Interactive capability improvement support device, interactive capability improvement support control method, and interactive capability improvement support control program

Abstract

To support improvement in interactive capability of a user while assuming a variety of interaction scenes.SOLUTION: An interactive capability improvement support device 30 comprises: a reception part 31 which receives information for selecting a scene 312 in which participants including a user and one or more machine learning models 311 interact with each other and the machine learning model 311 included in the participants; a structuring part 32 which structures an environment 321 in which the participants interact with each other in the selected scene 312; an acquisition part 33 which acquires interaction details 331 between the user and machine learning model 311 in the environment 321; and an evaluation part 34 which evaluates the interactive capability of the user from the acquired interaction details 331 based upon an evaluation standard 341 for evaluating interactive capability according to the interaction details 331.SELECTED DRAWING: Figure 8

Description

Technical Field

[0001] The present invention relates to a dialogue ability improvement support device, a dialogue ability improvement support control method, and a dialogue ability improvement support control program.

Background Art

[0002] Patent Document 1 discloses a system that promotes the health, learning, and purchasing of users through a dialogue that makes the user feel as if they are contacting a single person who cares about them as if they were themselves. This system analyzes the mental state of the user from the content of the user's speech on SNS (Social Networking Service) and gives advice to the user through SNS based on the analysis result.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] The technology of Patent Document 1 is not sufficient to solve the improvement of the user's dialogue ability assuming various dialogue scenarios.

[0005] The main object of the present invention is to support the improvement of the user's dialogue ability assuming various dialogue scenarios.

Means for Solving the Problems

[0006] An interactive ability improvement support device according to one aspect of the present invention includes: a reception unit that receives information for selecting a scene in which participants including a user and one or more machine learning models interact with each other, and the machine learning models included in the participants; a construction unit that constructs an environment in which the participants interact with each other in the selected scene; an acquisition unit that acquires the dialogue content between the user and the machine learning model in the environment; and an evaluation unit that evaluates the dialogue ability of the user from the acquired dialogue content based on an evaluation criterion for evaluating the dialogue ability according to the dialogue content.

[0007] In another aspect for achieving the above object, an interactive ability improvement support control method according to one aspect of the present invention is a method in which an information processing device receives information for selecting a scene in which participants including a user and one or more machine learning models interact with each other, and the machine learning models included in the participants, constructs an environment in which the participants interact with each other in the selected scene, acquires the dialogue content between the user and the machine learning model in the environment, and evaluates the dialogue ability of the user from the acquired dialogue content based on an evaluation criterion for evaluating the dialogue ability according to the dialogue content.

[0008] Furthermore, in a further aspect for achieving the above object, an interactive ability improvement support control program according to one aspect of the present invention causes a computer to execute a reception process for receiving information for selecting a scene in which participants including a user and one or more machine learning models interact with each other, and the machine learning models included in the participants, a construction process for constructing an environment in which the participants interact with each other in the selected scene, an acquisition process for acquiring the dialogue content between the user and the machine learning model in the environment, and an evaluation process for evaluating the dialogue ability of the user from the acquired dialogue content based on an evaluation criterion for evaluating the dialogue ability according to the dialogue content.

[0009] Furthermore, the present invention can also be realized by a computer-readable non-volatile recording medium storing such an interactive ability improvement support control program (computer program).

Advantages of the Invention

[0010] The present invention can assist in improving the user's dialogue ability assuming various dialogue scenarios.

Brief Description of the Drawings

[0011]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6A

Figure 6B

Figure 7

Figure 8

Figure 9A

Figure 9B

Figure 10

Figure 11A

Figure 11B

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

DETAILED DESCRIPTION OF THE INVENTION

[0012] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

[0013] <The First Embodiment> FIG. 1 is a block diagram showing the configuration of an interactive ability improvement support device 10 according to the present disclosure. The interactive ability improvement support device 10 provides an environment for performing conversations (communications) in various scenarios (scenes) where participants including a user and one or more machine learning models (artificial intelligence) are assumed, and supports the user in improving their interactive ability through such conversations. The machine learning model is, for example, a generative AI having a function of having a natural conversation with the user by a large language model. Note that there may be a plurality of users included in the participants.

[0014] (Definition of Machine Learning Model) Here, an overview of the machine learning model and the generation of the machine learning model will be described.

[0015] A machine learning model is a model that generates an answer corresponding to a request. As an example, when a query generated based on a request from a user or the like is input, the machine learning model outputs an answer corresponding to the query.

[0016] As an example, a machine learning model is composed of a language model. The language model may be, for example, what is called an LLM (Large Language Models), but is not limited thereto.

[0017] A language model is a machine learning model (also called a generative model) that inputs a language and outputs a language. A language model is one that has learned the relationship between words in a sentence, and is a model that generates a related string related to the target string from the target string. By using a language model that has learned sentences and texts in various contexts, it is possible to generate a related string with appropriate content related to the target string.

[0018] For example, the case of using a language model in question and answer will be described. The language model receives, as the target string, an input of a question such as "What kind of country is Japan?". The language model generates a string such as "Japan is an island country in the Northern Hemisphere." as an answer to the question.

[0019] The learning method of the language model is not particularly limited. As an example, it may be one that is learned to output at least one sentence including the input string. For a specific example, the language model is GPT (Generative Pretrained-Transformer) that outputs a sentence including the input string by predicting a highly probable string following the input string.

[0020] In addition to this, for example, T5 (Text-to-Text Transfer Transformer), BERT (Bidirectional Encoder Representations from Transformers), RoBERTa (Robustly optimized BERT approach), ELECTRA (Efficiently Learning an Encoder that Classifies Token Replacements Accurately), etc. are also language models.

[0021] Alternatively, the language model may output a natural language corresponding to the string input in an artificial language. Also, the content generated by the language model is not limited to a string. The language model may generate, for example, image data, video data, audio data, or other data formats corresponding to the input string.

[0022] The dialogue ability improvement support device 10 is communicably connected to the terminal device 20. The terminal device 20 is an information processing device such as a personal computer, a smartphone, or a tablet terminal. The terminal device 20 inputs the information input via the input operation by the user to the dialogue ability improvement support device 10 and displays the information output from the dialogue ability improvement support device 10 on the provided display screen 21.

[0023] The dialogue ability improvement support device 10 is an information processing device such as a server, and includes a reception unit 11, a construction unit 12, an acquisition unit 13, an evaluation unit 14, a presentation unit 15, a scene generation unit 16, a machine learning model generation unit 17, a storage unit 18, and a recommendation unit 19. The reception unit 11, the construction unit 12, the acquisition unit 13, the evaluation unit 14, the presentation unit 15, the scene generation unit 16, the machine learning model generation unit 17, and the recommendation unit 19 are examples of reception means, construction means, acquisition means, evaluation means, presentation means, scene generation means, machine learning model generation means, and recommendation means, respectively.

[0024] The storage unit 18 is a storage device such as a RAM (Random Access Memory) 903 or a hard disk 904, which will be described later with reference to FIG. 16. The storage unit 18 stores scene management information 181, machine learning model management information 182, user management information 183, dialogue content 184, evaluation criteria 185, evaluation results 186, scene generation criteria 187, and recommendation criteria 188. Details of the above-mentioned information stored in the storage unit 18 will be described later.

[0025] The reception unit 11 receives, from the terminal device 20, information input by the user for selecting a scene in which participants including the user and one or more machine learning models have a dialogue and the machine learning models included in the participants.

[0026] For example, when the user selects a scene for dialogue, the reception unit 11 displays the scene management information 181 stored in the storage unit 18 on the display screen 21 of the terminal device 20 as an option. Note that the scene management information 181 is, for example, given in advance by an administrator or the like of the dialogue ability improvement support device 10.

[0027] FIG. 2 is a diagram illustrating the content of the scene management information 181 in the dialogue ability improvement support device 10 according to the present disclosure. The scene management information 181 illustrated in FIG. 2 represents, for each individual dialogue scene, the scene name, the description of the scene, and the purpose of learning the dialogue. Note that the scene management information 181 illustrated in FIG. 2 is an example, and the scene management information 181 may include content different from that illustrated in FIG. 2.

