hereinafter also simply referred to as a "user") encounters a decrease in prediction accuracy (hereinafter also referred to as a "prediction error") during the operation of prediction processing using the machine learning model, the user requests the management system 1 for information regarding special circumstances that may affect the prediction. When there are special circumstances that may affect the prediction, the management system 1 provides information regarding the special circumstances. Further, when there is a prediction model that can be used under the special circumstances, the management system 1 proposes it. As a result, when a prediction error occurs due to special circumstances, the user can respond to the prediction error by changing the machine learning model used for prediction, etc.

[0011] As shown in FIG. 1, the management system 1 includes a server device 10 and a terminal device 20. The server device 10 and the terminal device 20 can communicate with each other via a wired or wireless network.

[0012] The terminal device 20 is operated by a user who designs and manages a machine learning model. The user designates a prediction task to the terminal device 20 and requests information on special circumstances that may affect the prediction. Although details will be described later, the terminal device 20 is equipped with a natural language model that can interpret natural language. The user inputs a prompt described in natural language to the terminal device 20. The user includes the designation of the prediction task and the request for information on special circumstances in this prompt. Note that a "prompt" refers to an instruction text to a generative AI (Artificial Intelligence) including a natural language model, etc. The terminal device 20 receives the input of the prompt, interprets the prompt using the natural language model, and recognizes the prediction task designated by the user and the request for information on special circumstances. Then, the terminal device 20 searches for and acquires special circumstances that may affect the designated prediction task using the natural language model.

[0013] The server device 10 stores prediction-related information prepared in association with various tasks in a database or the like. When the terminal device 20 finds special circumstances that can affect the prediction task specified by the user, it accesses the database of the server device 10, refers to the prediction-related information related to the prediction task specified by the user and the obtained special circumstances, and acquires information about the prediction model that can be used under those special circumstances.

[0014] Examples of the prediction tasks specified by the user include various prediction tasks such as prediction of product demand, prediction of power demand, prediction of weather, etc. In the following description, it is assumed that the prediction task specified by the user is the prediction of product demand in a certain store.

[0015] The terminal device 20 acquires information about the prediction model that can be used under special circumstances and presents it to the user. In this way, the user can obtain information about special circumstances considered to be a cause of prediction errors and the prediction model that can be used under those special circumstances by making an input in natural language.

[0016] [Natural language model] Here, the natural language model will be explained. The natural language model is one that has learned the relationships between words in a sentence and is a model that generates related strings related to the target string from the target string. By using a natural language model that has learned sentences and texts in various contexts, it is possible to generate related strings with appropriate content related to the target string. For example, the case of using a natural language model in question and answer will be explained. In this case, the natural language model receives an input of a question "What kind of country is Japan?" as the target string and generates a string such as "Japan is an island country in the Northern Hemisphere..." as an answer to the question.

[0017] The learning method of the natural language model is not particularly limited. As an example, it may be one that is learned to output at least one sentence including the input string. For a specific example, the natural language model can be GPT (Generative Pre-Training) that outputs a sentence including the input string by predicting a highly probable string following the input string. In addition to this, as the natural language model, for example, T5 (Text-to-Text Transfer Transformer), BERT (Bidirectional Encoder Representations from Transformers), RoBERTa (Robustly optimized BERT approach), ELECTRA (Efficiently Learning an Encoder that Classifies Token Replacements Accurately), etc. can also be used. In the present embodiment, a large language model may be used as the natural language model. Also, the natural language model may be one that can access the Internet or other specialized knowledge bases to obtain information.

[0018] [Hardware Configuration] (Server Device) FIG. 2 is a block diagram showing the hardware configuration of the server device 10. As shown in the figure, the server device 10 includes a processor 11, an interface (IF) 12, a ROM (Read Only Memory) 13, a RAM (Random Access Memory) 14, a database (DB) 15, and a recording medium 16. Each component is connected to each other through, for example, a bus 18.

[0019] The processor 11 is a computer such as a CPU (Central Processing Unit), and controls the entire server device 10 by executing a pre-prepared program. Specifically, as the processor 11, a CPU, GPU (Graphics Processing Unit), DSP (Digital Signal Processor), MPU (Micro Processing Unit), FPU (Floating Point number Processing Unit), PPU (Physics Processing Unit), TPU (TensorProcessingUnit), quantum processor, microcontroller, or a combination thereof can be used.

