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

Abstract

To provide an information processing device capable of reducing processing costs when providing response information indicative of a response to a prompt, an information processing method, and an information processing program.SOLUTION: An information processing device comprises an acceptance unit, a determination unit, and a provision unit. The acceptance unit accepts a query including a new prompt or a query for obtaining the new prompt from a user. The determination unit determines whether or not it is possible to respond to the new prompt using existing information. When the determination unit determines that it is possible to respond to the new prompt using the existing information, the provision unit provides the user with the existing information as response information indicative of a response to the new prompt.SELECTED DRAWING: Figure 3

Description

Technical Field

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

Background Art

[0002] In recent years, techniques for generating information using generative AI have been known. For example, Patent Document 1 discloses a technique using large language models (LLMs), which generates a prompt that adds a valid sentence as reference information to an input question sentence within a determined character limit range. A prompt is information input to generative AI, and is, for example, information indicating instructions, requests, etc. given to generative AI to execute a specific task.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, the above prior art has a problem that a processing cost is required to cause generative AI to generate response information indicating a response to a prompt.

[0005] The present application has been made in view of the above, and an object thereof is to provide an information processing apparatus, an information processing method, and an information processing program capable of reducing the processing cost when providing response information indicating a response to a prompt.

Means for Solving the Problems

[0006] The information processing apparatus according to the present application includes a reception unit, a determination unit, and a provision unit. The reception unit receives a query including a new prompt from a user or a query for obtaining a new prompt. The determination unit determines whether it is possible to respond to the new prompt using existing information that is existing information. The provision unit provides the user with the existing information as response information indicating a response to the new prompt when the determination unit determines that it is possible to respond to the new prompt using the existing information.

Effect of the Invention

[0007] According to one aspect of the embodiment, there is an effect that the processing cost when providing response information indicating a response to a prompt can be reduced.

Brief Description of the Drawings

[0008]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Embodiments for Carrying Out the Invention

[0009] Hereinafter, embodiments for carrying out the information processing apparatus, information processing method, and information processing program according to the present application (hereinafter referred to as "embodiments") will be described in detail with reference to the drawings. Note that the information processing apparatus, information processing method, and information processing program according to the present application are not limited by this embodiment. Also, in the following embodiments, the same parts are denoted by the same reference numerals, and redundant explanations are omitted.

[0010] 〔1. An Example of Information Processing〕 First, an example of information processing according to the embodiment will be described with reference to FIG. 1. FIG. 1 is a diagram for explaining the information processing according to the embodiment.

[0011] The information processing apparatus 1 shown in FIG. 1 is an information processing apparatus that cooperates with each terminal device 2 of the user U and provides various types of information to the user U online, and is realized by, for example, one or more servers or a cloud system. The terminal device 2 is, for example, a smartphone, a tablet, or a personal computer.

[0012] The information processing apparatus 1 has a plurality of generative AIs (Artificial Intelligence), and information can be provided using these plurality of generative AIs. The plurality of generative AIs in the information processing apparatus 1 are managed by the operator of the information processing apparatus 1, and in the following, the generative AI in the information processing apparatus 1 may be described as an in-house generative AI. The in-house generative AI is an example of the first generative AI.

[0013] In addition, the information processing device 1 can utilize a generative AI via an API (Application Programming Interface) for generative AI that operates in the information processing device 3. In the following, the generative AI possessed by the information processing device 3 may be described as a third-party generative AI, and the API for the generative AI may be described as a generative AI-API. The information processing device 3 is an example of another information processing device, and the third-party generative AI is an example of a second generative AI.

[0014] The generative AI is, for example, a text generation AI, an image generation AI, or a multimodal generative AI. The text generation AI is, for example, a large language model trained to estimate and output the next token from an input token sequence, such as a Transformer-based model, an RNN (Recurrent Neural Network)-based model, etc., but may also be a hybrid model thereof. Further, the text generation AI may be a composite system combined with an identification mechanism for preventing unauthorized use.

[0015] The Transformer-based model is, for example, GPT (Generative Pre-trained Transformer), PaLM2 (Pathways Language Model Version 2), etc., but is not limited to such examples. The RNN-based model is, for example, RWKV (Receptance Weighted Key Value), etc., but is not limited to such examples.

[0016] The image generation AI is an AI that generates an image from text, such as StackGAN (Generative Adversarial Networks), AttnGAN, T2I (Text-to-Image) with Transformers, Diffusion model, etc., but is not limited to such examples. Examples of the Diffusion model include DALL-E and Stable-Diffusion.

[0017] A multimodal generation AI is, for example, a generation AI that generates at least one of text, images, and audio from at least one of text, images, and audio. Multimodal generation AIs include, for example, but are not limited to, GPT-4 Turbo with vision, gemini, CM3Leon (Chameleon Multimodal Model), etc.

[0018] Note that it is desirable for the above-mentioned generation AI to be trained so as not to include personal information, etc. in its generation results.

[0019] As shown in FIG. 1, the information processing apparatus 1 receives a query including a new prompt, which is a new prompt from the user U (step S1). The user U can operate the terminal device 2 to input or select a prompt, whereby a usage request including the new prompt is transmitted from the terminal device 2 to the information processing apparatus 1. The information processing apparatus 1 receives the new prompt included in the usage request by receiving the usage request from the terminal device 2. The usage request is an example of a query.

[0020] Subsequently, when the information processing apparatus 1 receives a query including a new prompt in step S1, it selects one generation AI from a plurality of generation AIs including its own generation AI and other companies' generation AIs (step S2).

[0021] For example, the information processing apparatus 1 selects, based on the information of a plurality of items included in the new prompt, a generation AI to be used for generating response information indicating a response to the new prompt from among a plurality of generation AIs including its own generation AI and other companies' generation AIs.

[0022] Each of the plurality of items is an item indicating the type of a key item or a characteristic item, for example, an item indicating a semantic type such as a semantic field. Also, the information of an item is information indicating the item (information indicating the item itself) or information indicating the content of the item. The information indicating the content of an item is information indicating the content of a key item or information indicating the content of a characteristic item, etc.

[0023] The information processing apparatus 1 extracts the information of a plurality of items included in a new prompt by using techniques such as slot filtering. In slot filtering, each item is called a slot, but each item is not limited to a slot, and as described above, it may be an item indicating the type of a key item or a characteristic item.

[0024] For example, assume that the new prompt is a prompt input to a generation AI for generating advertisement content. In this case, the new prompt is, for example, the information of the character string "Please create advertisement content for a newly opened organic coffee shop featuring a comfortable interior. The advertisement content should be designed to target the local community."

[0025] In this case, the information of the plurality of items included in the new prompt is information indicating the content of items such as the information of the character string "comfortable interior", the information of the character string "newly opened", the information of the character string "organic coffee shop", and the information of the character string "local community". The information of the character string "comfortable interior" is information indicating the content of the item "design inside the store", and the information of the character string "newly opened" is information indicating the content of the item "status of the store".

[0026] Also, the information of the character string "organic coffee shop" is information indicating the content of the item "category of the store", and the information of the character string "local community" is information indicating the content of the item "residence of the target user". The information of the item "residence of the target user" is an example of information indicating the attributes of the target user U.

[0027] In addition, the information of a plurality of items included in the new prompt may be, instead of or in addition to the information indicating the content of the item, for example, the information indicating the item "store location", the information of the item "event", the information indicating the item "coupon or discount", the information indicating the item "product price", the item "target user age group", the information indicating the item "target user interest", and the like.

[0028] In addition, the information of the item may be, for example, the information of the item "tone", the information of the item "prompt format", and the like. The information indicating the content of the item "tone" is, for example, information such as a direct command like "do it" or an indirect command like "please do it". The information indicating the content of the item "prompt format" is, for example, information such as zero-shot, one-shot, few-shot, and the like.

[0029] The information processing device 1 extracts, for example, in a rule-based manner, the information indicating each of a plurality of items from the new prompt. For example, the information processing device 1 has dictionary information including a plurality of terms for each item, and can extract the information of a plurality of items from the new prompt using such dictionary information.

[0030] In addition, the information processing device 1 may have an extraction model that extracts the information of a plurality of items included in the new prompt, and can extract the information of a plurality of items included in the new prompt using such an extraction model.

[0031] The extraction model is a model learned to extract the information of a plurality of items from a prompt, and is a model learned to extract the information of a plurality of items from a prompt using learning information including, for each prompt, a combination of the prompt and the information of a plurality of items included in the prompt. The extraction model is, for example, an LSTM (Long Short-Term Memory) or Transformer-based model, but may be the above-described generative AI or other models.

[0032] In addition, the information processing device 1 can also extract information on a plurality of items from a new prompt using a language model such as a text generation AI. For example, the information processing device 1 can input information including instruction information indicating an instruction to extract information on a plurality of items from a new prompt and the new prompt into the text generation AI, and cause the text generation AI to extract information on a plurality of items.

[0033] In addition, information on a plurality of items is associated with each of a plurality of generation AIs including the in-house generated AI and the other company generated AI. The information processing device 1 can select, from among the plurality of generation AIs, the generation AI to be used for generating response information based on the comparison result between the information on the plurality of items associated with each of the plurality of generation AIs and the information on the plurality of items included in the new prompt.

[0034] For example, the information processing device 1 selects, from among the plurality of generation AIs, the generation AI with the highest degree of match between the information on the plurality of items associated therewith and the information on the plurality of items in the new prompt. The degree of match is, for example, the number of matches, the ratio of matches, or a value obtained by weighted addition of these between the information on the plurality of items associated therewith and the information on the plurality of items in the new prompt. However, when different weighting is performed for each item, it may also be a value obtained by adding the weights of the matching items.

[0035] In addition, the information processing device 1 can also select, from among the plurality of generation AIs, the generation AI whose degree of match between the information on the plurality of items associated therewith and the information on the plurality of items in the new prompt is equal to or higher than a threshold value. In this case, when there are a plurality of generation AIs whose degree of match between the information on the plurality of items associated therewith and the information on the plurality of items in the new prompt is equal to or higher than the threshold value, the information processing device 1 selects one generation AI from among these generation AIs randomly or according to a predetermined rule.

[0036] In addition, the information processing device 1 can also select, from among the plurality of generation AIs, the generation AI to be used for generating response information indicating a response to the new prompt based on the context of the user U and the information on the plurality of items included in the new prompt.

[0037] For example, for each of a plurality of generative AIs, in addition to information on a plurality of items, context may be associated. In this case, the information processing apparatus 1 selects, based on the comparison result between the combination of the information on a plurality of items and context associated with each of the plurality of generative AIs, the information on a plurality of items included in the new prompt, and the combination of the context of the user, a generative AI to be used for generating response information indicating a response to the new prompt.

[0038] For example, the information processing apparatus 1 selects, from among the plurality of generative AIs, the generative AI having the highest degree of match between the combination of the information on a plurality of items and context associated therewith, the information on a plurality of items included in the new prompt, and the context of the user U.