[0028] The scene management information 181 illustrated in FIG. 2 includes, as dialogue scenes, communication with friends, communication at work, communication at school, communication in public places, communication within the family, communication in emergencies, and the like. According to the scene management information 181, for example, communication with friends aims to simulate a conversation with friends in a relaxed environment and acquire the dialogue ability (dialogue skills) for building and maintaining friendships. According to the scene management information 181, for example, communication at work aims to simulate a conversation with other employees in a workplace meeting and acquire the dialogue ability for sharing opinions with other employees and smoothly advancing work. The user selects a scene in which they want to improve their dialogue ability from the scenes represented by the scene management information 181 displayed on the display screen 21.

[0029] The reception unit 11 displays the machine learning model management information 182 on the display screen 21 of the terminal device 20, for example, as an option when the user selects a machine learning model included in the dialogue participants. Note that the machine learning model management information 182 is, for example, provided in advance by an administrator or the like of the dialogue ability improvement support device 10.

[0030] FIG. 3 is a diagram illustrating the content of the machine learning model management information 182 in the dialogue ability improvement support device 10 according to the present disclosure. The machine learning model management information 182 illustrated in FIG. 3 represents the participation scene of the dialogue, its characteristics, and the learning data used when generating the machine learning model for each individual machine learning model with different characteristics. Note that the machine learning model management information 182 illustrated in FIG. 3 is an example, and the machine learning model management information 182 may include content different from that illustrated in FIG. 3.

[0031] The machine learning model management information 182 illustrated in FIG. 3 includes, as machine learning models, ordinary working people, engineers, entrepreneurs, teachers, student A, student B, and the like. According to the machine learning model management information 182, for example, an ordinary working person is a machine learning model generated by having general common sense as a working person and learning educational materials related to business communication and the like. According to the machine learning model management information 182, for example, an engineer is a machine learning model generated by having technical expertise as a characteristic and learning technical documents, product manuals, technical blogs, professional magazines, and the like. The user selects a machine learning model that they want to participate in the conversation in order to improve their dialogue ability from among the scenes represented by the scene management information 181 displayed on the display screen 21.

[0032] Here, each machine learning model (language model) is generated based on learning data. For example, when generating a machine learning model that reflects the thinking of experts, various learning data (data resources) such as public statements by experts in the media, related industry articles, and reporters are utilized and the machine learning model is trained (learned).

[0033] Alternatively, a machine learning model with specific characteristics may be generated by utilizing an existing machine learning model (language model). For example, transfer learning (fine-tuning) may be performed to train the weights of a pre-generated and learned model with new learning data. Specifically, by using an existing language model and performing additional training with unique learning data (dataset), the machine learning model may be given characteristics. By preparing the statements of teachers and students respectively as learning data and performing additional learning of these learning data on the basic machine learning model, the machine learning model may be characterized (personalized).

[0034] A "feedback" method may be used to generate a machine learning model. For example, an evaluator (such as the generator of the machine learning model) determines the appropriateness of the feedback for the output from the machine learning model. The judgment result is used as learning data and retraining is executed. As a result, the performance (output accuracy) of the machine learning model is improved. For example, when generating a machine learning model (teacher model) that reflects the teacher's thinking, the evaluator (the generator of the machine learning model) may provide feedback so that the output result of the teacher model "comes closer to a specific teacher".

[0035] In addition, regarding the characterization (personalization) of the machine learning model, in addition to generating the machine learning model using learning data, the machine learning model may be characterized by so-called "prompt engineering". That is, by devising questions and instructions (prompts) input to the machine learning model, the output of the machine learning model may be guided (regulated) in a specified manner. For example, when it is desired to obtain an answer like that of a teacher (when an existing machine learning model is to be used as a teacher model), a stereotyped sentence such as "Please think like a teacher" may be input to the above-mentioned machine learning model. Alternatively, related statements of the teacher, etc. may be obtained from a database, and the statements obtained from the database may be input to the machine learning model together with the above-mentioned stereotyped sentence.

[0036] The construction unit 12 shown in FIG. 1 constructs an environment in which participants including the machine learning model selected by the user as described above interact in the conversation scenario selected by the user as described above. Specifically, the construction unit 12 constructs the environment as a program and data processed by the dialogue ability improvement support device 10, and constructs the function of a user interface for the user and the machine learning model to have a dialogue. For example, the construction unit 12 constructs a function of displaying the utterance by the machine learning model included in the participant on the display screen 21, and inputting the user's utterance content represented as the character information input via the input operation by the user to the terminal device 20 or the result of voice recognition of the voice spoken by the user to the machine learning model included in the participant.

[0037] The construction unit 12 constructs, for example, an environment in which participants converse about topics included in a scenario according to a scenario given in advance by an administrator or the like of the dialogue ability improvement support device 10 for each dialogue scenario. The scenario includes, for example, the content of the speech when starting a dialogue by a certain machine learning model included in the participants, and the roles of the individual machine learning models included in the participants. The construction unit 12 may store information representing the constructed environment in the storage unit 18.

[0038] The acquisition unit 13 acquires the dialogue content 184 between the user and the machine learning model in the environment constructed by the construction unit 12 as described above.

[0039] FIG. 4 is a diagram showing a first example of the dialogue content 184 in the dialogue ability improvement support device 10 according to the present disclosure. FIG. 4 represents an example of a mode in which the dialogue content 184 is displayed on the display screen 21. In the example shown in FIG. 4, as the dialogue scenario, among the scenarios shown by the scenario management information 181 illustrated in FIG. 2, communication at school is selected by the user, and as the machine learning models participating in the dialogue, among the machine learning model management information 182 illustrated in FIG. 3, the teacher and student A are selected. Also, the participation of the teacher and student A in the dialogue may be specified in advance in a scenario related to the communication scenario at school.

[0040] According to the dialogue content 184 illustrated in FIG. 4, first, the dialogue is started by the teacher's statement "Before starting a new chapter today, let's review the previous content." Next, in response to the teacher's statement described above, student A makes a statement "Teacher, the previous content was about the method of calculating the area of a triangle, right?" Then, in response to the statement by student A, the user makes a statement "Umm, I think it was about calculating the area of a quadrilateral." Then, in response to the statement by the user, the teacher makes a statement "The correct answer is, as student A said, the method of calculating the area of a triangle. User, please check your notes again."

[0041] The evaluation unit 14 shown in FIG. 1 evaluates the user's dialogue ability from the dialogue content 184 acquired by the acquisition unit 13 based on an evaluation criterion 185 for evaluating the user's dialogue ability according to the dialogue content 184. The evaluation unit 14 analyzes the meaning of the user's utterance using, for example, an existing semantic analysis technique. In this case, the evaluation criterion 185 includes a criterion for evaluating how appropriate the user's utterance is for the scenario of the dialogue situation. The evaluation unit 14 analyzes the user's emotion indicated by the user's utterance using, for example, an existing sentiment analysis technique. In this case, the evaluation criterion 185 includes a criterion for evaluating whether the user's emotion is positive / affirmative or negative / negative. Note that the evaluation criterion 185 is, for example, given in advance by the administrator of the dialogue ability improvement support device 10 or the like.

[0042] FIG. 5 is a diagram illustrating an evaluation result 186 of the user's dialogue ability by the evaluation unit 14 for the dialogue content 184 illustrated in FIG. 4 in the dialogue ability improvement support device 10 according to the present disclosure. FIG. 5 shows an example of a mode in which the evaluation result 186 is displayed on the display screen 21.