[0020] Also, the processor 11 loads the programs stored in the ROM 13 and the recording medium 16 into the RAM 14 and executes each process coded in the program. The processor 11 functions as part or all of the server device 10.

[0021] The IF 12 transmits and receives data to and from an external device. Specifically, the server device 10 transmits and receives information to and from the terminal device 20 through the IF 12.

[0022] The ROM 13 stores various programs executed by the processor 11. The RAM 14 is used as a work memory during the execution of various processes by the processor 11. The DB 15 stores prediction-related information for various tasks. Note that the prediction-related information will be described later.

[0023] The recording medium 16 is a non-volatile and non-temporary recording medium such as a disk-shaped recording medium or a semiconductor memory. The recording medium 16 may be configured to be detachable from the server device 10. The recording medium 16 stores various programs executed by the processor 11.

[0024] (Terminal device) Figure 3 is a block diagram showing the hardware configuration of the terminal device 20. The terminal device 20 is, for example, a PC or a tablet terminal. As shown in the figure, the terminal device 20 includes a processor 21, an IF 22, a ROM 23, a RAM 24, a recording medium 25, an input unit 26, and a display unit 27. Each component is connected to each other through, for example, a bus 28.

[0025] The processor 21 is a computer such as a CPU, and controls the entire terminal device 20 by executing a pre-prepared program. Note that the processor 21 may be a GPU, an FPGA, a DSP, an ASIC, or the like.

[0026] The IF 22 transmits and receives data to and from an external device. Specifically, the terminal device 20 accesses the DB 15 of the server device 10 through the IF 22 and acquires prediction-related information.

[0027] The ROM 23 stores various programs executed by the processor 21. The RAM 24 is used as a working memory during the execution of various processes by the processor 21.

[0028] The recording medium 25 is a non-volatile and non-transitory recording medium such as a disk-shaped recording medium or a semiconductor memory. The recording medium 25 may be configured to be detachable from the terminal device 20. The recording medium 25 stores various programs executed by the processor 21.

[0029] The input unit 26 is an input device such as a keyboard, a mouse, or a touch panel, for example. The user inputs a prompt including task specification and a request for presenting special circumstances by operating the input unit 26. The display unit 27 is a display or the like that performs display based on the control of the processor 21. The answer to the prompt input by the user is displayed on the display unit 27 and presented to the user.

[0030] [Functional Configuration] FIG. 4 is a block diagram showing the functional configurations of the server device 10 and the terminal device 20. The server device 10 includes a prediction model DB 151, an explanatory variable DB 152, and a special circumstance DB 153 in the DB 15. On the other hand, the terminal device 20 includes a proposal unit 28 in addition to the input unit 26 and the display unit 27 described above.

[0031] As described above, the DB 15 stores prediction-related information. The prediction-related information includes prediction model data, explanatory variable data, and special circumstance data. The prediction model DB 151 stores the prediction model data, the explanatory variable DB 152 stores the explanatory variable data, and the special circumstance DB 153 stores the special circumstance data.

[0032] FIG. 5 shows an example of the prediction model data stored in the prediction model DB 151. The prediction model DB 151 stores prediction model data indicating a prediction model corresponding to each task for each task. In the example of FIG. 5, the prediction model DB 151 stores data of a plurality of prediction models used for the task of "product demand prediction". Specifically, the prediction model data includes a "prediction model ID", an "explanatory variable", a "prediction model", and an "accuracy". The "prediction model ID" is identification information of each prediction model. The "explanatory variable" indicates a variable used in each prediction model. The "prediction model" is a mathematical expression of each prediction model. In this example, for the sake of explanation, each prediction model is shown by a mathematical expression, but the prediction model is not limited to being shown by a mathematical expression. The "accuracy" indicates the accuracy of the prediction by each prediction model.

[0033] FIG. 6 shows an example of explanatory variable data stored in the explanatory variable DB 152. The explanatory variable DB 152 stores explanatory variable data indicating explanatory variables used for prediction corresponding to each task. In the example of FIG. 6, the explanatory variable DB 152 stores a plurality of pieces of explanatory variable data used for the task of "product demand prediction". Specifically, the explanatory variable data includes an "explanatory variable", a "relationship with the target variable", and a "numerical example". The "relationship with the target variable" is information indicating the relationship between each explanatory variable and the target variable of the prediction task (product demand in this example). The "numerical example" shows an example of numerical values that each explanatory variable can take.