[0039] When the context includes a plurality of elements, the degree of match of the context is the number of matching elements, the ratio of the matching elements, or a value obtained by weighted addition thereof. However, when different weights are assigned to each element, it may be a value obtained by adding the weights of the matching elements. The information processing apparatus 1 selects, from among the plurality of generative AIs, the generative AI having the highest value obtained by weighted addition of the degree of match of a plurality of items and the degree of match of the context.

[0040] Further, the information processing apparatus 1 can also select, from among the plurality of generative AIs, a generative AI having a degree of match between the combination of the information on a plurality of items and context associated therewith, the information on a plurality of items included in the new prompt, and the context of the user U that is equal to or greater than a threshold value.

[0041] In this case, when there are a plurality of generative AIs having a degree of match between the combination of the information on a plurality of items and context associated therewith, the information on a plurality of items included in the new prompt, and the context of the user U that is equal to or greater than the threshold value, the information processing apparatus 1 selects one generative AI therefrom randomly or according to a predetermined rule.

[0042] The context of user U is the context when the advertising content is provided to user U by the information processing apparatus 1. The context of user U is the current situation of user U and the situation surrounding user U, etc.

[0043] For example, the elements of the context of user U include the attributes of user U, the current location of user U, the current time, the physical environment where user U is located, the social environment where user U is located, the motion state of user U, and the emotion of user U, etc. The attributes of user U are, for example, demographic attributes, psychographic attributes, etc.

[0044] Demographic attributes are demographic attributes and include a plurality of attribute elements such as age, gender, occupation, place of residence, annual income, family composition, etc. Psychographic attributes are psychological attributes and include a plurality of attribute elements related to, for example, interests, lifestyle, values, etc. The attributes of user U, when the attribute is a demographic attribute, are, for example, gender, age (age group), place of residence, occupation, or a combination of two or more of these, but are not limited to such examples.

[0045] The physical environment where user U is located is, for example, temperature, humidity, weather, illuminance, indoor, outdoor, or a combination of two or more of these, but is not limited to such examples. The social environment where user U is located is, for example, the economic situation, the political situation, popular goods and services, or a combination of two or more of these, but is not limited to such examples.

[0046] The motion state of user U is, for example, a running state, a walking state, a sitting state, etc., but is not limited to such examples. Also, the emotion of user U is, for example, a laughing state, an angry state, a troubled state, etc., but is not limited to such examples.

[0047] Note that the context of the user U may be, for example, a combination of two or more of the attributes of the user U, the current location of the user U, the current time, the physical environment in which the user U is placed, the social environment in which the user U is placed, the motion state of the user U, and the emotion of the user U.

[0048] The information processing apparatus 1 can acquire information indicating the attributes of the user U associated with the identification information of the user U included in the usage request from an internal storage unit or an external information processing apparatus. The identification information of the user U is, for example, the ID (Identifier) of the user U or a cookie of the browser of the terminal device 2, but is not limited to such examples.

[0049] In addition, the information processing apparatus 1 can acquire information indicating the physical environment in which the user U is placed, information indicating the social environment in which the user U is placed, information indicating the current location of the user U, information indicating the motion state of the user U, and information indicating the emotion of the user U from the terminal device 2 or an external information processing apparatus.

[0050] In addition, the information processing apparatus 1 has a selection model, which is a model for selecting one of the self-generated AI and the other-generated AI and is obtained by learning, and can select one of the self-generated AI and the other-generated AI using such a selection model.

[0051] For example, the information processing apparatus 1 can generate a selection model by using, as learning information, information including, for each prompt, the information of the prompt and information indicating the evaluation by the user U of the provided information generated using the generation AI based on the information including this prompt and provided to the user U.

[0052] In this case, the information of the prompt is the prompt or a plurality of terms included in the prompt. The information indicating the evaluation is, for example, the evaluation value in a five-level evaluation. The information indicating the evaluation is, for example, multiplied by +1 for one of the evaluation values when using the in-house generated AI and multiplied by -1 for the other, and is used for generating the selection model. The provided information is the generated information generated in the past using the generative AI or the information based on this generated information.

[0053] The information processing device 1 inputs the information of the new prompt into the selection model, and determines whether to use the in-house generated AI or the other company's generated AI depending on whether the value output from such a selection model is a positive value or a negative value.

[0054] Further, the information processing device 1 may have a selection model that is a model obtained by learning and is a model for determining whether a response to a new prompt is possible using the generated information generated using the in-house generated AI. In this case, the information processing device 1 can also determine whether a response to a new prompt is possible using the generated information generated using the in-house generated AI, using the selection model.

[0055] The selection model is, for example, a model learned using learning information including, for each prompt, the prompt input to the in-house generated AI in the past and the evaluation information indicating the evaluation by the user U of the provided information generated using such a prompt. The evaluation information is information indicating the evaluation by the user U who provided the generated information generated by the in-house generated AI or the information based on such generated information as the provided information, and is, for example, information indicating the evaluation value of a five-level evaluation.

[0056] In this case, when the information processing device 1 inputs the new prompt or each term included in the new prompt into the selection model, if the value output from the selection model is equal to or greater than the threshold value, it is determined that a response to the new prompt is possible using the generated information generated using the in-house generated AI.

[0057] Further, when there are multiple in-house generated AIs, the selection model may be a model that inputs a new prompt or each term included in the new prompt and outputs a value (score) for each in-house generated AI. The information processing apparatus 1 can select the in-house generated AI corresponding to the highest value whose value output from the selection model is equal to or greater than the threshold value.

[0058] Note that when the information processing apparatus 1 includes information specifying a generated AI in the usage request, the information processing apparatus 1 can also select the generated AI specified in the usage request from among the plurality of generated AIs.

[0059] When the information processing apparatus 1 selects an externally generated AI from among the plurality of generated AIs in step S2, the information processing apparatus 1 determines whether it is possible to respond to the new prompt using past generated information, which is information generated using an externally generated AI in the past (step S3A). The past generated information is an example of existing information.

[0060] For example, the information processing apparatus 1 determines whether it is possible to respond to the new prompt using the past generated information based on the prompt comparison result, which is the result of comparing the past prompt, which is the past prompt used to generate the past generated information, with the new prompt.

[0061] The information processing apparatus 1 calculates a score indicating the comparison result between the past prompt used to generate the past generated information and the new prompt, and determines that it is possible to respond to the new prompt when the calculated score satisfies a predetermined condition (for example, when it is equal to or greater than a threshold value or equal to or less than a threshold value). The score indicating the comparison result between the past prompt and the new prompt is indicated by, for example, the similarity or dissimilarity of the vectorized prompts. The vectorization of the prompt is performed, for example, by embedding (Embedding) using a sentence embedding model (for example, a transformer-based model).

[0062] Embedding by a text embedding model is, for example, embedding by text-embedding-ada provided by OpenAI, BERT, etc., but is not limited to such examples. Note that the vectorization of content is not limited to embedding by a text embedding model, and for example, content may be vectorized by Doc2Vec, the average of word embeddings, etc. For word embeddings, for example, Word2Vec, fastText, etc. are used.

[0063] The information processing device 1 vectorizes each past prompt and vectorizes the new prompt. Then, the information processing device 1 determines whether there is a past prompt whose similarity to the new prompt is equal to or greater than a threshold or whether there is a past prompt whose difference from the new prompt is equal to or less than a threshold. When the information processing device 1 determines that there is a past prompt whose similarity to the new prompt is equal to or greater than the threshold or a past prompt whose difference from the new prompt is equal to or less than the threshold, it determines that a response to the new prompt can be made using the past generated information generated by the generation AI using such a past prompt.

[0064] The similarity between the new prompt and the past prompt is the cosine similarity, but it may also be the Jaccard similarity, etc., or may be the reciprocal of the Euclidean distance, the reciprocal of the Manhattan distance, etc. Also, the similarity between the new prompt and the past prompt is, for example, the reciprocal of the similarity between the new prompt and the past prompt.

[0065] Also, the information processing device 1 can also extract information on a plurality of items from the past prompts. The information processing device 1 can extract information on a plurality of items from each past prompt using the dictionary information, extraction model, or generation AI described above.

[0066] In this case, when the information processing apparatus 1 determines that there is a past prompt whose degree of match with the information of a plurality of items of the new prompt is equal to or higher than a threshold value, it determines that a response to the new prompt can be made using the past generated information generated by the generation AI using such a past prompt.

[0067] Further, the information processing apparatus 1 has a determination model obtained by learning, which is a model for determining whether a response to a new prompt can be made using past generated information, and can determine whether a response to the new prompt can be made using such a determination model.

[0068] For example, the information processing apparatus 1 can generate a determination model by using, as learning information, information including, for each prompt, the information of the prompt and information indicating an evaluation by the user U of the provided information generated using the other company's generation AI and provided to the user U based on the information including this prompt.

[0069] In this case, the information of the prompt is the prompt or a plurality of terms included in the prompt. The information indicating the evaluation is, for example, an evaluation value in a five-level evaluation. The provided information is past generated information generated using the other company's generation AI or information based on this past generated information.

[0070] The information processing apparatus 1 inputs the information of the new prompt into the determination model, and when the value output from such a determination model is equal to or higher than a threshold value, it determines that a response to the new prompt can be made using past generated information.

[0071] When the information processing apparatus 1 determines that a response to the new prompt can be made using past generated information, it determines the past generated information associated with the past prompt having the highest similarity or degree of match among the above-described new prompts. The past generated information associated with the past prompt is information generated by the other company's generation AI when the information including the past prompt was input to the other company's generation AI in the past.

[0072] When the information processing apparatus 1 selects an AI-generated by another company from among a plurality of generative AIs in step S2 and determines that it is possible to respond to a new prompt using past generation information, the information processing apparatus 1 provides the user U with the past generation information or information based on the past generation information as response information indicating a response to the new prompt (step S4A-1). In this case, the past generation information is past generation information that information including the target prompt was input as input information to the AI-generated by another company and output from the AI-generated by another company.

[0073] The target prompt is a past prompt whose vector similarity to the new prompt is equal to or greater than a threshold value or a past prompt whose degree of coincidence of information of a plurality of items is equal to or greater than a threshold value. When there are a plurality of past prompts whose similarity or degree of coincidence is equal to or greater than the threshold value, the target prompt is the past prompt with the highest similarity or degree of coincidence or a past prompt selected randomly or according to a predetermined rule from among a plurality of past prompts whose similarity or degree of coincidence is equal to or greater than the threshold value.

[0074] Note that the target prompt may be a past prompt whose score obtained by weighted addition of the similarity and the degree of coincidence is equal to or greater than a threshold value. In this case, in step S3A, the information processing apparatus 1 determines that it is possible to respond to a new prompt using past generation information when there is a past prompt whose score is equal to or greater than the threshold value.

[0075] The past generation information is information output from the AI-generated by another company when information including a prompt from the user U is input to the AI-generated by another company. However, when the prompt from the user U is a scenario or the like, the past generation information may be information output from the AI-generated by another company by repeatedly inputting a dialogue history including such a prompt and the history of information output from the AI-generated by another company to the AI-generated by another company.

[0076] The scenario includes, for example, but is not limited to, part or all of string information indicating the definition of a task executed by the generative AI, the constraint conditions of the task in the generative AI, the definition of the behavior and tone of the generative AI, the output format of the generative AI, and the like.