[0043] According to the evaluation result 186 illustrated in FIG. 4, regarding the speed of the user's reaction, the evaluation unit 14 evaluates that "You reacted quickly." This evaluation is due to the fact that the time interval between the statement by student A and the statement by the user is shorter than the threshold value indicated by the evaluation criterion 185. Also, according to the evaluation result 186 illustrated in FIG. 4, regarding the accuracy of the content of the user's statement, the evaluation unit 14 evaluates and makes a suggestion that "Your answer was incorrect. It is recommended to review the method of calculating the area of a triangle." This evaluation and suggestion are based on the evaluation criterion 185 that indicates pointing out and correcting the mistake when there is a mistake in the statement by the user. And the evaluation unit 14 evaluates and makes a suggestion regarding the method of communication that "Taking time to think using the word 'um' is a good approach. However, if you are not confident in your answer, you can also consider answering in the form of a question. Example: 'Was it not the calculation of the area of a quadrilateral?'". Furthermore, the evaluation unit 14 evaluates and makes a suggestion as the next proposal that "It is important to have an attitude of checking your answer by yourself after making a mistake. Next time, let's try to admit the mistake and make an effort to check your answer by yourself." These evaluations and suggestions by the evaluation unit 14 are based on the evaluation criterion 185 that associates the content of the statement by the user with the content of the evaluation and suggestion for the user who made the statement.

[0044] FIGS. 6A and 6B are diagrams showing a second example of the dialogue content 184 in the dialogue ability improvement support device 10 according to the present disclosure. FIGS. 6A and 6B represent an example of the manner in which the dialogue content 184 is displayed on the display screen 21, similar to FIG. 4. In the example shown in FIGS. 6A and 6B, among the dialogue scenes indicated by the scene management information 181 as the dialogue scene, communication in a business negotiation with another company (not shown in FIG. 2) is selected by the user, and as the machine learning models participating in the dialogue, the salespersons of Company A, Company B, and Company C (not shown in FIG. 3) are selected.

[0045] According to the conversation content 184 illustrated in FIG. 6A, in response to the user's statement in the negotiation, "Please tell me the offering price of your company's module product that meets our company's requirements." the salespersons of Company A, Company B, and Company C respectively make statements regarding the offering price of their own products and, if necessary, the superiority of their own products over those of other companies.

[0046] In the examples shown in FIGS. 6A and 6B, the display screen 21 of the terminal device 20 is assumed to have a touch panel function. And when an input operation is performed by a user who drags and moves a display area representing a statement by a certain machine learning model to the bottom of the screen on the display screen 21, the construction unit 12 constructs a dialogue environment in which the machine learning model makes a statement in response to a statement by another machine learning model.

[0047] In the examples shown in FIGS. 6A and 6B, a user who is interested in Company A, after receiving the statements by the salespersons of Company B and Company C, drags and moves the display area representing the statement of the salesperson of Company A to the bottom of the screen as illustrated in FIG. 6A to confirm how the salesperson of Company A appeals to the superiority of its own products. As a result, as illustrated in FIG. 6B, the salesperson of Company A makes a statement appealing to the superiority of its own products over those of other companies, saying, "It is 1 million yen per lot, but if you purchase two lots, there will be a 20% discount and it will be the same price as Company B, and it is not inferior to Company C in terms of performance."

[0048] By constructing a dialogue environment such as that illustrated in FIGS. 6A and 6B by the construction unit 12, the user can practice dialogue to improve the dialogue ability in negotiation so that the user can determine the purchase destination that can be purchased under the optimal conditions in the negotiation with multiple companies for determining the purchase destination of a certain product.

[0049] Here, in the construction unit 12, a method for the machine learning model to generate a dialogue will be described with reference to FIGS. 7 to 12. FIGS. 7 to 12 are diagrams for explaining a method for the machine learning model to generate a dialogue, taking the dialogue scene in the childcare support service as an example in the dialogue ability improvement support device 10. The construction unit 12 includes a dialogue generation unit 120 (not shown in FIG. 1).

[0050] The dialogue generation unit 120 includes basic functions, intervention functions, flip functions, and the like. The dialogue generation unit 120 realizes these functions using a plurality of machine learning models selected by the user.

[0051] (Basic function) First, the basic function will be described.

[0052] The basic function is a function of sequentially presenting answers from a plurality of machine learning models to a user's question (answer request). In the basic function, the second and subsequent machine learning models answer while referring to the answers from each of the machine learning models that have answered so far.

[0053] The request (user's question) and answer (response of the machine learning model) in the basic function are described as the first request and the first answer. For one first request, a first answer is obtained from each of a plurality of machine learning models. The plurality of first answers can be different from each other.

[0054] Note that the first answer (answer of the first machine learning model) to the first request is described as the first request [1]. Similarly, the second answer to the first request is described as the first request [2], and the third answer is described as the first answer [3]. The first answer from the k-th machine learning model is denoted as the first answer [k] (k is a positive integer, the same hereinafter).

[0055] Furthermore, the plurality of first answers are collectively described as the first answer set.

[0056] The dialogue generation unit 120 acquires a first request from the user. The dialogue generation unit 120 inputs the acquired first request into one machine learning model among a plurality of machine learning models selected by the user.

[0057] Specifically, the dialogue generation unit 120 selects one machine learning model from among the plurality of machine learning models selected by the user. Further, the dialogue generation unit 120 generates a query based on the first request and the scene selected by the user so that a response corresponding to the scene selected by the user can be obtained. The dialogue generation unit 120 inputs the generated query into the one selected machine learning model.

[0058] For example, when the user selects three models, namely, the supporter A model, the supporter B model, and the supporter C model illustrated in FIG. 7, the dialogue generation unit 120 selects the supporter A model from among the three machine learning models. The dialogue generation unit 120 inputs a query into the selected teacher model.

[0059] When the dialogue generation unit 120 acquires output data (first response [1]) from the machine learning model, the dialogue generation unit 120 selects one machine learning model from among the plurality of machine learning models selected by the user into which the query has not been input. For example, in the above example, the supporter B model is selected from among the supporter B model and the supporter C model.

[0060] The dialogue generation unit 120 generates a query for inputting into the machine learning model selected second based on the first request, the scene selected by the user, and the output data (first response) of the machine learning model for which the response has been obtained. In the above example, the dialogue generation unit 120 generates a query (query for the supporter B model) for inputting into the supporter B model according to the first request, the scene (scene), and the first response [1] of the supporter A model.

[0061] The dialogue generation unit 120 inputs the generated query into the machine learning model selected second to obtain output data (first response [2]). In the above example, the dialogue generation unit 120 obtains the first response [2] from the supporter B model.

[0062] When the dialogue generation unit 120 obtains the first response [2] from the second machine learning model, the dialogue generation unit 120 selects the third machine learning model from which to obtain a response. In the above example, the supporter C model is selected.

[0063] When the dialogue generation unit 120 selects the third machine learning model, the dialogue generation unit 120, in the same manner as in the case of the second machine learning model, generates a query for the third machine learning model based on the first request, the scenario, and the output data (the first response [1], the first response [2]) of the machine learning models for which responses have already been obtained. The dialogue generation unit 120 uses the generated query to obtain the first response [3] from the third machine learning model. In the above example, the dialogue generation unit 120 obtains the first response [3] from the supporter C model.

[0064] The inputs and outputs related to the above three machine learning models are summarized as shown in FIG. 7. As shown in FIG. 7, the dialogue generation unit 120 adds the user's question (response request) and the responses (output data) of the machine learning models for which responses have already been obtained in the scenario, and inputs them to the subsequent machine learning model, thereby obtaining the response of the machine learning model that refers to the responses of the previous machine learning models.

[0065] When the dialogue generation unit 120 obtains the first responses [1] to [3] from each of the machine learning models selected by the user for the user's first request, the dialogue generation unit 120 displays the first responses (the first response set) of each of the machine learning models on the display screen 21 of the terminal device 20 (see FIG. 8).

[0066] Note that FIG. 8 shows an example of the first request and the first responses when the user selects "communication of child-rearing support service" as the dialogue scenario as described above and selects three models, namely, the supporter A model, the supporter B model, and the supporter C model, as the machine learning models.