[0034] FIG. 7 shows an example of special circumstance data stored in the special circumstance DB 153. The special circumstance DB 153 stores special circumstance data regarding special circumstances that may affect the prediction corresponding to each task. In the example of FIG. 7, the special circumstance DB 153 stores special circumstance data that may affect the task of "product demand prediction". Specifically, the special circumstance data includes a "special circumstance ID", a "type of special circumstance", a "duration", and a "usable prediction model".

[0035] The "special circumstance ID" is identification information for each special circumstance. The "type of special circumstance" indicates the result of classifying each special circumstance based on its nature. Specifically, the type of special circumstance includes various events. When various events are held in the vicinity of the store targeted for demand prediction, the product demand fluctuates due to the event, and the demand prediction is affected. Therefore, various events are included in the special circumstances. Also, the type of special circumstance includes information dissemination by media such as SNS and TV. When a store or product is introduced on SNS, TV, etc., the product demand fluctuates due to the spread of the information, and the demand prediction is affected. Therefore, information dissemination by media is included in the special circumstances. Also, the prevalence of infectious diseases and illnesses such as the coronavirus and influenza can affect the demand for products. For example, due to the prevalence of an infectious disease, the demand for infection prevention goods such as masks and mouthwash increases. Therefore, the prevalence of infectious diseases and illnesses is included in the special circumstances.

[0036] "Duration" generally indicates the period during which each special circumstance is considered to continue. Basically, the longer the duration, the greater the variation in product demand due to special circumstances. Therefore, the duration of each special circumstance is a factor that affects demand forecasting.

[0037] "Predictive models that can be used" shows examples of predictive models that can be suitably used under each special circumstance. In the example of Fig. 7, predictive model 1 can be used during long-term and short-term sports events, and predictive model 2 can be used during music events and public events. Note that the parenthesized predictive model 1x for long-term sports events will be described later. The predictive models that can be used under each special circumstance are determined in advance based on the explanatory variables included in each predictive model and their combinations, etc. For example, although not illustrated in Fig. 7, if there are outdoor events and indoor events as special circumstances, a predictive model that includes "weather" as an explanatory variable can be set as the predictive model that can be used for outdoor events, and a predictive model that does not include "weather" as an explanatory variable can be set as the predictive model that can be used for indoor events.

[0038] [Generation of Special Circumstance Data] Next, the processing during the generation of special circumstance data as described above will be explained. When a prediction error occurs during the operation of the machine learning model, the user may think that one of the causes is not a special circumstance as described above. In this case, the user inquires of the terminal device 20 about the presence or absence of special circumstances that can affect the prediction. Specifically, the user operates the input unit 26 to input a prompt as illustrated in Fig. 8(A).

[0039] The input prompt is input to the proposal unit 28 shown in FIG. 4. The proposal unit 28 is configured using a natural language model and outputs an answer to the input prompt to the display unit 27. For example, as illustrated in FIG. 8(B), the proposal unit 28 answers that a sports event was held for three days last month. A user who sees this can infer that the demand for products has increased due to the sports event and that there may have been a prediction error. Then, the user can collect data on each explanatory variable during the period when the sports event was actually held and data such as the actual sales of the store, and create a prediction model suitable for use during the sports event by re-learning the prediction model. For example, the user can generate a prediction model 1x suitable for use during the sports event based on the prediction model 1 used during the normal period and register it in the prediction model DB 151 as shown in FIG. 5. Also, the user can register the prediction model 1x as the "usable prediction model" corresponding to the "long-term sports event" in the special circumstances DB 153 based on the answer message illustrated in FIG. 8(B) (see FIG. 7).

[0040] Note that if the user knows that the "type of special circumstances" and the "duration" can be registered in the special circumstances DB 153, the user may input a prompt as shown in FIG. 8(C) so that the answer message from the proposal unit 28 includes the type of special circumstances and the duration.