[0077] The information based on the past generated information is, for example, when the past generated information is intention information including information indicating the intention type and information indicating the intention content, the information collected by the information processing apparatus 1 or the information processing apparatus 3 based on the information indicating the intention type and the information indicating the intention content.

[0078] The intention type is, for example, the type of intention of additional information, and the information indicating the intention type is, for example, information specifying a function or function corresponding to the intention type. Further, the intention content is the content of the intention of the additional information, and the information indicating the intention content is, for example, information indicating an argument of a function or a parameter of a function corresponding to the intention type.

[0079] When information including intention definition information for extracting the intention type and the intention content in addition to the prompt from the user U is input as input information, the other company generated AI outputs intention information. When the generation AI-API is an API provided by OpenAI (registered trademark) company, the information processing apparatus 1 can cause the other company generated AI to generate intention information by using the function of function calling.

[0080] Further, the past generated information or the information based on the past generated information may be information output from the other company generated AI by inputting information including the collected information collected as described above and instruction information for instructing processing on such collected information to the other company generated AI. Note that the information based on the past generated information is information generated in the past, but may also be newly generated information.

[0081] Also, when the information processing apparatus 1 selects an AI generated by another company from among a plurality of generative AIs in step S2 and determines that it is not possible to respond to a new prompt using past generation information, the information including the new prompt is input as input information to the AI generated by another company via the generative AI-API, and the response information indicating the response to the new prompt or the information for generating the response information is generated as new generation information by the AI generated by another company, and such new generation information is acquired via the generative AI-API (step S4A-2). The information for generating the response information is, for example, the intention information described above. The method for generating the new generation information is the same as the method for generating the past generation information described above.

[0082] Then, the information processing apparatus 1 provides the user U with the new generation information acquired in step S4A-2 or the information based on the new generation information as response information (step S4A-3). The method for generating the information based on the new generation information is the same as the method for generating the information based on the past generation information described above.

[0083] Also, when the information processing apparatus 1 selects its own generative AI from among a plurality of generative AIs in step S2, the information including the new prompt is input to the selected generative AI of the company itself, and the response information indicating the response to the new prompt or the information for generating such response information is generated as new generation information by the generative AI of the company itself (step S3B). The information for generating the response information is, for example, the intention information described above. The method for generating the new generation information is the same as the method for generating the past generation information described above, but is not limited to such an example.

[0084] Then, the information processing apparatus 1 provides the user U with the new generation information acquired in step S3A or the information based on the new generation information as response information (step S4B). The method for generating the information based on the new generation information is the same as the method for generating the information based on the past generation information described above, but is not limited to such an example.

[0085] Note that when the information processing apparatus 1 selects its own generated AI from among a plurality of generated AIs in step S2, it can also determine whether it is possible to respond to a new prompt using past generated information, which is information generated in the past using the selected own generated AI.

[0086] In this case, when the information processing apparatus 1 determines that it is possible to respond to a new prompt using past generated information, it provides the user U with the past generated information or information based on the past generated information as response information indicating a response to the new prompt. Otherwise, it performs the processes of steps S3B and S4B.

[0087] In the example described above, the information processing apparatus 1 receives a query including a new prompt, but instead of or in addition to the query including the new prompt, it can receive a query for obtaining a new prompt. In this case, the information processing apparatus 1 processes or treats the information included in the query for obtaining a new prompt to obtain a new prompt.

[0088] For example, the information processing apparatus 1 has dictionary information associating one or more keywords with a prompt for each of the one or more keywords. The information processing apparatus 1 extracts one or more keywords from the information included in the query for obtaining a new prompt, and obtains as the new prompt the prompt associated with the extracted one or more keywords.

[0089] Further, the information processing apparatus 1 may have dictionary information associating vector information with a prompt. In this case, the information processing apparatus 1 vectorizes the information included in the query for obtaining a new prompt, and can obtain as the new prompt the prompt associated with the vector information having the highest similarity to the vectorized information or the vector information equal to or higher than a threshold value. The vector information is information obtained by vectorizing the corresponding prompt, but may be information obtained by vectorizing information different from the corresponding prompt.

[0090] Also, in the above example, the information processing apparatus 1 determines whether it is possible to respond to a new prompt using past generated information, but it can also determine whether it is possible to respond to a new prompt using existing information other than past generated information. The existing information other than past generated information is, for example, information other than the information generated using a past generation AI, such as information generated by an AI other than the generation AI or information created by a person.

[0091] In this case, the information processing apparatus 1 has, for example, dictionary information including information obtained by combining a past prompt and existing information for each past prompt, and determines whether it is possible to respond to a new prompt using existing information other than past generated information based on the comparison result between the new prompt and the past prompt.

[0092] Also, the information processing apparatus 1 can determine whether it is possible to respond to a new prompt using a model that determines whether it is possible to respond to a new prompt using existing information other than past generated information and that is obtained by learning.

[0093] Also, the information processing apparatus 1 can calculate the degree of coincidence between the information of a plurality of items of the past prompt and the information of a plurality of items of the new prompt as a score indicating the comparison result between the past prompt and the new prompt. Also, the information processing apparatus 1 can calculate the degree of difference between the information of a plurality of items of the past prompt and the information of a plurality of items of the new prompt as a score indicating the comparison result between the past prompt and the new prompt. In this case, the information processing apparatus 1 determines that it is possible to respond to the new prompt when the calculated score satisfies a predetermined condition (for example, when the degree of coincidence is equal to or greater than a threshold value or the degree of difference is equal to or less than a threshold value).

[0094] In addition, the information processing apparatus 1 can also calculate a value obtained by weighted addition of the degree of coincidence and the degree of difference between the information of a plurality of items of the past prompt and the information of a plurality of items of the new prompt as a score indicating the comparison result between the past prompt and the new prompt. In this case, when the calculated score is less than or equal to a threshold value or greater than or equal to the threshold value, the information processing apparatus 1 determines that a response to the new prompt is possible.

[0095] Also, in the above-described example, the information processing apparatus 1 selects, from among a plurality of generation AIs, a generation AI to be used for generating response information based on the degree of coincidence between the information of a plurality of items associated therewith and the information of a plurality of items of the new prompt, but is not limited to such an example. For example, the information processing apparatus 1 can also select, from among a plurality of generation AIs, a generation AI to be used for generating response information based on the degree of difference between the information of a plurality of items associated therewith and the information of a plurality of items of the new prompt.

[0096] For example, the information processing apparatus 1 can select, from among a plurality of generation AIs, a generation AI having the lowest degree of difference or a generation AI having a degree of difference less than or equal to a threshold value between the information of a plurality of items associated therewith and the information of a plurality of items of the new prompt as a generation AI to be used for generating response information. Also, the information processing apparatus 1 can also select, from among a plurality of generation AIs, a generation AI to be used for generating response information based on the result of weighted addition of the degree of coincidence and the degree of difference.

[0097] Also, in the above-described example, the information processing apparatus 1 selects, from among a plurality of generation AIs, a generation AI to be used for generating response information indicating a response to the prompt based on the information of a plurality of items included in the prompt, but is not limited to such an example. For example, the information processing apparatus 1 can also select an AI other than the generation AI. The AI other than the generation AI may be, for example, a natural language processing model that analyzes, classifies, or responds to the prompt, or may be an AI agent that performs processing according to the prompt. In this case, the information processing apparatus 1 selects, from among a plurality of AIs, an AI to be used for generating response information indicating a response to the prompt based on the information of a plurality of items included in the prompt.

[0098] In this way, the information processing apparatus 1 receives a query including a new prompt from the user U or a query for obtaining a new prompt, determines whether it is possible to respond to the new prompt using existing information which is the existing information, and when it is determined that it is possible to respond to the new prompt using the existing information, provides the user U with the existing information as response information indicating the response to the new prompt. Thereby, the information processing apparatus 1 can reduce the processing cost when providing the response information indicating the response to the prompt.

[0099] Also, the information processing apparatus 1 receives a query including a prompt from the user U or a query for obtaining a prompt, selects, from among a plurality of AIs, an AI to be used for generating response information indicating a response to the prompt based on information of a plurality of items included in the prompt, and provides the user U with the response information generated using the selected AI. Thereby, the information processing apparatus 1 can improve the accuracy of appropriately obtaining a response to the prompt.

[0100] Hereinafter, the configuration of the information processing system including the information processing apparatus 1 and the plurality of terminal devices 2 that perform such processing will be described in detail.

[0101] 〔2. Configuration of Information Processing System〕 FIG. 2 is a diagram showing an example of the configuration of the information processing system according to the embodiment. As shown in FIG. 2, the information processing system 100 according to the embodiment includes an information processing apparatus 1, a plurality of terminal devices 2, and an information processing apparatus 3.

[0102] The plurality of terminal devices 2 are used by different users U. The terminal device 2 is, for example, a notebook PC (Personal Computer), a desktop PC, a smartphone, a tablet PC, or a wearable device. The wearable device is, for example, smart glasses or a smart watch, but is not limited to such examples.

[0103] Each of the information processing apparatus 1, the terminal apparatus 2, and the information processing apparatus 3 is connected to be communicable with each other by wire or wirelessly via the network N. Note that the information processing system 100 shown in FIG. 2 may include a plurality of information processing apparatuses 1 and information processing apparatuses 3 and the like.

[0104] The network N includes, for example, a WAN (Wide Area Network) such as the Internet and a mobile communication network such as LTE (Long Term Evolution), 4G (4th Generation), and 5G (5th Generation: the 5th generation mobile communication system).

[0105] The terminal apparatus 2 can be connected to the network N via a mobile communication network, short-range wireless communication such as Bluetooth (registered trademark), or wireless LAN (Local Area Network), and communicate with the information processing apparatus 1, the information processing apparatus 3, and the like.

[0106] 〔3. Configuration of Information Processing Apparatus 1〕 FIG. 3 is a diagram showing an example of the configuration of the information processing apparatus 1 according to the embodiment. As shown in FIG. 3, the information processing apparatus 1 includes a communication unit 10, a storage unit 11, and a processing unit 12.

[0107] 〔3.1. Communication Unit 10〕 The communication unit 10 is realized by, for example, a communication module or a NIC (Network Interface Card). Then, the communication unit 10 is connected to the network N by wire or wirelessly, and transmits and receives information to and from other various devices. For example, the communication unit 10 transmits and receives information to and from the terminal apparatus 2 and the information processing apparatus 3 via the network N.

[0108] 〔3.2. Storage Unit 11〕 The storage unit 11 is implemented by, for example, a semiconductor memory element such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk. The storage unit 11 includes a user information storage unit 20, a past prompt storage unit 21, and a generation AI related information storage unit 22.

[0109] 〔3.2.1. User Information Storage Unit 20〕 The user information storage unit 20 stores user information including information about the user U. FIG. 4 is a diagram showing an example of a user information table stored in the user information storage unit 20 of the information processing apparatus 1 according to the embodiment. As shown in FIG. 4, the user information table stored in the user information storage unit 20 includes items such as "user ID" and "context".

[0110] The "user ID" is identification information for identifying the user U. The "context" is information indicating the context of the user U corresponding to the "user ID". The context of the user U is the current situation of the user U or the situation surrounding the user U.