[0067] Also, in the display screen 21 illustrated in FIG. 8, the person (icon) on the left side indicates the user (questioner). Also, the person (icon) on the right side indicates the machine learning model (answerer). Further, the dialogue generation unit 120 makes it possible for the user to easily identify the answerer by changing the color or the like of the icon corresponding to each machine learning model. For example, in FIG. 8, the icons on the right side correspond to the supporter A model, the supporter B model, and the supporter C model in order from the top.

[0068] A user who considers that the troubles or the like have been resolved by the first answer (first answer set) from each machine learning model ends the childcare support service.

[0069] Here, the functions that respond to the user's desire for more answers and answers with deeper content are the "intervention function" and the "flip function".

[0070] (Intervention function) The intervention function is a function of sequentially obtaining new answers from a plurality of machine learning models by the user adding (intervening) a question to the first answer (first answer set) obtained by the basic function. The intervention function is a function of answering additional questions by the user.

[0071] In the intervention function, from the second and subsequent machine learning models, new answers that refer to the new answers of each machine learning model obtained so far can be obtained.

[0072] The requests and new answers added in the intervention function are described as second requests and second answers. For one second request, second answers are obtained from each of a plurality of machine learning models. The plurality of second answers are collectively described as a second answer set.

[0073] Note that the intervention function may be further executed for the second answer and subsequent answers. The new requests and new answers in the intervention function performed for the nth answer (n is a positive integer) are described as the (n + 1)th request and the (n + 1)th answer, and the plurality of (n + 1)th answers are collectively described as the (n + 1)th answer set.

[0074] For example, in the state shown in FIG. 8 (the state in which the first answer set is displayed), the user inputs an additional question (second request) to the dialogue generation unit 120 (see FIG. 9A).

[0075] When acquiring the additional question (second request) from the user, the dialogue generation unit 120 generates a query from the second request of the user, the scene selected by the user, the first request, and / or the first answer set. The dialogue generation unit 120 inputs the generated query into one of the plurality of machine learning models selected by the user.

[0076] For example, the dialogue generation unit 120 inputs the generated query into the machine learning model first selected in the basic function (in the above example, the supporter A model). The dialogue generation unit 120 obtains a second answer [1] from the machine learning model.

[0077] Similar to the basic function, the dialogue generation unit 120 generates a query for input into the second selected machine learning model. The dialogue generation unit 120 generates a query based on the second request, the scene, the first request, and / or the first answer set, and the second answer [1] of the machine learning model first selected. The dialogue generation unit 120 inputs the generated query into the second machine learning model (in the above example, the supporter B model).

[0078] The dialogue generation unit 120 performs the same processing for the third machine learning model (supporter C model).

[0079] When the dialogue generation unit 120 obtains a second answer (second answer set) from each machine learning model selected by the user for the second request of the user, the dialogue generation unit 120 displays the second answer on the display screen 21 of the terminal device 20 (see FIG. 9B).

[0080] The inputs and outputs related to the three machine learning models in the intervention function are summarized as shown in FIG. 10. Note that in FIG. 10, the first answer set obtained by the basic function is used for generating each query.

[0081] In this way, the user can repeatedly ask questions to the dialogue generation unit 120 (the machine learning model selected by the user) using the intervention function.

[0082] (Flip function) The flip function is a function that allows the user to change (flip) the order in which multiple machine learning models are used for the answer obtained by the basic function.

[0083] In the flip function, for at least one machine learning model, a new first answer is obtained by referring to the already obtained first answer from each machine learning model whose order after the change is earlier than that of the machine learning model.

[0084] Here, the order of the machine learning models after being changed by the flip function is denoted as k(new). For example, the dialogue generation unit 120 causes one machine learning model [k(new)] to refer to the first answers [1] to [k(new)-1] obtained from each machine learning model [1] to [k(new)-1] whose order after the change is earlier than that of the machine learning model [k(new)]. Thereby, the dialogue generation unit 120 acquires a new first answer [k(new)] corresponding to the first request from the machine learning model [k(new)]. Note that the dialogue generation unit 120 may acquire a new nth answer based on an order change instruction for the nth answer set.

[0085] For example, a user who has confirmed the answers of each machine learning model shown in FIG. 8 wants to know more about the answer of the supporter A model. Specifically, the user wants to know the answer of the supporter A model based on the answers of other machine learning models (supporter B model, supporter C model) to his / her question.

[0086] In this case, the user gives an "order change instruction" to the dialogue generation unit 120 so that the supporter A model answers last. Although the specific operation content of the order change instruction will be described later, as shown in FIG. 11A, the user performs an operation to move the first answer [1] of the supporter A model to the first answer [3] of the last supporter C model.

[0087] Upon receiving an order change instruction, the dialogue generation unit 120 generates a query for input to the machine learning model whose order has been changed by the user. Specifically, the dialogue generation unit 120 generates a query based on the first request, the scenario, the first answer [2] of the machine learning model that has already been answered, and the first answer [3].

[0088] The dialogue generation unit 120 inputs the generated query to the machine learning model whose order has been changed by the user. As described above, when the teacher model is changed to answer last, the dialogue generation unit 120 obtains the first answer [3(new)] from the supporter A model. The dialogue generation unit 120 displays the obtained first answer [3(new)] on the display screen 21 of the terminal device 20 (see FIG. 11B).

[0089] The inputs and outputs related to the three machine learning models in the flip function are summarized as shown in FIG. 12. As shown in FIG. 12, the supporter A model selected by the dialogue generation unit 120 as the machine learning model that answers first in the basic function has its order changed to be the machine learning model that answers last by the user. That is, the order of answering, such as the supporter A model, the supporter B model, and the supporter C model, is changed to the order of answering, such as the supporter B model, the supporter C model, and the supporter A model, by the user utilizing the flip function.

[0090] The presentation unit 15 shown in FIG. 1 presents to the user by displaying, on the display screen 21 of the terminal device 20, the evaluation result 186 exemplified in FIG. 5, for example, by the evaluation unit 14.

[0091] The scenario generation unit 16 generates a dialogue scenario from the user management information 183 based on a scenario generation criterion 187 for generating a dialogue scenario according to the characteristics of the user represented by the user management information 183. The scenario generation unit 16 may be a machine learning model including a large language model. The scenario generation unit 16 may store information representing the generated scenario in the storage unit 18.

[0092] The user management information 183 represents the characteristics of the user based on at least any one of, for example, attributes including age, gender, occupation, hobbies, personality, etc. related to the user, characteristics indicated by the record of the previous dialogue content 184 using the dialogue ability improvement support device 10, and characteristics indicated by the record of the evaluation results 186. The user management information 183 may be generated or updated, for example, by the user himself / herself, or may be generated or updated by the acquisition unit 13 or the evaluation unit 14. The acquisition unit 13 may, for example, acquire the characteristics of the user including the user's hobbies by analyzing the acquired dialogue content 184, and may reflect the acquired characteristics of the user in the user management information 183. The acquisition unit 13 may, for example, as shown in FIG. 6A, estimate the trend of the keywords included in the display area from the history of the display area representing the utterances by the machine learning model that the user has dragged and moved to the bottom of the screen so far, and acquire the characteristics of the user including the user's hobbies from the estimation result.