[0041] In this way, when a prediction error occurs, the user can create a prompt and inquire of the terminal device 20 to investigate whether the occurrence of special circumstances contributed to the prediction error. Then, if the answer from the terminal device 20 suggests that the occurrence of special circumstances was the cause of the prediction error, by using the data obtained during the occurrence of those special circumstances to perform retraining of the prediction model, etc., it becomes possible to generate and register a prediction model that can be suitably used under those special circumstances. In this way, as illustrated in FIG. 7, the special circumstances DB 153 is constructed. In the example of FIG. 1, one terminal device 20 is connected to the server device 10, but by connecting a plurality of terminal devices 20 used by a plurality of users to the server device 10 and sharing the DB 15 of the server device 10 among the plurality of users, it is possible to efficiently construct and update special circumstances data and prediction model data.

[0042] [Proposal of Model under Special Circumstances] Next, an example of proposing a model that can be used under special circumstances will be described. When a prediction error occurs during the operation of a machine learning model, the user considers the presence or absence of special circumstances as described above as one cause. In this case, the user inquires of the terminal device 20 about the presence or absence of special circumstances that can affect the prediction and the prediction model that can be used if there are special circumstances. For example, the user operates the input unit 26 to input a prompt as shown in FIG. 9(A).

[0043] The input prompt is input to the proposal unit 28 shown in FIG. 4. The proposal unit 28 interprets the input prompt and generates information on special circumstances that can affect the specified prediction task. Next, the proposal unit 28 accesses the special circumstances DB 153 of the server device 10 to search for a prediction model that can be used under the special circumstances. If information indicating a prediction model that can be used under the special circumstances is in the special circumstances DB 153, the proposal unit 28 acquires information indicating the usable prediction model, such as a prediction model ID. Further, the proposal unit 28 acquires information on the prediction model from the prediction model DB 151 based on the acquired prediction model ID. Then, the proposal unit 28 displays information on the special circumstances and the prediction model that can be used under the special circumstances on the display unit 27.

[0044] Now, assume that based on the prompt shown in FIG. 9(A), the proposal unit 28 has acquired information on a "4-day sports event" as a special circumstance. The proposal unit 28 first accesses the special circumstances DB 153 to determine whether information on special circumstances that match the "4-day sports event" is registered. In the example of FIG. 7, a "long-term sports event" with special circumstance ID = 1 is registered. Therefore, the proposal unit 28 determines that the "long-term sports event" with special circumstance ID = 1 matches the "4-day sports event". Next, the proposal unit 28 acquires "prediction models 1 and 1x" as prediction models that can be used in the "long-term sports event" with special circumstance ID = 1, accesses the prediction model DB 151 shown in FIG. 5, and acquires information on prediction models 1 and 1x. Then, the proposal unit 28 generates a response message including the acquired special circumstance, information on the special circumstance, and information on the model that can be used under the special circumstance, and displays it on the display unit 27. For example, the proposal unit 28 generates and displays a response message as shown in FIG. 9(B). Thereby, the user can obtain information on special circumstances considered to be the cause of prediction errors and information on prediction models that can be used under the special circumstances.

[0045] Even if the proposal unit 28 has found a special circumstance (hypothetically, "voting in an election") for the prompt shown in FIG. 9(A), but the information regarding the special circumstance is not registered in the special circumstance DB 153, the proposal unit 28 cannot propose a prediction model that can be used under that special circumstance. In this case, the proposal unit 28 displays a response message as shown in FIG. 9(C), for example. In this case, as described in the item [Generation of Special Circumstance Data] above, the user can collect data on each explanatory variable when voting in an election in the past and data such as store sales, and by performing learning of the prediction model, create a prediction model that can be used for the election voting date and register it in the prediction model DB 151 and the special circumstance DB 153.

[0046] When the proposal unit 28 has found a plurality of special circumstances for the prompt input by the user, it may obtain information on the prediction models that can be used from the special circumstance DB 153 for each special circumstance and display it on the display unit 27. Thereby, the user can consider a plurality of special circumstances as the cause of prediction errors and examine the use of a prediction model that is considered appropriate.

[0047] In this way, when a prediction error occurs, the user can input a prompt described in natural language to the terminal device 20, and obtain information regarding the presence or absence of special circumstances that may affect the prediction, and the proposal of a prediction model that can be used when there are special circumstances.