[0111] For example, the elements of the context of the user U include the attributes of the user U, the current location of the user U, the current time, the physical environment in which the user U is placed, the social environment in which the user U is placed, the movement state of the user U, and the emotions of the user U. The attributes of the user U are, for example, demographic attributes, psychographic attributes, etc.

[0112] Demographic attributes are demographic attributes and include, for example, a plurality of attribute elements such as age, gender, occupation, place of residence, annual income, and family composition. Psychographic attributes are psychological attributes and include, for example, a plurality of attribute elements related to interests, lifestyle, values, etc. The attributes of the user U are, when the attribute is a demographic attribute, for example, gender, age (age group), place of residence, occupation, or a combination of two or more of these, but are not limited to such examples.

[0113] The physical environment where the user U is located is, for example, temperature, humidity, weather, illuminance, indoor, outdoor, or a combination of two or more of these, but is not limited to such examples. The social environment where the user U is located is, for example, economic situation, political situation, popular goods and services, or a combination of two or more of these, but is not limited to such examples.

[0114] The motion state of the user U is, for example, a running state, a walking state, a sitting state, etc., but is not limited to such examples. Also, the emotion of the user U is, for example, a laughing state, an angry state, a troubled state, etc., but is not limited to such examples.

[0115] Note that the context of the user U may be, for example, a combination of two or more of the attributes of the user U, the current position of the user U, the current time, the physical environment where the user U is located, the social environment where the user U is located, the motion state of the user U, and the emotion of the user U.

[0116] 〔3.2.2. Past Prompt Memory Unit 21〕 The past prompt memory unit 21 stores various information related to past prompts. A past prompt is a prompt that was received in the past by the reception unit 31 and input to the generation AI to generate generated information.

[0117] FIG. 5 is a diagram showing an example of a past prompt table stored in the past prompt memory unit 21 of the information processing apparatus 1 according to the embodiment. In the example shown in FIG. 5, the past prompt table stored in the past prompt memory unit 21 includes information on items such as "prompt ID", "past prompt", "past generated information", and "generation AI ID".

[0118] The "prompt ID" is an identifier for identifying a past prompt and is information assigned to each past prompt. The "past prompt" is the past prompt corresponding to the "prompt ID".

[0119] "Past generated information" is past generated information that is information output by the generative AI in the past after information including the past prompt corresponding to the "prompt ID" is input to the generative AI. The past generated information is past generated information of the company's own generative AI or past generated information of another company's generative AI.

[0120] "Generative AI ID" is an identifier that identifies the generative AI into which information including the past prompt corresponding to the "prompt ID" is input, and is information assigned to each generative AI.

[0121] [3.2.3. Generative AI related information storage unit 22] The generative AI related information storage unit 22 stores various information about each of the plurality of generative AIs. The plurality of generative AIs includes a plurality of the company's own generative AIs and other companies' generative AIs. FIG. 6 is a diagram showing an example of a generative AI related information table stored in the generative AI related information storage unit 22 of the information processing apparatus 1 according to the embodiment.

[0122] In the example shown in FIG. 6, the generative AI related information table stored in the generative AI related information storage unit 22 includes information on items such as "generative AI ID", "generative AI name", "plurality of item information", and "context". "Generative AI ID" is an identifier that identifies the generative AI, and is information assigned to each generative AI. "Generative AI name" is information indicating the name of the generative AI corresponding to the "generative AI ID".

[0123] The generative AI is, for example, a text generative AI, an image generative AI, or a multimodal generative AI. The text generative AI is, for example, a large language model trained to estimate and output the next token from the input token sequence, and is, for example, a transformer-based model, an RNN-based model, etc., but may also be a hybrid model thereof. Further, the text generative AI may be a composite system combined with an identification mechanism for preventing unauthorized use.

[0124] Transformer-based models include, for example, but are not limited to, GPT, PaLM2, etc. RNN-based models include, for example, but are not limited to, RWKV, etc.

[0125] Image generation AI is AI that generates images from text, and includes, for example, but is not limited to, StackGAN, AttnGAN, T2I with Transformers, Diffusion models, etc. Examples of Diffusion models include DALL-E and Stable-Diffusion.

[0126] Multimodal generation AI is generation AI that generates at least one of text, images, and audio from at least one of text, images, and audio. Multimodal generation AI includes, for example, but is not limited to, GPT-4 Turbo with vision, gemini, CM3Leon, etc.

[0127] It is desirable that the above-mentioned generation AI be trained so as not to include personal information or the like in its generation results.

[0128] "Multiple-item information" is information on multiple items associated with the generation AI corresponding to the "generation AI ID". Each of the multiple items is an item indicating the type of the key item or the characteristic item, and is, for example, an item indicating a semantic type such as a semantic field. Also, the information on the item is information indicating the item (information indicating the item itself) or information indicating the content of the item. The information indicating the content of the item is information indicating the content of the key item or information indicating the content of the characteristic item, etc.

[0129] "Context" is information indicating the context associated with the generation AI corresponding to the "generation AI ID". The context is, for example, information indicating the context of the user U to which the generation AI is applied. The context of the user U is the current situation of the user U or the situation surrounding the user U, etc.

[0130] [3.3. Processing Unit 12] The processing unit 12 is a controller and is realized, for example, by a processor such as a CPU (Central Processing Unit) or an MPU (Micro Processing Unit), by executing various programs (corresponding to an example of an information processing program) stored in a storage device inside the information processing apparatus 1 using a RAM or the like as a work area.

[0131] Also, the processing unit 12 is a controller and may be realized by an integrated circuit such as an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or a GPGPU (General Purpose Graphic Processing Unit).

[0132] As shown in FIG. 3, the processing unit 12 includes an acquisition unit 30, a reception unit 31, a selection unit 32, a determination unit 33, a provision unit 34, and a learning unit 35, and realizes or executes the functions and operations of information processing described below. Note that the internal configuration of the processing unit 12 is not limited to the configuration shown in FIG. 3, and may be any other configuration as long as it can perform the information processing described later.

[0133] [3.3.1. Acquisition Unit 30] The acquisition unit 30 acquires various information from an external device or the like via the network N and the communication unit 10, or acquires various information from the storage unit 11.

[0134] For example, the acquisition unit 30 acquires information of the user U from the terminal device 2 or an external information processing device, and stores the acquired information of the user U in the user information storage unit 20. The information of the user U is, for example, the context of the user U.

[0135] In addition, the acquisition unit 30 acquires various types of information from the user information storage unit 20, the past prompt storage unit 21, the generative AI-related information storage unit 22, and the like.

[0136] [3.3.2. Reception Unit 31] The reception unit 31 receives a query including a prompt from the user U or a query for obtaining a new prompt. For example, the reception unit 31 receives a new prompt, which is a new prompt from the user U.

[0137] When the reception unit 31 receives a query for obtaining a new prompt, it processes or treats the information included in the query for obtaining the new prompt to obtain the new prompt. For example, the reception unit 31 has dictionary information associating one or more keywords with a prompt for each of the one or more keywords. The reception unit 31 extracts one or more keywords from the information included in the query for obtaining the new prompt, and obtains the prompt associated with the extracted one or more keywords as the new prompt.

[0138] In addition, the reception unit 31 may have dictionary information associating vector information with a prompt. In this case, the reception unit 31 vectorizes the information included in the query for obtaining the new prompt, and can obtain, as the new prompt, the prompt associated with the vector information having the highest similarity to the vectorized information or the vector information equal to or higher than a threshold value. The vector information is information obtained by vectorizing the corresponding prompt, but may be information obtained by vectorizing information different from the corresponding prompt.

[0139] Note that the process of processing or treating the information included in the query for obtaining the new prompt to obtain the new prompt is not limited to the process by the reception unit 31, and may be, for example, the process of the acquisition unit in the processing unit 12.

[0140] The user U can operate the terminal device 2 to input or select information for obtaining a prompt, whereby a usage request including a new prompt or a usage request for obtaining a new prompt is transmitted from the terminal device 2 to the information processing device 1. The reception unit 31 receives the new prompt included in the usage request by receiving the usage request from the terminal device 2. The usage request including the new prompt or the usage request for obtaining a new prompt is an example of a query including the above-described prompt or a query for obtaining a new prompt.

[0141] In addition, the reception unit 31 receives evaluation information which is information indicating an evaluation by the user U of the provided information provided by the provision unit 34. The user U can operate the terminal device 2 to input information indicating an evaluation of the provided information provided by the provision unit 34. Thereby, the evaluation information transmitted from the terminal device 2 to the information processing device 1 is, for example, information indicating an evaluation value of a five-level evaluation, but is not limited to such an example.

[0142] [3.3.3. Selection unit 32] The selection unit 32 selects, based on information of a plurality of items included in the new prompt received by the reception unit 31, a generation AI to be used for generating response information indicating a response to the new prompt from among a plurality of generation AIs.

[0143] The plurality of generation AIs include a self-generation AI which is a generation AI included in the information processing device 1 and an other-company generation AI which is a generation AI included in the information processing device 3. The self-generation AI is an example of a first generation AI, and the other-company generation AI is an example of a second generation AI. The information processing device 3 is an example of another information processing device different from the information processing device 1.

[0144] Each of the plurality of items is an item indicating a type of a key matter or a characteristic matter, for example, an item indicating a semantic type such as a semantic field. Also, the information of the item is information indicating the content of the key matter or information indicating the content of the characteristic matter.

[0145] The selection unit 32 extracts information on a plurality of items included in the new prompt using techniques such as slot filtering. In slot filtering, each item is called a slot, but each item is not limited to a slot, and as described above, any item that indicates the type of key item or characteristic item may be used.

[0146] For example, assume that the new prompt is a prompt input to the generation AI to generate advertisement content. In this case, the new prompt is, for example, the information of the character string "Please create advertisement content for a newly opened organic coffee shop featuring a comfortable interior. The advertisement content should be designed to target the local community."

[0147] In this case, the information on a plurality of items included in the new prompt is information indicating the content of items such as the information of the character string "comfortable interior", the information of the character string "newly opened", the information of the character string "organic coffee shop", and the information of the character string "local community". The information of the character string "comfortable interior" is information indicating the content of the item "interior design of the store", and the information of the character string "newly opened" is information indicating the content of the item "status of the store".

[0148] Also, the information of the character string "organic coffee shop" is information indicating the content of the item "category of the store", and the information of the character string "local community" is information indicating the content of the item "residence of the target user". The information of the item "residence of the target user" is an example of information indicating the attributes of the target user U.

[0149] Also, the information on a plurality of items included in the new prompt may be information indicating items such as, instead of or in addition to the information indicating the content of the item, for example, the information indicating the item "location of the store", the information of the item "event", the information indicating the item "coupon or discount", the information indicating the item "product price", the item "age group of the target user", and the information indicating the item "interests of the target user".