[0093] An example of the generation of a scenario by the scenario generation unit 16 will be described. For example, in a past dialogue in which a certain user used the dialogue ability improvement support device 10, the machine learning model X and the machine learning model Y, which are participants, opposed each other, and although the user was in a position to mediate between the machine learning model X and the machine learning model Y in that scenario, it is assumed that the mediation failed because the user sided with either the machine learning model X or the machine learning model Y. In this case, the evaluation unit 14 generates an evaluation result 186 indicating that the user lacks the dialogue ability to mediate between the opposing parties, and reflects the feature that the user lacks the dialogue ability to mediate between the opposing parties in the user management information 183. Specifically, for example, in a dialogue scenario where the future business direction is discussed, the evaluation unit 14 identifies from the keywords included in the outputs from the machine learning models X and Y that the machine learning model X advocates the importance of price and the machine learning model Y advocates the importance of performance. Then, the evaluation unit 14 identifies from the keywords included in the user's speech that the user advocates the importance of price without considering the importance of performance (i.e., sides with the machine learning model X). The evaluation unit 14 further identifies from the keywords included in the output from the machine learning model Y that the machine learning model Y has great dissatisfaction in response to the user's speech. Thereby, the evaluation unit 14 determines that the mediation between the user and the machine learning models X and Y has failed. Then, the scenario generation unit 16 newly generates a dialogue scenario in which a plurality of machine learning models with different positions oppose each other so that the user can learn the dialogue ability to appropriately mediate between the opposing parties. In this case, the scenario generation criterion 187 represents that when the user lacks the dialogue ability to mediate between the opposing parties, a dialogue scenario in which a plurality of machine learning models with different positions oppose each other is generated.

[0094] The machine learning model generation unit 17 trains (generates or updates) a machine learning model, for example, by learning the words and actions of real or fictional characters. The machine learning model generation unit 17 trains a machine learning model that reproduces ways of thinking including the values of modern celebrities or historical greats, for example, by learning information representing the words and actions of modern celebrities or historical greats. Alternatively, the machine learning model generation unit 17 trains a machine learning model that reproduces ways of thinking including the values of fictional characters that appear in works such as novels, by learning the words and actions of such fictional characters. At this time, the machine learning model generation unit 17 may train the machine learning model by repeatedly predicting the words and actions of the character who will be the model of the machine learning model to be generated. The machine learning model generation unit 17 registers the newly generated machine learning model in the machine learning model management information 182. The machine learning model generation unit 17 may store information representing the generated machine learning model in the storage unit 18.

[0095] The machine learning model generation unit 17 also determines, as a function of the machine learning model to be generated, whether the relevance (direction of the dialogue) between the dialogue content 184 and the dialogue scene meets a predetermined criterion, and if the relevance does not meet the predetermined criterion, includes a function of guiding the dialogue with the user so that the dialogue content 184 meets the predetermined criterion. However, the predetermined criterion is to be given in advance, for example, by the administrator of the dialogue ability improvement support device 10 or the like.

[0096] For example, in a negotiation scenario as illustrated in FIGS. 6A and 6B, assume that the user mentions a topic whose relevance to the negotiation is below a predetermined standard (for example, a topic related to a famous person or an event with low relevance to the negotiation). In this case, the machine learning model generated by the machine learning model generation unit 17, for example, does not touch on the topic mentioned by the user and makes a statement to return the topic of the conversation to the negotiation. The machine learning model, for example, calculates the distance in the feature vector space for a feature vector calculated from the features of keywords representing the conversation scenario and a feature vector calculated from the features of keywords included in the user's utterance, and may have a function of calculating the relevance (similarity) between the conversation scenario and the user's utterance from the calculated distance.

[0097] Note that since existing technologies can be applied to vectorize text data, detailed descriptions are omitted. For example, existing embedding models such as Word2Vec, FastText, GloVe, etc. or transformer models such as BERT, RoBERTa, GPT, etc. can be used to convert the text data of the user and the machine learning model into feature vectors.

[0098] The recommendation unit 19 has a function of recommending (proposing) at least one of the machine learning models included in the conversation scenario and the participants to a certain user represented by the user management information 183 based on the recommendation criterion 188. However, the recommendation criterion 188 is a criterion for recommending at least one of the machine learning models included in the conversation scenario and the participants to the user according to the characteristics of the user represented by the user management information 183. Note that the recommendation criterion 188 is, for example, given in advance by the administrator of the dialogue ability improvement support device 10 or the like. The recommendation unit 19 displays the recommended result on the display screen 21 of the terminal device 20.

[0099] For example, regarding a user who participated in a dialogue scenario where a machine learning model with generally fast discussion progress and fast individual responses participated, assume that an evaluation result 186 is obtained, indicating that the user's response is slow and the user is not keeping up with the discussion based on the content of the user's speech. In this case, the recommendation unit 19 recommends, as participants in the dialogue, a combination of machine learning models that make the discussion progress slower and have slower individual responses, based on the characteristics of the user indicated by the user management information 183 in which the above-described evaluation result 186 is reflected. Conversely, regarding a user who participated in a dialogue scenario where a machine learning model with generally slow discussion progress and slow individual responses participated, there may be a case where an evaluation result 186 is obtained, indicating that the user's response is fast and the user can respond to a faster discussion based on the content of the user's speech. In this case, the recommendation unit 19 recommends, as participants in the dialogue, a combination of machine learning models that make the discussion progress faster and have faster individual responses, based on the characteristics of the user indicated by the user management information 183 in which the above-described evaluation result 186 is reflected.

[0100] Also, regarding a user who participated in a dialogue scenario where a machine learning model with high expertise in a certain technical field participated, assume that an evaluation result 186 is obtained, indicating that the user makes few speeches and does not fully understand the technical content being discussed based on the content of the user's speech. In this case, the recommendation unit 19 recommends, as participants in the dialogue, a combination of machine learning models with slightly lower expertise in that technical field, based on the characteristics of the user indicated by the user management information 183 in which the above-described evaluation result 186 is reflected. Conversely, regarding a user who participated in a dialogue scenario where a machine learning model with not very high expertise in a certain technical field participated, assume that an evaluation result 186 is obtained, indicating that the user can respond to a discussion with higher expertise based on the content of the user's speech. In this case, the recommendation unit 19 recommends, as participants in the dialogue, a combination of machine learning models with higher expertise in that technical field, based on the characteristics of the user indicated by the user management information 183 in which the above-described evaluation result 186 is reflected.

[0101] As described above, the recommendation unit 19 recommends to the user, as a participant in the conversation, a combination of machine learning models capable of conducting a conversation commensurate with the user's conversation ability. That is, the recommendation criterion 188 correlates the user's conversation ability with the conversation ability of the machine learning models.

[0102] Also, the recommendation criterion 188 may be a criterion for recommending to the user at least either the conversation scenario or the machine learning model included in the participants in the conversation according to a situation where the user's characteristics change over time. However, in this case, the evaluation unit 14 shall manage the user management information 183 representing the characteristics of the user in time series. And in this case, based on the fact that the user's conversation ability gradually improves as the user uses the conversation ability improvement support apparatus 10, the recommendation unit 19 recommends at least either the conversation scenario or a combination of machine learning models capable of conducting a conversation according to the conversation ability of the user, as a participant in the conversation.

[0103] Next, with reference to the flowchart of FIG. 13, the operation (processing) of the conversation ability improvement support apparatus 10 according to the present disclosure will be described in detail.

[0104] The reception unit 11 receives information indicating the start of use of the conversation ability improvement support apparatus 10, which is input from the terminal device 20 via an input operation by the user on the terminal device 20 (step S101). The recommendation unit 19 displays, on the display screen 21, a selection menu of the conversation scenario and the machine learning model, including the conversation scenario and the machine learning model to be recommended to the user, based on the scenario management information 181, the machine learning model management information 182, the user management information 183, and the recommendation criterion 188 (step S102).

[0105] The reception unit 11 receives information representing the dialogue scenario and the machine learning model selected by the user in the selection menu (step S103). The construction unit 12 constructs an environment in which the selected machine learning model and the user interact in the dialogue scenario selected by the user (step S104). The acquisition unit 13 acquires the dialogue content 184 between the user and the machine learning model that has taken place in the environment constructed by the construction unit 12 (step S105).

[0106] When the relevance between the dialogue content 184 and the dialogue scenario satisfies a predetermined criterion (Yes in step S106), the process proceeds to step S108. When the relevance between the dialogue content 184 and the dialogue scenario does not satisfy the predetermined criterion (No in step S106), the machine learning model participating in the dialogue induces the dialogue with the user so that the dialogue content 184 satisfies the predetermined criterion (step S107).