[0048] FIG. 10 shows a flowchart of the prediction model proposal process. The prediction model proposal process is a process of proposing a prediction model that can be used under a certain special circumstance, as described above. This process is realized by the processor 21 shown in FIG. 3 executing a program prepared in advance and operating as an element shown in FIG. 3.

[0049] First, when the user inputs a prompt, the proposal unit 28 receives the input prompt (step S51). Next, the proposal unit 28 searches for and acquires special circumstances related to the task specified in the prompt (step S52). Next, the proposal unit 28 accesses the special circumstances DB 153 and the prediction model DB 151 of the server device 10, and acquires information on the prediction model that can be used in the acquired special circumstances (step S53). Next, the proposal unit 28 displays the special circumstances and the information on the prediction model that can be used in the special circumstances on the display unit 27 (step S54). Then, the prediction model proposal process ends.

[0050] [Modification Example] Next, a modification example of the above embodiment will be described. The following modification examples can be applied in appropriate combinations.

[0051] (Modification Example 1) In the above embodiment, the proposal unit 28 including the natural language model is provided in the terminal device 20. Instead, one proposal unit may be provided in the server device 10, and the proposal unit may provide the user with information on special circumstances and information on the prediction model that can be used. In this case, the terminal device 20 transmits the prompt input by the user to the server device 10. In the server device 10, first, the proposal unit acquires information on special circumstances based on the prompt. Next, the proposal unit accesses the DB 15, refers to the special circumstances data and the prediction model data, and acquires information on the prediction model that can be used in the special circumstances. Then, the server device 10 transmits the acquired information on the special circumstances and the prediction model to the terminal device 20. The terminal device 20 may display the received information as a response message.

[0052] (Modification Example 2) The prediction model DB 151, the explanatory variable DB 152, and the special circumstances DB 153 stored in the DB 15 of the server device 10 may store data described in natural language. In this case, the proposal unit proposes a usable prediction model using the special circumstances data and the prediction model data described in natural language.

[0053] <Second Embodiment> FIG. 11 is a block diagram showing the functional configuration of the information processing apparatus according to the second embodiment. The information processing apparatus 70 includes a prompt acquisition unit 71, a special circumstance acquisition unit 72, a model information acquisition unit 73, and an output unit 74.

[0054] FIG. 12 is a flowchart of the processing by the information processing apparatus according to the second embodiment. The prompt acquisition unit 71 acquires a prompt described in natural language and including a request for specifying a prediction task and information on special circumstances that can affect the prediction task (step S71). The special circumstance acquisition unit 72 interprets the prompt using a language model and acquires information on special circumstances that can affect the specified prediction task (step S72). The model information acquisition unit 73 acquires model information on a model that can be used when the special circumstances apply (step S73). The output unit 74 outputs an answer including the information on the special circumstances and the acquired model information (step S74).

[0055] According to the information processing apparatus 70 of the second embodiment, it is possible to propose a prediction model in consideration of the occurrence of special circumstances.

[0056] Some or all of the above embodiments may be described as follows in the following supplementary notes, but are not limited thereto.

[0057] (Supplementary Note 1) A prompt acquisition unit that acquires a prompt described in natural language and including a request for specifying a prediction task and information on special circumstances that can affect the prediction task, A special circumstance acquisition unit that interprets the prompt using a language model and acquires information on special circumstances that can affect the specified prediction task, A model information acquisition unit that acquires model information on a model that can be used when the special circumstances apply, An output unit that outputs an answer including the information on the special circumstances and the acquired model information, An information processing apparatus comprising the same.

[0058] (Appendix 2) The model information acquisition means is the information processing apparatus described in Appendix 1 that acquires the model information from a storage unit storing special circumstance data indicating the relationship between special circumstances and models that can be used when the special circumstances apply.

[0059] (Appendix 3) The special circumstance data is described in natural language, The model information acquisition means is the information processing apparatus described in Appendix 2 that interprets information regarding the special circumstances and acquires the model information by referring to the special circumstance data.

[0060] (Appendix 4) The special circumstance data includes the type of each special circumstance, The model information acquisition means is the information processing apparatus described in Appendix 3 that acquires the model information based on the type of the special circumstance.