[0150] Also, the new prompt may be, for example, a prompt input to the generation AI to generate the calibration result as generation information. In this case, the new prompt may be, for example, information such as the character string "You are a professional proofreader. Detect typos, fluctuations, and other misnotations in the input text. The maximum number of detected items is 20. Please check each sentence carefully. The output format is as follows.\n{Error type}\n<Correct>{Original misnotation}\n<Error>{Corrected misnotation}\n\n{Error type} is a character string indicating the error type, {Original misnotation} is a character string indicating the text before the misnotation, and {Corrected misnotation} is a character string indicating the corrected text. ". Such a new prompt is, for example, the prompt first input to the generation AI, and is then input to the generation AI together with the text to be proofread additionally input from the user U. The text to be proofread is also a prompt.

[0151] The information of the character string "You are a professional proofreader." is information indicating the content of the item "Role of AI", the information of the character string "Detect typos, fluctuations, and other misnotations in the input text." is information of the item "Detection target", the information of the character string "The maximum number of detected items is 20." is information indicating the content of the item "Number of detections", and the information of the character string "Please check each sentence carefully." is information indicating the content of the item "Detection procedure".

[0152] Also, the character string "The output format is as follows.\n{Error type}\n<Correct>{Original misnotation}\n<Error>{Corrected misnotation}\n\n{Error type} is a character string indicating the error type, {Original misnotation} is a character string indicating the text before the misnotation, and {Corrected misnotation} is a character string indicating the corrected text." is information indicating the content of the item "Output format".

[0153] Note that the information in the item "Output Format" may further be information on subdivided items. For example, it may be information such as the item "Output Format - Misspelling Type", the item "Output Format - Misspelling Content", the item "Output Format - Correction Content", etc., and may also include information such as the item "Output Format - Detection Reason".

[0154] Also, the new prompt may be a prompt input to the generative AI to generate information indicating a problem or an answer as generation information. In this case, the new prompt is, for example, information in the character string "You are a professional teacher. Under the following constraints, follow the following procedures to provide educational services to the user.\nConstraints\n1. Meet the needs of individual learners: Provide customized guidance tailored to the learner's knowledge level, interests, and learning style.\n2. Abide by educational principles: Based on teaching methods for including educational accuracy and understanding...\nProcedures\n1. Grasp the learning goals of the user: Confirm the topics and goals the user wants to learn.\n2. Create a customized learning plan: Create an individual learning plan based on the user's needs and goals..." Such a new prompt is, for example, a prompt first input to the generative AI to communicate with the user U in a chat format, and is input to the generative AI together with the prompt additionally input from the user U.

[0155] The information in the character string "You are a professional teacher." is information indicating the content of the item "Role of AI", the information in the character string "1. Meet the needs of individual learners:..." is information indicating the content of the item "Constraints - Needs Correspondence", the information in the character string "2. Abide by educational principles:..." is information indicating the content of the item "Constraints - Educational Principles",..., the information in the character string "1. Grasp the learning goals of the user:..." is information indicating the content of the item "Procedures - Goal Grasping",..., and the information in the character string "2. Create a customized learning plan:..." is information indicating the content of the item "Procedures - Learning Plan Creation, etc.".

[0156] Also, the information of the items may be, for example, the information of the item "tone", the information of the item "prompt format", etc. The information indicating the content of the item "tone" is, for example, information such as a direct command like "do it" or an indirect command like "please do it". The item "prompt format" is information such as zero-shot, one-shot, few-shot, etc.

[0157] The selection unit 32 extracts information of a plurality of items from a new prompt, for example, based on rules. For example, the selection unit 32 has dictionary information including a plurality of terms for each item, and can extract information of a plurality of items from the new prompt using such dictionary information.

[0158] Also, the selection unit 32 has an extraction model that extracts information of a plurality of items included in the new prompt received by the reception unit 31, and can also extract information of a plurality of items included in the new prompt using such extraction model.

[0159] The extraction model is a model learned to extract information of a plurality of items from a prompt, and is a model learned to extract information of a plurality of items from a prompt using learning information including, for each prompt, a combination of the prompt and the information of the plurality of items included in the prompt. The extraction model is, for example, an LSTM or Transformer-based model, but may also be the above-described generative AI or other models.

[0160] Also, the selection unit 32 can also extract information of a plurality of items from a new prompt using a language model such as a text generation AI. For example, the selection unit 32 can input information including instruction information indicating an instruction to extract information of a plurality of items from the new prompt and the new prompt into the text generation AI, and cause the text generation AI to extract information of a plurality of items.

[0161] In addition, a plurality of pieces of information are associated with each of a plurality of generative AIs including self-generated AI and other-generated AI. For example, a plurality of pieces of information are associated with each of the plurality of generative AIs in the generative AI-related information storage unit 22.

[0162] Based on the comparison result between the plurality of pieces of information associated with each of the plurality of generative AIs and the plurality of pieces of information included in the new prompt, the selection unit 32 selects, from among the plurality of generative AIs, the generative AI to be used for generating response information.

[0163] For example, the selection unit 32 selects, from among the plurality of generative AIs, the generative AI with the highest degree of match between the plurality of pieces of associated information and the plurality of pieces of information in the new prompt. The degree of match is, for example, the number of matching pieces of information between the plurality of pieces of associated information and the plurality of pieces of information in the new prompt, the ratio of matching, or a value obtained by weighted addition of these. However, when different weightings are performed for each item, it may also be the value obtained by adding the weights of the matching items.

[0164] In addition, the selection unit 32 can also select, from among the plurality of generative AIs, the generative AI whose degree of match between the plurality of pieces of associated information and the plurality of pieces of information in the new prompt is equal to or greater than a threshold value. In this case, when there are a plurality of generative AIs whose degree of match between the plurality of pieces of associated information and the plurality of pieces of information in the new prompt is equal to or greater than the threshold value, the selection unit 32 selects one generative AI from among these generative AIs randomly or according to a predetermined rule.

[0165] In addition, the selection unit 32 can also select, from among the plurality of generative AIs, the generative AI to be used for generating response information based on the context of the user U and the plurality of pieces of information included in the new prompt.

[0166] For example, in addition to information on multiple items, context may be associated with each of the plurality of generative AIs. In this case, the selection unit 32 selects, based on the comparison result between the combination of information on multiple items and context associated with each of the plurality of generative AIs, the combination of information on multiple items included in the new prompt, and the combination of the user's context, the generative AI to be used for generating response information indicating a response to the new prompt.

[0167] For example, the selection unit 32 selects, from among the plurality of generative AIs, the generative AI with the highest degree of match between the combination of information on multiple items and context associated therewith, the combination of information on multiple items included in the new prompt, and the context of the user U.

[0168] When the context includes a plurality of elements, the degree of match of the context is the number of matching elements, the ratio of matching elements, or a value obtained by weighted addition thereof. However, when different weights are assigned to each element, it may be a value obtained by adding the weights of the matching elements. The selection unit 32 selects, from among the plurality of generative AIs, the generative AI with the highest value obtained by weighted addition of the degree of match of multiple items and the degree of match of the context.

[0169] In addition, the selection unit 32 can also select, from among the plurality of generative AIs, the generative AI to be used for generating response information based on the comparison result between the combination of information on multiple items and context associated with each of the plurality of generative AIs, the combination of information on multiple items included in the new prompt, and the context of the user U.

[0170] In this case, the selection unit 32 selects, from among the plurality of generative AIs, the generative AI with the highest degree of match between the combination of information on multiple items and context associated therewith, the combination of information on multiple items included in the new prompt, and the context of the user U. The degree of match is, for example, a value obtained by weighted addition of the degree of match between the information on multiple associated items and the information on multiple items of the new prompt, and the degree of match between the associated context and the context of the user U.

[0171] Further, when there are multiple generative AIs whose degree of match between the combination of information and context of a plurality of associated items and the information of the plurality of items included in the new prompt and the context of the user U is equal to or higher than a threshold value, the selection unit 32 selects one generative AI from among these generative AIs randomly or according to a predetermined rule.

[0172] Further, the selection unit 32 is a model for selecting one generative AI from among the in-house generative AI and the third-party generative AI, and has a selection model obtained by learning, and can also select one generative AI from among the in-house generative AI and the third-party generative AI using such a selection model.

[0173] The selection model is generated by the learning unit 35 using, as learning information, information including, for each prompt, for example, the information of the prompt and information indicating an evaluation by the user U of the provided information generated using the generative AI based on the information including this prompt and provided to the user U. The selection model is, for example, an LSTM or Transformer-based model, but may be the above-described generative AI or other models.

[0174] The selection unit 32 inputs the information of the new prompt into the selection model, and determines whether to use the in-house generative AI or the third-party generative AI depending on whether the value output from such a selection model is a positive value or a negative value.

[0175] Further, the selection unit 32 may have, as a selection model, a model obtained by learning that determines whether a response to the new prompt is possible using the generated information generated using the in-house generative AI. In this case, the selection unit 32 can also determine whether a response to the new prompt is possible using the generated information generated using the in-house generative AI using the selection model.

[0176] The selection model is a model trained using learning information that includes, for each prompt, for example, prompts previously input to the in-house generated AI and evaluation information indicating the evaluation by the user U of the provided information generated using such prompts. The evaluation information is information indicating the evaluation by the user U who provided the generated information generated by the in-house generated AI or information based on such generated information as the provided information, and is, for example, information indicating the evaluation value of a five-level evaluation.

[0177] In this case, when the selection unit 32 inputs the new prompt or each term included in the new prompt to the selection model, if the value output from the selection model is equal to or greater than the threshold value, it is determined that a response to the new prompt can be made using the generated information generated using the in-house generated AI.

[0178] Also, when there are multiple in-house generated AIs, the selection model may be a model that inputs the new prompt or each term included in the new prompt and outputs a value (score) for each in-house generated AI. The selection unit 32 can select the in-house generated AI corresponding to the highest value among the values output from the selection model that are equal to or greater than the threshold value.

[0179] Note that when the utilization request includes information specifying the generation AI, the selection unit 32 can also select the generation AI specified in the utilization request from among the multiple generation AIs.

[0180] Also, the selection unit 32 can select the generation AI to be used for generating the response information from among the multiple generation AIs based on the degree of difference between the information of the multiple associated items and the information of the multiple items of the new prompt. For example, the selection unit 32 selects the generation AI with the lowest degree of difference or the generation AI with a degree of difference equal to or less than the threshold value between the information of the multiple associated items and the information of the multiple items of the new prompt as the generation AI to be used for generating the response information from among the multiple generation AIs.

[0181] Further, the selection unit 32 can also select an AI other than the generative AI. The AI other than the generative AI may be, for example, a natural language processing model that performs analysis, classification, response, etc. of prompts. In this case, the selection unit 32 selects, from among a plurality of AIs, an AI to be used for generating response information indicating a response to the prompt based on information on a plurality of items included in the prompt.

[0182] [3.3.4. Determination unit 33] The determination unit 33 determines whether it is possible to respond to a new prompt using past generated information, which is information generated using the generative AI in the past.

[0183] For example, the determination unit 33 determines whether it is possible to respond to a new prompt using past generated information, which is information generated using an AI generated by another company in the past. For example, when an AI generated by another company is selected from among a plurality of generative AIs by the selection unit 32, the determination unit 33 determines whether it is possible to respond to a new prompt using past generated information, which is information generated using an AI generated by another company in the past.