[0107] When the dialogue content 184 does not indicate that the dialogue has been completed (i.e., indicates that the dialogue is ongoing) (No in step S108), the process returns to step S105. When the dialogue content 184 indicates that the dialogue has been completed (Yes in step S108), the evaluation unit 14 evaluates the user's dialogue ability based on the dialogue content 184 and the evaluation criterion 185 (step S109). The presentation unit 15 displays the evaluation result 186 of the user's dialogue ability on the display screen 21 of the terminal device 20 (step S110), and the overall process ends.

[0108] The dialogue ability improvement support device 10 according to the present disclosure can support the improvement of the user's dialogue ability assuming various dialogue scenarios. The reason is that the dialogue ability improvement support device 10 constructs an environment in which participants including the user and one or more machine learning models interact in various scenarios, evaluates the user's dialogue ability from the dialogue content, and presents the evaluation result to the user.

[0109] Hereinafter, the effects achieved by the dialogue ability improvement support device 10 according to the present disclosure will be described in detail.

[0110] For example, many people with developmental disabilities struggle to communicate effectively with others in their daily lives. There is a therapeutic support system called AI-PAC (ABA Integrated Program for Autism speCtrum disorders) for people with developmental disabilities, but since AI-PAC focuses on the function of recommending teaching materials, it often cannot be used for actual conversation practice or in various situations that people with developmental disabilities face in their daily lives. Also, not only people with developmental disabilities but also healthy individuals may suffer disadvantages due to failures in communicating with others in various situations in their daily lives. To solve such problems, it is necessary to support the improvement of the user's communication ability assuming various communication scenarios.

[0111] In response to such problems, the conversation ability improvement support device 10 according to the present disclosure receives information for respectively selecting a scenario in which participants including a user and one or more machine learning models interact with each other and the machine learning models included in the participants. The conversation ability improvement support device 10 constructs an environment in which the participants interact with each other in the selected scenario. The conversation ability improvement support device 10 acquires the conversation content 184 between the user and the machine learning model in the environment. The conversation ability improvement support device 10 evaluates the user's conversation ability from the acquired conversation content 184 based on an evaluation criterion 185 for evaluating the conversation ability according to the conversation content 184. Then, the conversation ability improvement support device 10 presents the evaluation result 186 of the conversation ability to the user. That is, the conversation ability improvement support device 10 provides the user with an environment for interacting with a machine learning model assuming various conversation scenarios, and presents (feeds back) the evaluation result of the user's conversation ability based on the conversation content to the user, thereby supporting the improvement of the user's conversation ability assuming various conversation scenarios.

[0112] In addition, the dialogue ability improvement support device 10 according to the present disclosure generates a dialogue scenario based on a scenario generation criterion 187 for generating a dialogue scenario according to the characteristics of the user from user management information 183 representing the characteristics of the user. Thereby, since the dialogue ability improvement support device 10 generates a dialogue scenario to be a learning target according to the characteristics of the user, it is possible to surely support the improvement of the user's dialogue ability assuming various dialogue scenarios.

[0113] In addition, the dialogue ability improvement support device 10 according to the present disclosure acquires user management information 183 representing the characteristics of the user including the user's preferences from the dialogue content 184. Thereby, since the dialogue ability improvement support device 10 can manage the characteristics of the user with high accuracy, it is possible to surely support the improvement of the user's dialogue ability assuming various dialogue scenarios.

[0114] In addition, the dialogue ability improvement support device 10 according to the present disclosure trains a machine learning model participating in the dialogue by learning the actions of real or fictional characters. Thereby, the dialogue ability improvement support device 10 can efficiently generate a machine learning model having various characteristics.

[0115] In addition, as a function of the machine learning model, the dialogue ability improvement support device 10 according to the present disclosure determines whether the relevance between the dialogue content 184 and the dialogue scenario satisfies a predetermined criterion, and if the relevance does not satisfy the predetermined criterion, includes a function of guiding the dialogue with the user so that the dialogue content 184 satisfies the predetermined criterion. That is, since the dialogue ability improvement support device 10 generates a machine learning model having a function of proceeding with the dialogue so that the content of the dialogue does not diverge, it is possible to surely support the improvement of the user's dialogue ability assuming various dialogue scenarios.

[0116] In addition, the dialogue ability improvement support device 10 according to the present disclosure recommends at least one of the dialogue scene and the machine learning model participating in the dialogue to the user based on a recommendation criterion 188 for recommending at least one of the dialogue scene and the machine learning model participating in the dialogue to the user according to the characteristics of the user. As a result, the user can appropriately perform learning to improve the dialogue ability, so that the dialogue ability improvement support device 10 can surely support the improvement of the user's dialogue ability assuming various dialogue scenes.

[0117] Further, the dialogue ability improvement support device 10 according to the present disclosure may manage information representing the characteristics of the user in time series, and use a recommendation criterion 188 for recommending at least one of the dialogue scene and the machine learning model included in the participants to the user according to the situation where the characteristics of the user change over time. In this case, the dialogue ability improvement support device 10 recommends at least one of the dialogue scene and the machine learning model included in the participants to the user from the situation where the characteristics of the user change over time. That is, since the dialogue ability improvement support device 10 recommends the machine learning model included in the dialogue scene and the participants based on the fact that the user's dialogue ability improves over time as the user accumulates dialogue learning, it can surely support the improvement of the user's dialogue ability assuming various dialogue scenes.

[0118] In addition, the dialogue ability improvement support device 10 according to the present disclosure may notify the evaluation result 186 to, for example, the guardians and supporters of users with developmental disabilities.

[0119] <Second Embodiment> FIG. 14 is a block diagram showing the configuration of a dialogue ability improvement support device 30 according to the present disclosure. The dialogue ability improvement support device 30 includes a reception unit 31, a construction unit 32, an acquisition unit 33, and an evaluation unit 34. The reception unit 31, the construction unit 32, the acquisition unit 33, and the evaluation unit 34 are examples of reception means, construction means, acquisition means, and evaluation means, respectively.

[0120] The reception unit 31 receives information for selecting, respectively, a scene 312 where participants including a user and one or more machine learning models 311 interact with each other, and the machine learning model 311 included in the participants. The machine learning model 311 is, for example, a machine learning model similar to the machine learning model managed by the machine learning model management information 182 related to the dialogue ability improvement support device 10. The scene 312 is, for example, a scene similar to the scene managed by the scene management information 181 related to the dialogue ability improvement support device 10. The reception unit 31 operates in the same manner as, for example, the reception unit 11 related to the dialogue ability improvement support device 10.

[0121] The construction unit 32 constructs an environment 321 in which the participants interact with each other in the selected scene 312. The environment 321 is, for example, an environment similar to the dialogue environment constructed by the construction unit 12 related to the dialogue ability improvement support device 10. The construction unit 32 operates in the same manner as, for example, the construction unit 12 related to the dialogue ability improvement support device 10.

[0122] The acquisition unit 33 acquires the dialogue content 331 between the user and the machine learning model 311 in the environment 321. The dialogue content 331 is, for example, information similar to the dialogue content 184 related to the dialogue ability improvement support device 10. The acquisition unit 33 operates in the same manner as, for example, the acquisition unit 13 related to the dialogue ability improvement support device 10.

[0123] The evaluation unit 34 evaluates the dialogue ability of the user from the acquired dialogue content 331 based on an evaluation criterion 341 for evaluating the dialogue ability according to the dialogue content 331. The evaluation criterion 341 is, for example, a criterion similar to the evaluation criterion 185 related to the dialogue ability improvement support device 10. The evaluation unit 34 operates in the same manner as, for example, the evaluation unit 14 related to the dialogue ability improvement support device 10.

[0124] Next, with reference to the flowchart of FIG. 15, the operation (processing) of the dialogue ability improvement support device 30 according to the present disclosure will be described in detail.