[0061] (Appendix 5) The prompt includes a request for information regarding the type of the special circumstance, The response includes the type of the special circumstance. The information processing apparatus described in Appendix 4.

[0062] (Appendix 6) The special circumstance data includes the duration of each special circumstance, The model information acquisition means is the information processing apparatus described in Appendix 3 that acquires the model information based on the duration of the special circumstance.

[0063] (Appendix 7) The prompt includes a request for information regarding the duration of the special circumstance, The response includes the duration of the special circumstance. The information processing apparatus described in Appendix 6.

[0064] (Appendix 8) When there is no model that can be used when the special circumstances apply, the model information acquisition means outputs a message indicating that there is no available model. The information processing apparatus described in Appendix 1.

[0065] (Appendix 9) Executed by a computer, obtain a prompt described in natural language and including a specification of a prediction task and a request for information on special circumstances that may affect the prediction task, interpret the prompt using a language model to obtain information on special circumstances that may affect the specified prediction task, obtain model information on a model that can be used when the special circumstances apply, An information processing method for outputting an answer including the information on the special circumstances and the obtained model information.

[0066] (Appendix 10) obtain a prompt described in natural language and including a specification of a prediction task and a request for information on special circumstances that may affect the prediction task, interpret the prompt using a language model to obtain information on special circumstances that may affect the specified prediction task, obtain model information on a model that can be used when the special circumstances apply, A program for causing a computer to execute a process of outputting an answer including the information on the special circumstances and the obtained model information.

[0067] Although the present disclosure has been described with reference to the embodiments and examples above, the present disclosure is not limited to the above embodiments and examples. Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present disclosure within the scope of the present disclosure.

Explanation of Reference Numerals

[0068] 10 Server device 20 Terminal device 21 Processor 26 Input unit 27 Display unit 28 Proposal unit 15 Database (DB) 151 Prediction model DB 152 Explanation Variable DB 153 Special Circumstances DB

Claims

1. Prompt acquisition means for acquiring a prompt described in natural language and including a specification of a prediction task and a request for information on special circumstances that can affect the prediction task; Special circumstance acquisition means for interpreting the prompt using a language model and acquiring information on special circumstances that can affect the specified prediction task; Model information acquisition means for acquiring model information on a model that can be used when the special circumstances apply; Output means for outputting an answer including the information on the special circumstances and the acquired model information; An information processing apparatus comprising the above.

2. The information processing apparatus according to claim 1, wherein the model information acquisition means acquires the model information from a storage unit storing special circumstance data indicating the relationship between special circumstances and a model that can be used when the special circumstances apply.

3. The special circumstance data is described in natural language, The information processing apparatus according to claim 2, wherein the model information acquisition means interprets the information on the special circumstances and acquires the model information by referring to the special circumstance data.

4. The special circumstance data includes the type of each special circumstance, The information processing apparatus according to claim 3, wherein the model information acquisition means acquires the model information based on the type of the special circumstance.

5. The prompt includes a request for information on the type of the special circumstance, The information processing apparatus according to claim 4, wherein the answer includes the type of the special circumstance.

6. The special circumstance data includes the duration of each special circumstance, The information processing apparatus according to claim 3, wherein the model information acquisition means acquires the model information based on the duration of the special circumstance.

7. The prompt includes a request for information on the duration of the special circumstance, The information processing apparatus according to claim 6, wherein the answer includes the duration of the special circumstance.

8. The information processing apparatus according to claim 1, wherein when there is no model that can be used when the special circumstances apply, the model information acquisition means outputs a message indicating that there is no available model.

9. Executed by a computer, Acquire a prompt described in natural language and including a specification of a prediction task and a request for information on special circumstances that can affect the prediction task, Interpret the prompt using a language model and acquire information on special circumstances that can affect the specified prediction task, Obtain model information regarding a model that can be used when the special circumstances apply, An information processing method for outputting an answer including information regarding the special circumstances and the obtained model information.

10. Obtain a prompt described in natural language and including a specification of a prediction task and a request for information regarding special circumstances that can affect the prediction task, Interpret the prompt using a language model to obtain information regarding special circumstances that can affect the specified prediction task, Obtain model information regarding a model that can be used when the special circumstances apply, A program for causing a computer to execute a process of outputting an answer including the information regarding the special circumstances and the obtained model information.

Images