[0184] For example, the determination unit 33 determines whether it is possible to respond to a new prompt using past generated information based on a prompt comparison result, which is a result of comparison between a past prompt, which is a past prompt used for generating the past generated information, and the new prompt. The past prompt, which is a past prompt used for generating the past generated information, is, for example, a past prompt stored in the past prompt storage unit 21.

[0185] The determination unit 33 calculates a score indicating the comparison result between the past prompt used for generating past generation information and the new prompt, and determines that a response to the new prompt is possible when the calculated score satisfies a predetermined condition (for example, when it is equal to or greater than a threshold value or equal to or less than a threshold value). The score indicating the comparison result between the past prompt and the new prompt is represented by, for example, the similarity or dissimilarity of the vectorized prompts. The vectorization of the prompt is performed, for example, by embedding (Embedding) using a sentence embedding model (for example, a transformer-based model).

[0186] The embedding by the sentence embedding model is, for example, an embedding by text-embedding-ada, BERT, etc. provided by OpenAI, but is not limited to such examples. Note that the vectorization of the content is not limited to the embedding by the sentence embedding model, and for example, the content may be vectorized by Doc2Vec, the average of word embeddings, etc. For word embeddings, for example, Word2Vec, fastText, etc. are used.

[0187] The determination unit 33 vectorizes each past prompt and vectorizes the new prompt. Then, the determination unit 33 determines whether there is a past prompt whose similarity to the new prompt is equal to or greater than the threshold value or whether there is a past prompt whose dissimilarity to the new prompt is equal to or less than the threshold value. When the determination unit 33 determines that there is a past prompt whose similarity to the new prompt is equal to or greater than the threshold value or a past prompt whose dissimilarity to the new prompt is equal to or less than the threshold value, it determines that a response to the new prompt can be made using the past generation information, which is the generation information generated in the past by an AI generated by another company, using such a past prompt.

[0188] The similarity between the new prompt and the past prompt is the cosine similarity, but it may also be the Jaccard similarity or the like, or may be the reciprocal of the Euclidean distance, the reciprocal of the Manhattan distance, or the like. Further, the similarity between the new prompt and the past prompt is, for example, the reciprocal of the similarity between the new prompt and the past prompt.

[0189] Further, the determination unit 33 can also extract information on a plurality of items from the past prompts. The determination unit 33 can extract information on a plurality of items from each past prompt using the dictionary information, extraction model, or generation AI described above.

[0190] The determination unit 33 can also determine whether a response to the new prompt is possible based on the degree of coincidence between the information on the plurality of items included in the past prompt and the information on the plurality of items included in the new prompt. For example, when the determination unit 33 determines that there is a past prompt whose degree of coincidence with the information on the plurality of items of the new prompt is equal to or greater than a threshold value, the determination unit 33 determines that a response to the new prompt can be made using the past generated information generated by a third-party generation AI using such a past prompt.

[0191] The determination unit 33 can also determine whether a response to the new prompt is possible using a determination model that determines whether a response to the new prompt is possible using past generated information and that is obtained by learning.

[0192] The determination model is a model that determines whether a response to the new prompt is possible using past generated information and that is obtained by learning, and is generated by a learning unit 35 described later. The determination model is an example of a model that determines whether a response to the new prompt is possible using generated information generated using a third-party generation AI different from the in-house generation AI.

[0193] The determination unit 33 inputs the information of the new prompt into the determination model, and determines that a response to the new prompt is possible using past generated information when the value output from such a determination model is equal to or greater than a threshold value. The information of the prompt is the prompt or a plurality of terms included in the prompt.

[0194] In addition, the determination unit 33 can also determine whether a response to the new prompt is possible using past generated information, which is information generated using the in-house generated AI in the past. For example, when the in-house generated AI is selected from among a plurality of generated AIs by the selection unit 32, the determination unit 33 determines whether a response to the new prompt is possible using past generated information, which is information generated in the past using the selected in-house generated AI.

[0195] In this case, the determination unit 33 can perform the determination of whether a response to the new prompt is possible using past generated information, which is information generated using the in-house generated AI in the past, in the same manner as the determination method of whether a response to the new prompt is possible using past generated information, which is information generated using the other company's generated AI in the past.

[0196] In addition, the determination unit 33 determines whether a response to the new prompt is possible using past generated information, but can also determine whether a response to the new prompt is possible using existing information other than past generated information. The existing information other than past generated information is, for example, information other than information generated using the generated AI in the past, such as information generated by an AI other than the generated AI or information created by a person.

[0197] For example, the determination unit 33 has dictionary information including information obtained by combining the past prompt and the existing information for each past prompt, and determines whether a response to the new prompt is possible using existing information other than past generated information based on the comparison result between the new prompt and the past prompt.

[0198] Further, the determination unit 33 is a model that determines whether a response to a new prompt is possible using existing information other than past generation information, and can also determine whether a response to a new prompt is possible using a model obtained through learning.

[0199] Further, the determination unit 33 can also calculate the degree of coincidence between the information of a plurality of items of the past prompt and the information of a plurality of items of the new prompt as a score indicating the comparison result between the past prompt and the new prompt. Further, the determination unit 33 can also calculate the degree of difference between the information of a plurality of items of the past prompt and the information of a plurality of items of the new prompt as a score indicating the comparison result between the past prompt and the new prompt. In this case, when the calculated score satisfies a predetermined condition (for example, when the degree of coincidence is equal to or greater than the threshold value or the degree of difference is equal to or less than the threshold value), the determination unit 33 determines that a response to the new prompt is possible.

[0200] Further, the determination unit 33 can also calculate a value obtained by weighted addition of the degree of coincidence and the degree of difference between the information of a plurality of items of the past prompt and the information of a plurality of items of the new prompt as a score indicating the comparison result between the past prompt and the new prompt. In this case, when the calculated score is equal to or less than the threshold value or equal to or greater than the threshold value, the determination unit 33 determines that a response to the new prompt is possible.

[0201] 〔3.3.5. Providing Unit 34〕 The providing unit 34 provides various types of information. For example, the providing unit 34 provides various types of information to the user U by transmitting various types of information to the terminal device 2 via the communication unit 10 and the network N. The providing unit 34 provides, for example, response information generated using the AI selected by the selection unit 32 to the user U.

[0202] For example, the providing unit 34 provides the user U with the response information generated using the generation AI selected by the selection unit 32. For example, when the determination unit 33 determines that a response to a new prompt is possible using past generation information, the providing unit 34 provides the user U with the past generation information or information based on the past generation information as response information indicating a response to the new prompt. Note that the information based on the past generation information may be information generated in the past but newly generated based on the past generation information. The information based on the generation information is information generated by the providing unit 34 based on the generation information.

[0203] For example, when the providing unit 34 determines that a response to a new prompt is possible using past generation information, it determines the past generation information associated with the past prompt having the highest similarity or match degree among the above-described new prompts.

[0204] The past generation information is past generation information that information including the target prompt was input as input information to the generation AI and output from another company's generation AI. The target prompt is, for example, a past prompt having a vector similarity with the new prompt equal to or greater than a threshold value or a past prompt having a match degree of information of a plurality of items equal to or greater than a threshold value. When there are a plurality of past prompts having a similarity or match degree equal to or greater than the threshold value, the target prompt is a past prompt having the highest similarity or match degree or a past prompt randomly or selected according to a predetermined rule among a plurality of past prompts having a similarity or match degree equal to or greater than the threshold value.

[0205] Note that the target prompt may be a past prompt having a score obtained by weighted addition of the similarity and the match degree equal to or greater than a threshold value. In this case, when there is a past prompt having a score equal to or greater than the threshold value, the above-described determination unit 33 can also determine that a response to the new prompt is possible using the past generation information.

[0206] When the determination unit 33 determines that the past generation information associated with the past prompt can be used to respond to a new prompt using the past generation information generated by an AI generated by another company, it is the information generated by the AI generated by another company when the information including the past prompt was input to the AI generated by another company in the past.

[0207] Also, when the determination unit 33 determines that the past generation information associated with the past prompt can be used to respond to a new prompt using the past generation information generated by the in-house generated AI, it is the information generated by the in-house generated AI when the information including the past prompt was input to the in-house generated AI in the past.

[0208] The past generation information is the information output from the generation AI when information including a prompt from the user U is input to the generation AI. However, when the prompt from the user U is a scenario or the like, the past generation information may also be the information output from the generation AI by repeatedly inputting a dialogue history including such a prompt and the history of the information output from the generation AI to the generation AI.

[0209] The scenario includes, for example, part or all of the information of a character string indicating the definition of the task executed by the generation AI, the constraint conditions of the task in the generation AI, the definition of the behavior and tone of the generation AI, the output format of the generation AI, etc., but is not limited to such examples.

[0210] The information based on the past generation information is, for example, when the past generation information is intention information including information indicating the intention type and information indicating the intention content, the information collected by the providing unit 34 or the information processing apparatus 3 based on the information indicating the intention type and the information indicating the intention content.

[0211] The intention type is, for example, the type of intention of additional information, and the information indicating the intention type is, for example, information specifying a function or function corresponding to the intention type. Also, the intention content is the content of the intention of the additional information, and the information indicating the intention content is, for example, information indicating the argument of the function or the parameter of the function corresponding to the intention type.

[0212] When the generation AI receives, as input information, information including intention definition information for extracting an intention type and intention content in addition to a prompt from the user U, it can output intention information. When the generation AI-API is an API provided by OpenAI, the providing unit 34 can cause the generation AI to generate intention information by using the function of function calling.

[0213] Also, the past generation information may be information output from the generation AI by inputting, into the generation AI, information including the collected information collected as described above and instruction information for instructing processing on such collected information.

[0214] The providing unit 34 includes an acquisition processing unit 40 and a providing processing unit 41. When the determination unit 33 determines that it is possible to respond to a new prompt by using the past generation information of another company's generation AI, the acquisition processing unit 40 acquires the past generation information of another company's generation AI from the past prompt storage unit 21 or the like.

[0215] Also, when the determination unit 33 determines that it is not possible to respond to a new prompt by using the past generation information of another company's generation AI, the acquisition processing unit 40 inputs, as input information, information including the new prompt to another company's generation AI via the generation AI-API, and causes another company's generation AI to generate response information indicating a response to the new prompt or information for generating such response information, and acquires the generated information as new generation information via the generation AI-API.

[0216] Also, when the in-house generation AI is selected by the selection unit 32 from among a plurality of generation AIs, the providing unit 34 inputs information including the new prompt to the in-house generation AI selected by the selection unit 32, and causes the in-house generation AI to generate response information indicating a response to the new prompt or information for generating such response information as new generation information, and acquires the generated information.

[0217] Further, when the determination unit 33 determines that it is possible to respond to a new prompt using the past generation information of the company-generated AI, the acquisition processing unit 40 acquires the past generation information of the company-generated AI from the past prompt storage unit 21 or the like.

[0218] Also, when the determination unit 33 determines that it is not possible to respond to a new prompt using the past generation information of the company-generated AI, the acquisition processing unit 40 inputs the information including the new prompt as input information to the company-generated AI, and acquires, as new generation information, response information indicating a response to the new prompt or information for generating the response information from the company-generated AI.