[0125] The reception unit 31 receives information for selecting, respectively, a scene 312 in which participants including a user and one or more machine learning models 311 interact with each other, and the machine learning model 311 included in the participants (step S201). The construction unit 32 constructs an environment 321 in which the participants interact with each other in the selected scene 312 (step S202).

[0126] The acquisition unit 33 acquires the conversation content 331 between the user and the machine learning model 311 in the environment 321 (step S203). The evaluation unit 34 evaluates the user's conversation ability from the acquired conversation content 331 based on the evaluation criteria 341 (step S204), and the overall process ends.

[0127] The conversation ability improvement support device 30 according to the present disclosure can support the improvement of the user's conversation ability assuming various scenes. The reason is that the conversation ability improvement support device 30 constructs an environment in which participants including a user and one or more machine learning models interact in various scenes, evaluates the user's conversation ability from the conversation content, and presents the evaluation result to the user.

[0128] <Hardware Configuration Example> In each of the above-described embodiments, each unit in the conversation ability improvement support devices 10 and 30 shown in FIGS. 1 and 14 can be realized by dedicated HW (Hardware) (electronic circuit). Further, in FIGS. 1 and 8, at least the following configurations can be regarded as functional (processing) units (software modules) of a software program including instructions executed by a processor. · The reception units 11 and 31, · The construction units 12 and 32, · The acquisition units 13 and 33, · The evaluation units 14 and 34, · The presentation unit 15, · The scene generation unit 16, · The machine learning model generation unit 17, · The recommendation unit 19, · The storage control function in the storage unit 18.

[0129] However, the division of each part shown in these drawings is a configuration for convenience of explanation, and various configurations can be assumed in implementation. An example of the hardware environment in this case will be described with reference to FIG. 16.

[0130] FIG. 16 is a diagram exemplarily explaining the configuration of an information processing apparatus 900 (computer) capable of realizing the dialogue ability improvement support apparatus according to the present disclosure. That is, FIG. 16 shows the configuration of a computer (information processing apparatus) capable of realizing the dialogue ability improvement support apparatus shown in FIGS. 1 and 14, and represents a hardware environment capable of realizing each function in the above-described embodiment. However, each part in the above-described dialogue ability improvement support apparatus may be provided dispersedly in a plurality of information processing apparatuses 900, or at least some of its functions may be provided in a server or the like constituting a cloud computing environment.

[0131] The information processing apparatus 900 shown in FIG. 10 includes the following as components. · CPU (Central Processing Unit) 901, · ROM (Read Only Memory) 902, · RAM (Random Access Memory) 903, · Hard disk (storage device) 904, · Communication interface 905, · Bus 906 (communication line), · Reader / writer 908 capable of reading and writing data stored in a recording medium 907 such as a CD-ROM (Compact Disc Read Only Memory), · Input / output interface 909 such as a monitor, speaker, and keyboard.

[0132] That is, the information processing apparatus 900 including the above-described components is a general computer in which these components are connected via a bus 906. The information processing apparatus 900 may include a plurality of CPUs 901, or may include a CPU 901 configured by a multi-core. The information processing apparatus 900 may also not include a part of the above-described configuration.

[0133] Also, the processor used by the information processing apparatus 900 is not limited to a CPU. For example, the processor may be an MPU (Micro Processing Unit), a DSP (Digital Signal Processor), a TPU (Tensor Processing Unit), or a GPU (Graphics Processing Unit). Alternatively, the processor may be an FPGA (Field Programmable Gate Array), an ASIC (Application Specific Integrated Circuit), or the like. The processor executes various programs including an operating system (OS; Operating System).

[0134] Then, the present invention described by taking the above-described embodiment as an example supplies a computer program capable of realizing the following functions to the information processing apparatus 900 shown in FIG. 16. The functions are the above-described configurations in the block diagrams (FIGS. 1 and 14) referred to in the description of the embodiment, or the functions of the flowcharts (FIGS. 13 and 15). The present invention is then achieved by reading the computer program into the CPU 901 of the hardware, interpreting, and executing it. Further, the computer program supplied into the apparatus may be stored in a readable and writable volatile memory (RAM 903), or a non-volatile storage device such as a ROM 902 or a hard disk 904.

[0135] In the above case, a general procedure at present can be adopted as a method for supplying the computer program into the hardware. Examples of the procedure include a method of installing it into the apparatus via various recording media 907 such as a CD-ROM, and a method of downloading it from the outside via a communication line such as the Internet. In such a case, the present invention can be regarded as being constituted by the code constituting such a computer program or the recording medium 907 storing the code.

[0136] The present invention has been described by taking the above-described embodiments as exemplary examples. However, the present invention is not limited to the above-described embodiments. That is, within the scope of the present invention, various aspects understandable by those skilled in the art can be applied.

[0137] Note that part or all of each of the above-described embodiments may also be described as follows. However, the present invention exemplified by each of the above-described embodiments is not limited to the following.

[0138] (Appendix 1) Reception means for receiving information for respectively selecting a scene in which participants including a user and one or more machine learning models interact with each other, and the machine learning models included in the participants; Construction means for constructing an environment in which the participants interact with each other in the selected scene; Acquisition means for acquiring the dialogue content between the user and the machine learning model in the environment; Evaluation means for evaluating the dialogue ability of the user from the acquired dialogue content based on an evaluation criterion for evaluating the dialogue ability according to the dialogue content; A dialogue ability improvement support device comprising:

[0139] (Appendix 2) Further comprising scene generation means for generating the scene from information representing the characteristics of the user based on a scene generation criterion for generating the scene according to the characteristics of the user. The dialogue ability improvement support device according to Appendix 1.

[0140] (Appendix 3) The information representing the characteristics of the user represents at least any one of the age, gender, occupation, hobbies, personality, achievements of the dialogue content, and achievements of the evaluation results of the dialogue ability of the user. The dialogue ability improvement support device according to Appendix 2.

[0141] (Appendix 4) The acquisition means acquires information representing the characteristics of the user, including the user's preferences, from the conversation content. The dialogue ability improvement support device according to appendix 2 or appendix 3.

[0142] (Appendix 5) Further comprising a machine learning model generation means for training the machine learning model by learning the actions of real or fictional characters. The dialogue ability improvement support device according to any one of appendices 1 to 4.

[0143] (Appendix 6) As a function of the machine learning model, the machine learning model generation means determines whether the relevance between the conversation content and the scene meets a predetermined criterion, and if the relevance does not meet the predetermined criterion, includes a function of guiding the conversation with the user so that the conversation content meets the predetermined criterion. The dialogue ability improvement support device according to appendix 5.

[0144] (Appendix 7) Further comprising a recommendation means for recommending at least one of the scene and the machine learning model to the user based on a recommendation criterion for recommending at least one of the scene and the machine learning model to the user according to the characteristics of the user. The dialogue ability improvement support device according to any one of appendices 1 to 6.

[0145] (Appendix 8) The evaluation means manages the information representing the characteristics of the user in time series. The recommendation means recommends at least one of the scene and the machine learning model to the user based on the recommendation criterion for recommending at least one of the scene and the machine learning model to the user according to the situation where the characteristics of the user change over time, from the situation where the characteristics of the user change over time. The dialogue ability improvement support device according to appendix 7.

[0146] (Appendix 9) An information processing apparatus receives information for respectively selecting a scenario in which participants including a user and one or more machine learning models interact with each other and the machine learning models included in the participants, constructs an environment in which the participants interact with each other in the selected scenario, obtains the conversation content between the user and the machine learning model in the environment, evaluates the conversation ability of the user from the obtained conversation content based on an evaluation criterion for evaluating the conversation ability according to the conversation content, Conversation ability improvement support method.

[0147] (Appendix 10) generates the scenario from information representing the characteristics of the user based on a scenario generation criterion for generating the scenario according to the characteristics of the user, The conversation ability improvement support method according to Appendix 9.

[0148] (Appendix 11) The information representing the characteristics of the user represents at least any one of the age, gender, occupation, hobbies, personality, track record of the conversation content, and track record of the evaluation result of the conversation ability of the user, The conversation ability improvement support method according to Appendix 10.