[0219] The providing processing unit 41 provides the user U with the new generation information acquired by the acquisition processing unit 40 or information based on the new generation information as response information. For example, the providing processing unit 41 provides the response information to the user U by transmitting the new generation information acquired by the acquisition processing unit 40 or information based on the new generation information as response information to the terminal device 2.

[0220] For example, when the new generation information is intention information including information indicating an intention type and information indicating an intention content, the providing processing unit 41 collects information corresponding to the new generation information from the information processing device 1 or the information processing device 3 based on the information indicating the intention type and the information indicating the intention content.

[0221] The intention type is, for example, the type of intention of additional information, and the information indicating the intention type is, for example, information specifying a function or function corresponding to the intention type. Also, the intention content is the content of the intention of the additional information, and the information indicating the intention content is, for example, information indicating an argument of a function or a parameter of a function corresponding to the intention type.

[0222] Further, the new generation information or information based on the new generation information may be information output from the generation AI by inputting, to the generation AI, information including the collected information, which is the information collected as described above, and instruction information for instructing processing of such collected information.

[0223] [Learning Department 35] The learning department 35 generates various models through machine learning. For example, the learning department 35 generates a plurality of models including the above-described extraction model, selection model, and determination model, etc.

[0224] For example, the learning department 35 generates an extraction model using learning information that includes, for each prompt, a combination of the prompt and information on a plurality of items included in the prompt. The extraction model is a model learned to extract information on a plurality of items from the information of the prompt. The extraction model is, for example, an LSTM or Transformer-based model, but may also be the above-described generative AI, or other models. For example, the extraction model may be a neural network or a gradient boosting decision tree (GBDT), etc.

[0225] In addition, the learning department 35 uses, as learning information, information that includes, for each prompt, the information of the prompt and information indicating the evaluation by the user U of the provided information generated using generative AI based on the information including this prompt and provided to the user U, to generate a selection model.

[0226] The selection model is a model for determining whether to use the company's own generative AI or a third-party generative AI from the information of the prompt. The selection model is, for example, an LSTM or Transformer-based model, but may also be the above-described generative AI, or other models. For example, the selection model may be a neural network or a gradient boosting decision tree, etc.

[0227] The selection unit 32 inputs the information of the prompt into the selection model, and determines either the company's own generative AI or a third-party generative AI depending on whether the value output from such a selection model is a positive value or a negative value.

[0228] In this case, the information of the prompt is the prompt or a plurality of terms included in the prompt. The information indicating the evaluation is, for example, the evaluation value in a five-level evaluation. The information indicating the evaluation is, for example, used for generating the selection model by multiplying +1 to one of the evaluation values when using the company's own generated AI and multiplying -1 to the other when using the other company's generated AI. The provided information is the generated information generated in the past using the generation AI or the information based on this generated information.

[0229] Further, the selection model may be a model for selecting one generation AI from among three or more generation AIs including one or more other company's generated AIs and two or more company's own generated AIs. In this case, the learning unit 35 uses, as learning information, information including, for each prompt, the information of the prompt and the information indicating the evaluation by the user U for the provided information generated using the generation AI based on the information including this prompt and provided to the user U, to generate the selection model.

[0230] The learning information includes, for example, as the information indicating the evaluation, the information indicating the evaluation for the generated information generated using the other company's generated AI or the information based on the generated information and the information indicating the evaluation for the generated information generated using the company's own generated AI or the information based on the generated information.

[0231] The selection unit 32 inputs the information of the prompt into the selection model, and selects, from among the plurality of generation AIs, the generation AI with the highest score for each generation AI output from such selection model.

[0232] Further, the learning unit 35 uses, as learning information, information including, for each prompt, the information of the prompt and the information indicating the evaluation by the user U for the provided information generated using the other company's generated AI based on the information including this prompt and provided to the user U, to generate the determination model.

[0233] In this case, the information of the prompt is the prompt or a plurality of terms included in the prompt. The information indicating the evaluation is, for example, an evaluation value in a five-level evaluation. The provided information is past generation information generated using an AI generated by another company or information based on this past generation information.

[0234] The determination unit 33 can also determine whether a response to a new prompt is possible using a determination model. The determination model is, for example, an LSTM or Transformer-based model, but may also be the above-described generation AI or other models. For example, the determination model may be a neural network or a gradient boosting decision tree.

[0235] For example, the determination unit 33 inputs the information of the prompt into the determination model, and determines that a response to the prompt is possible using the past generation information when the value output from such a determination model is equal to or greater than a threshold value.

[0236] [4. Processing procedure] Next, the information processing procedure by the processing unit 12 of the information processing apparatus 1 according to the embodiment will be described. FIG. 7 is a flowchart showing an example of information processing by the processing unit 12 of the information processing apparatus 1 according to the embodiment.

[0237] As shown in FIG. 7, the processing unit 12 of the information processing apparatus 1 determines whether a new prompt from the user U has been received (step S10). When the processing unit 12 determines that a new prompt has been received (step S10: Yes), it selects one generation AI from a plurality of generation AIs (step S11).

[0238] Subsequently, the processing unit 12 determines whether a response to the new prompt is possible using the past generation information (step S12). When the processing unit 12 determines that a response to the new prompt is possible using the past generation information (step S12: Yes), it provides the user U with the past generation information or information based on the past generation information as response information indicating the response to the new prompt (step S13).

[0239] Also, when the processing unit 12 determines that a response to a new prompt cannot be made using past generation information (step S12: No), the processing unit 12 acquires response information indicating a response to the new prompt or information for generating the response information from the generation AI as new generation information (step S14). Then, the processing unit 12 provides the user U with the new generation information or information based on the new generation information as response information (step S15).

[0240] When the processing in step S13 is completed, when the processing in step S15 is completed, or when it is determined that no new prompt has been received (step S10: No), the processing unit 12 determines whether the operation end timing has arrived (step S16). The processing unit 12 determines that the operation end timing has arrived, for example, when the power of the information processing apparatus 1 is turned off.

[0241] When the processing unit 12 determines that the operation end timing has not arrived (step S16: No), the process proceeds to step S10. When the processing unit 12 determines that the operation end timing has arrived (step S16: Yes), the process shown in FIG. 7 ends.

[0242] [5. Modification Example] In the example described above, the evaluation by the user U of the provided information is a five - level evaluation, but it may be an evaluation of four levels or less, or an evaluation of six levels or more. Also, the information indicating the evaluation by the user U of the provided information may be, for example, the number of posts or reposts to X (old Twitter), or an evaluation value calculated from the review of the user U.

[0243] In addition to the acquisition unit 30, reception unit 31, selection unit 32, determination unit 33, provision unit 34, and learning unit 35, the processing unit 12 may include a generation unit, an evaluation request unit, and the like. FIG. 8 is a diagram showing another example of the configuration of the information processing apparatus 1 according to the embodiment. The information processing apparatus 1 shown in FIG. 8 includes a generation unit 36, an evaluation request unit 37, etc., in addition to the acquisition unit 30, reception unit 31, selection unit 32, determination unit 33, provision unit 34, and learning unit 35.

[0244] The reception unit 31 receives a reference prompt that is a reference prompt. The generation unit 36 generates a plurality of similar prompts that are a plurality of prompts whose contents are similar to the reference prompt received by the reception unit 31. The acquisition unit 30 acquires the reference prompt received by the reception unit 31 and the plurality of similar prompts generated by the generation unit 36.

[0245] The provision unit 34 provides the user U with reference generation information that is information generated using a generation AI based on information including the reference prompt. Further, the provision unit 34 provides the user U with similar generation information that is information generated for each similar prompt using a generation AI based on information including the corresponding similar prompt among the plurality of similar prompts. The provision of the reference generation information and the similar generation information to the user U is performed by transmitting the reference generation information and the similar generation information to the terminal device 2 of the user U.

[0246] The reception unit 31 receives reference evaluation information indicating an evaluation of the reference generation information that is information generated using a generation AI based on information including the reference prompt, and similar evaluation information indicating an evaluation of the similar generation information that is information generated for each similar prompt using a generation AI based on information including the corresponding similar prompt among the plurality of similar prompts.

[0247] The learning unit 35 causes the determination model to learn the relationship between the content of the change of the similar prompt with respect to the reference prompt and the evaluation of the change based on the difference between the reference prompt and the similar prompt and the difference between the reference evaluation information and the similar evaluation information. The determination model is, for example, an LSTM (Long Short-Term Memory), neural network, or Transformer-based model, but may be the above-described generative AI or other models.

[0248] Also, the generation unit 36 generates a plurality of similar prompts by replacing one or more terms included in the reference prompt with synonyms or equivalent terms. The generation unit 36 has, for example, a thesaurus or synonym dictionary, and uses these dictionaries to generate a plurality of similar prompts.

[0249] Also, the generation unit 36 generates a plurality of similar prompts by replacing the endings of one or more sentences included in the reference prompt with endings of different expressions. The generation unit 36 has, for example, an ending dictionary including terms indicating a plurality of types of endings, and uses such an ending dictionary to generate a plurality of similar prompts.

[0250] Also, the generation unit 36 generates a plurality of similar prompts by adding a specific sentence to the reference prompt. The generation unit 36 uses, for example, a generative AI trained to add a specific sentence to add a specific sentence to the reference prompt to generate a plurality of similar prompts.

[0251] Also, the learning unit 35 generates a determination model so as to output a similar prompt with a high evaluation when information including the reference prompt is input to the generative AI. The determination model is, for example, an LSTM, neural network, or Transformer-based model, but may be the above-described generative AI or other models.

[0252] In addition, the learning unit 35 generates a determination model for each category of reference prompts. The categories are, for example, questions, instructions, opinions (assertions), etc., but are not limited to such examples. For example, they may be classifications such as politics, economy, international, technology, science, sports, entertainment, health, society, etc.

[0253] In addition, the learning unit 35 can also generate a determination model based on the difference between the reference prompt and the similar prompt, the difference between the reference evaluation information and the similar evaluation information, and the difference between the number of tokens in the reference prompt and the number of tokens in the similar prompt.

[0254] In addition, the evaluation request unit 37 transmits an evaluation request including the reference generation information and the similar generation information to the terminal device (for example, terminal device 2) of each evaluator (for example, user U). The reception unit 31 receives the reference evaluation information and the similar evaluation information transmitted from the terminal device 2 of each user U. The evaluator is each user of the crowdsourcing destination or a predetermined evaluator.

[0255] In addition, the learning unit 35 learns the model for each combination of the attribute items of the evaluator. The attributes of the evaluator are, for example, the demographic attributes or psychographic attributes of the evaluator.

[0256] In addition, when the learning unit 35 can suppress a decrease in the output accuracy of the model within a predetermined range with a combination of some of the plurality of attribute items reduced, the learning unit 35 learns the model for each combination of attribute items that does not include some of the attribute items as the combination of the attribute items of the evaluator. The model is, for example, an LSTM, a neural network, or a Transformer-based model, but may be the above-described generative AI or other models.

[0257] Further, the learning unit 35 generates, by learning, a model that detects, for each category of the reference prompt, an item that has been changed in the similar prompt among the information of a plurality of items included in the reference prompt and is estimated to have a higher similar prompt than the reference prompt. The model is, for example, an LSTM, a neural network, or a Transformer-based model, but may be the above-described generative AI or other models.

[0258] 〔6. Hardware Configuration〕 The information processing apparatus 1 according to the above-described embodiment is realized by a computer 80 having a configuration as shown in FIG. 9, for example. FIG. 9 is a hardware configuration diagram showing an example of the computer 80 that realizes the functions of the information processing apparatus 1 according to the embodiment. The computer 80 includes a CPU 81, a RAM 82, a ROM (Read Only Memory) 83, an HDD (Hard Disk Drive) 84, a communication interface (I / F) 85, an input / output interface (I / F) 86, and a media interface (I / F) 87.

[0259] The CPU 81 operates based on a program stored in the ROM 83 or the HDD 84 and controls each part. The ROM 83 stores a boot program executed by the CPU 81 when the computer 80 is started up, a program depending on the hardware of the computer 80, and the like.

[0260] The HDD 84 stores a program executed by the CPU 81 and data used by such a program. The communication interface 85 receives data from other devices via the network N (see FIG. 2) and sends it to the CPU 81, and sends the data generated by the CPU 81 to other devices via the network N.

[0261] The CPU 81 controls output devices such as displays and printers, and input devices such as keyboards or mice via the input / output interface 86. The CPU 81 acquires data from the input device via the input / output interface 86. Also, the CPU 81 outputs the data generated via the input / output interface 86 to the output device.

[0262] The media interface 87 reads a program or data stored in the recording medium 88 and provides it to the CPU 81 via the RAM 82. The CPU 81 loads such a program from the recording medium 88 onto the RAM 82 via the media interface 87 and executes the loaded program. The recording medium 88 is, for example, an optical recording medium such as a DVD (Digital Versatile Disc), PD (Phase change rewritable Disk), a magneto-optical recording medium such as an MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory.

[0263] For example, when the computer 80 functions as the information processing apparatus 1 according to the embodiment, the CPU 81 of the computer 80 realizes the functions of the processing unit 12 by executing the program loaded onto the RAM 82. Also, the data in the storage unit 11 is stored in the HDD 84. The CPU 81 of the computer 80 reads and executes these programs from the recording medium 88, but as another example, these programs may be acquired from other devices via the network N.

[0264] 〔7. Others〕 In addition, among the processes described in the above embodiments, all or part of the processes described as being automatically performed can be manually performed, or all or part of the processes described as being manually performed can be automatically performed by a known method. In addition, the processing procedures, specific names, and information including various data and parameters shown in the above documents and drawings can be arbitrarily changed unless otherwise specified. For example, the various information shown in each figure is not limited to the illustrated information.

[0265] In addition, each component of each illustrated device is a functional concept and does not necessarily have to be physically configured as shown in the figure. That is, the specific form of the distribution and integration of each device is not limited to that shown in the figure, and all or part of it can be functionally or physically distributed and integrated in any unit according to various loads and usage situations.

[0266] For example, the above-described information processing apparatus 1 may be realized by a terminal device and a server computer, may be realized by a plurality of server computers, or may be realized by calling an external platform or the like by an API or network computing depending on the function, and the configuration can be flexibly changed.

[0267] In addition, the above-described embodiments and modifications can be appropriately combined as long as the processing contents do not conflict.

[0268] 〔8. Effects〕 As described above, the information processing apparatus 1 according to the embodiment includes a reception unit 31, a determination unit 33, and a provision unit 34. The reception unit 31 receives a query including a new prompt from the user U or a query for obtaining a new prompt. The determination unit 33 determines whether it is possible to respond to the new prompt using existing information which is existing information. The existing information is, for example, past generated information which is information generated using a generation AI in the past. The provision unit 34 provides the user U with the existing information as response information indicating a response to the new prompt when it is determined by the determination unit 33 that it is possible to respond to the new prompt using the existing information. Thereby, the information processing apparatus 1 can reduce the processing cost when providing response information indicating a response to the prompt.

[0269] Further, the provision unit 34 includes an acquisition processing unit 40 and a provision processing unit 41. When it is determined by the determination unit 33 that it is not possible to respond to the new prompt using the existing information, the acquisition processing unit 40 inputs information including the new prompt to the generation AI as input information, and acquires from the generation AI new generated information which is response information indicating a response to the new prompt or information for generating the response information. The provision processing unit 41 provides the user U with the new generated information acquired by the acquisition processing unit 40 or information based on the new generated information as response information. Thereby, the information processing apparatus 1 can appropriately provide the user U with response information indicating a response to the prompt.

[0270] Further, the determination unit 33 determines whether it is possible to respond to the new prompt based on the result of comparison between the past prompt used for generating the past generated information and the new prompt. Thereby, the information processing apparatus 1 can appropriately provide the user U with response information indicating a response to the prompt.

[0271] Further, when the score indicating the comparison result between the past prompt used for generating the past generation information and the new prompt satisfies a predetermined condition (for example, when it is equal to or greater than a threshold value or equal to or less than a threshold value), the determination unit 33 determines that a response to the new prompt is possible. Thereby, the information processing apparatus 1 can appropriately provide the user U with response information indicating a response to the prompt.

[0272] Further, the determination unit 33 determines whether a response to the prompt is possible based on the degree of match or difference between the information of a plurality of items included in the past prompt used for generating the past generation information and the information of a plurality of items included in the new prompt. Thereby, the information processing apparatus 1 can appropriately provide the user U with response information indicating a response to the prompt.

[0273] Further, the determination unit 33 determines whether a response to the new prompt is possible using a model that determines whether a response to the new prompt is possible using existing information and that is obtained by learning. Thereby, the information processing apparatus 1 can appropriately provide the user U with response information indicating a response to the prompt.

[0274] Further, the information processing apparatus 1 includes a selection unit 32 that selects one generation AI from among the generation AI and another generation AI different from the generation AI. The generation AI is an example of an AI generated by another company, and the other generation AI is an example of an in-house generated AI. The providing unit 34 includes an acquisition processing unit 40 and a providing processing unit 41. When another generation AI is selected by the selection unit 32, the acquisition processing unit 40 inputs the new prompt as input information to the other generation AI and acquires, as new generation information, response information indicating a response to the new prompt or information for generating the response information generated by the other generation AI. The providing processing unit 41 provides the user U with the new generation information acquired by the acquisition processing unit 40 or information based on the new generation information as response information. Thereby, the information processing apparatus 1 can appropriately provide the user U with response information indicating a response to the prompt.

[0275] In addition, the information processing apparatus 1 includes a learning unit 35 that generates a selection model based on information indicating an evaluation of generated information generated using a generation AI (for example, an AI generated by another company) or information based on the generated information and information indicating an evaluation of generated information generated using another generation AI (for example, an in-house generated AI) or information based on the generated information. As a result, the information processing apparatus 1 can appropriately provide the user U with response information indicating a response to the prompt.

[0276] As described above, the embodiments of the present application have been described in detail with reference to the drawings. However, this is merely an example, and the present invention can be implemented in other forms with various modifications and improvements based on the knowledge of those skilled in the art, including the aspects described in the column of the disclosure of the invention.

[0277] In addition, the "section (section, module, unit)" described above can be read as "means" or "circuit". For example, the acquisition unit can be read as an acquisition means or an acquisition circuit.

Explanation of Reference Numerals

[0278] 1, 3 Information processing apparatus 2 Terminal device 10 Communication unit 11 Storage unit 12 Processing unit 20 User information storage unit 21 Past prompt storage unit 22 Generation AI related information storage unit 30 Acquisition unit 31 Reception unit 32 Selection unit 33 Judgment unit 34 Provision unit 35 Learning unit 36 Generation unit 37 Evaluation request unit 40 Acquisition processing unit 41 Provision processing unit 100 Information processing system N Network

Claims

1. A receiving unit that receives a query including a new prompt from a user or a query for obtaining the new prompt, A determination unit that determines whether it is possible to respond to the new prompt using existing information that is existing information, A providing unit that, when the determination unit determines that it is possible to respond to the new prompt using the existing information, provides the user with the existing information as response information indicating a response to the new prompt. An information processing apparatus characterized by the above.

2. The providing unit, When the determination unit determines that it is not possible to respond to the new prompt using the existing information, inputs information including the new prompt into a generation AI as input information, and obtains, as new generation information, response information indicating a response to the new prompt or information for generating the response information from the generation AI, A providing processing unit that provides the user with the new generation information obtained by the obtaining processing unit or information based on the new generation information as the response information. The information processing apparatus according to claim 1, characterized by the above.

3. The existing information is, Past generation information that is information generated using the generation AI in the past The information processing apparatus according to claim 2, characterized by the above.

4. The determination unit, Determines whether it is possible to respond to the new prompt based on the result of comparison between the past prompt used for generating the past generation information and the new prompt The information processing apparatus according to claim 3, characterized by the above.

5. The determination unit, Determines that it is possible to respond to the new prompt when the score indicating the comparison result between the past prompt used for generating the past generation information and the new prompt is greater than or equal to a threshold value or less than or equal to the threshold value The information processing apparatus according to claim 4, characterized by the above.

6. The determination unit, Determines whether it is possible to respond to the prompt based on the degree of coincidence or difference between the information of a plurality of items included in the past prompt used for generating the past generation information and the information of a plurality of items included in the new prompt The information processing apparatus according to claim 3, characterized by the above.

7. The determination unit, A model that determines whether a response to the new prompt can be made using the existing information, and determines whether a response to the new prompt can be made using a model obtained by learning. The information processing apparatus according to claim 1 or 2, characterized in that.

8. A selection unit that selects one of the generation AIs from among the generation AI and another generation AI different from the generation AI using a selection model, The providing unit, When the other generation AI is selected by the selection unit, the new prompt is input as input information to the other generation AI, and response information indicating a response to the new prompt or information for generating the response information is generated as new generation information by the other generation AI and acquired by an acquisition processing unit; A providing processing unit that provides the new generation information acquired by the acquisition processing unit or information based on the new generation information to the user as the response information. The information processing apparatus according to claim 2, characterized in that.

9. A learning unit that generates the selection model based on information indicating an evaluation of generated information generated using the generation AI or information based on the generated information and information indicating an evaluation of generated information generated using the other generation AI or information based on the generated information. The information processing apparatus according to claim 8, characterized in that.

10. An information processing method executed by a computer, A reception step of receiving a query including a new prompt from a user or a query for obtaining the new prompt, A determination step of determining whether a response to the new prompt can be made using existing information that is existing information, When it is determined by the determination step that a response to the new prompt can be made using the existing information, a providing step of providing the existing information to the user as response information indicating a response to the new prompt. An information processing method characterized by that.

11. A reception procedure for receiving a query including a new prompt from a user or a query for obtaining the new prompt, A determination procedure for determining whether a response to the new prompt can be made using existing information that is existing information, When it is determined by the determination procedure that a response to the new prompt can be made using the existing information, the computer is caused to execute a providing procedure of providing the existing information to the user as response information indicating a response to the new prompt. An information processing program characterized by the above.

Images