[0149] (Appendix 12) obtains information representing the characteristics of the user including the hobbies of the user from the conversation content, The conversation ability improvement support method according to Appendix 10 or Appendix 11.

[0150] (Appendix 13) trains the machine learning model by learning the actions and words of real or fictional characters, The conversation ability improvement support method according to any one of Appendices 9 to 12.

[0151] (Appendix 14) As a function of the machine learning model, determine whether the relevance between the conversation content and the scenario meets a predetermined criterion, and if the relevance does not meet the predetermined criterion, include a function of guiding the conversation with the user so that the conversation content meets the predetermined criterion. The conversation ability improvement support method described in Supplementary Note 13.

[0152] (Supplementary Note 15) Based on a recommendation criterion for recommending at least one of the scenario and the machine learning model to the user according to the characteristics of the user, recommend at least one of the scenario and the machine learning model to the user from the characteristics of the user. The conversation ability improvement support method according to any one of Supplementary Notes 9 to 14.

[0153] (Supplementary Note 16) Manage the information representing the characteristics of the user in time series. Based on the recommendation criterion for recommending at least one of the scenario and the machine learning model to the user according to the situation where the characteristics of the user change over time, recommend at least one of the scenario and the machine learning model to the user from the situation where the characteristics of the user change over time. The conversation ability improvement support method described in Supplementary Note 15.

[0154] (Supplementary Note 17) A reception process for receiving information for selecting a scenario in which participants including a user and one or more machine learning models interact with each other, and the machine learning models included in the participants, A construction process for constructing an environment in which the participants interact with each other in the selected scenario, An acquisition process for acquiring the conversation content between the user and the machine learning model in the environment, An evaluation process for evaluating the conversation ability of the user from the acquired conversation content based on an evaluation criterion for evaluating the conversation ability according to the conversation content, A conversation ability improvement support program for causing a computer to execute.

[0155] (Appendix 18) For further causing the computer to execute a scene generation process of generating the scene from information representing the characteristics of the user based on a scene generation criterion for generating the scene according to the characteristics of the user, The dialogue ability improvement support program according to Appendix 17.

[0156] (Appendix 19) The information representing the characteristics of the user represents at least any one of the age, gender, occupation, hobbies, personality, record of the dialogue content, and record of the evaluation result of the dialogue ability of the user. The dialogue ability improvement support program according to Appendix 18.

[0157] (Appendix 20) The acquisition process acquires information representing the characteristics of the user including the hobbies of the user from the dialogue content. The dialogue ability improvement support program according to Appendix 18 or Appendix 19.

[0158] (Appendix 21) For further causing the computer to execute a machine learning model generation process of training the machine learning model by learning the actions of real or fictional characters, The dialogue ability improvement support program according to any one of Appendices 17 to 20.

[0159] (Appendix 22) The machine learning model generation process includes, as a function of the machine learning model, determining whether the relevance between the dialogue content and the scene satisfies a predetermined criterion, and if the relevance does not satisfy the predetermined criterion, including a function of guiding the dialogue with the user so that the dialogue content satisfies the predetermined criterion. The dialogue ability improvement support program according to Appendix 21.

[0160] (Appendix 23) Based on the recommendation criteria for recommending at least one of the scenarios and the machine learning model to the user according to the characteristics of the user, further causing the computer to execute a recommendation process for recommending at least one of the scenarios and the machine learning model to the user from the characteristics of the user. The dialogue ability improvement support program according to any one of Appendices 17 to 22.

[0161] (Appendix 24) The evaluation process manages the information representing the characteristics of the user in time series. The recommendation process is based on the recommendation criteria for recommending at least one of the scenarios and the machine learning model to the user according to the situation where the characteristics of the user change over time, and recommends at least one of the scenarios and the machine learning model to the user from the situation where the characteristics of the user change over time. The dialogue ability improvement support program according to Appendix 23.

Explanation of symbols

[0162] 10 Dialogue ability improvement support device 11 Reception unit 12 Construction unit 13 Acquisition unit 14 Evaluation unit 15 Presentation unit 16 Scenario generation unit 17 Machine learning model generation unit 18 Storage unit 19 Recommendation unit 181 Scenario management information 182 Machine learning model management information 183 User management information 184 Dialogue content 185 Evaluation criteria 186 Evaluation result 187 Scenario generation criteria 188 Recommendation criteria 20 Terminal device 21 Display screen 30 Dialogue ability improvement support device 31 Reception unit 311 Machine learning model 312 Scenario 32 Construction section 321 Environment 33 Acquisition section 331 Conversation content 34 Evaluation section 341 Evaluation criteria 900 Information processing apparatus 901 CPU 902 ROM 903 RAM 904 Hard disk (storage device) 905 Communication interface 906 Bus 907 Recording medium 908 Reader / writer 909 Input / output interface

Claims

1. Receiving means for receiving information for selecting a scene in which participants including a user and one or more machine learning models interact with each other and the machine learning models included in the participants, respectively; Constructing means for constructing an environment in which the participants interact with each other in the selected scene; Obtaining means for obtaining the dialogue content between the user and the machine learning model in the environment; Evaluation means for evaluating the dialogue ability of the user from the obtained dialogue content based on an evaluation criterion for evaluating the dialogue ability according to the dialogue content; A dialogue ability improvement support device comprising:

2. Further comprising scene generation means for generating the scene based on information representing the characteristics of the user based on a scene generation criterion for generating the scene according to the characteristics of the user; The dialogue ability improvement support device according to claim 1.

3. The information representing the characteristics of the user represents at least any one of the age, gender, occupation, hobbies, personality, performance of the dialogue content, and performance of the evaluation result of the dialogue ability of the user; The dialogue ability improvement support device according to claim 2.

4. The obtaining means obtains information representing the characteristics of the user including the hobbies of the user from the dialogue content; The dialogue ability improvement support device according to claim 2 or claim 3.

5. Further comprising machine learning model generation means for training the machine learning model by learning the actions of real or fictional characters; The dialogue ability improvement support device according to claim 1 or claim 2.

6. As a function of the machine learning model, the machine learning model generation means determines whether the relevance between the dialogue content and the scene satisfies a predetermined criterion, and if the relevance does not satisfy the predetermined criterion, includes a function of inducing a dialogue with the user so that the dialogue content satisfies the predetermined criterion; The dialogue ability improvement support device according to claim 5.

7. Further comprising recommendation means for recommending at least any one of the scene and the machine learning model to the user based on a recommendation criterion for recommending at least any one of the scene and the machine learning model to the user according to the characteristics of the user; The dialogue ability improvement support device according to claim 1 or claim 2.

8. The evaluation means manages the information representing the characteristics of the user in time series; The recommendation means recommends at least one of the scenario and the machine learning model to the user based on the recommendation criteria for recommending at least one of the scenario and the machine learning model to the user according to the situation where the characteristics of the user change over time. The dialogue ability improvement support device according to claim 7.

9. By an information processing device receives information for respectively selecting a scenario in which participants including a user and one or more machine learning models interact with each other and the machine learning models included in the participants, constructs an environment in which the participants interact with each other in the selected scenario, acquires the dialogue content between the user and the machine learning model in the environment, evaluates the dialogue ability of the user from the acquired dialogue content based on an evaluation criterion for evaluating the dialogue ability according to the dialogue content, A dialogue ability improvement support method.

10. a reception process of receiving information for respectively selecting a scenario in which participants including a user and one or more machine learning models interact with each other and the machine learning models included in the participants, a construction process of constructing an environment in which the participants interact in the selected scenario, an acquisition process of acquiring the dialogue content between the user and the machine learning model in the environment, an evaluation process of evaluating the dialogue ability of the user from the acquired dialogue content based on an evaluation criterion for evaluating the dialogue ability according to the dialogue content, A dialogue ability improvement support program for causing a computer to execute.

Images