Information processing device, information processing method, and information processing program
The information processing system addresses the psychological burden of first-time communication by generating and evaluating utterance candidates for Digital Twin partners, enabling users to make informed choices and create positive impressions, thus improving interaction quality.
Patent Information
- Application Number
- PCT/JP2023/019401
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-05-24
- Publication Date
- 2024-11-28
AI Technical Summary
Users face a significant psychological burden when communicating with someone for the first time, as they cannot gauge the other person's impressions or reactions until they meet, leading to uncertainty and anxiety in matchmaking, sales, and recruitment scenarios.
An information processing system that generates utterance candidates for a Digital Twin dialogue partner, calculates an evaluation value based on conversation history and personality, and provides feedback to users on their utterance choices, helping them make impressions that are likely to be well-received.
This system reduces the psychological burden of initial communication by allowing users to assess and improve their responses, ensuring they leave a positive impression, thereby enhancing interaction outcomes.
Smart Images

Figure JP2023019401_28112024_PF_FP_ABST
Abstract
Description
Information processing device, information processing method, and information processing program
[0001] The present invention relates to an information processing device, an information processing method, and an information processing program.
[0002] Conventionally, there are matching services in the fields of matchmaking, sales, recruitment, etc. that present potential partners who are compatible with each other based on information such as a user's profile, attributes, and behavioral history (see Non-Patent Documents 1 and 2).
[0003] Best Dating Sites for Real Relationships in 2023, [online], [accessed March 9, 2023], Internet <https: / / chicagoreader.com / reader-partners / best-dating-sites-for-real-relationships / > Dating Apps Thrive in China, but Not Just for Romance, [online], [accessed March 9, 2023], Internet <https: / / www.nytimes.com / 2022 / 09 / 27 / technology / dating-apps-china.html>
[0004] However, the atmosphere of the person presented in the matching service (e.g., behavior, reactions when speaking, etc., which cannot be seen in a photograph), impressions, etc., cannot be known until the user actually meets the person and communicates with them. Therefore, there is a problem that communicating with a person presented in the matching service for the first time can be a heavy psychological burden for the user. Therefore, an object of the present invention is to solve the above-mentioned problem and reduce the psychological burden of communicating with a person for the first time.
[0005] In order to solve the above-mentioned problems, the present invention is characterized by comprising an utterance generation unit that generates utterance candidates for a DT (Digital Twin) of a dialogue partner in a dialogue between the user and the DT of the dialogue partner, and an utterance evaluation unit that calculates an evaluation value that quantifies the impression that the dialogue partner has of the user's utterance candidates based on a dialogue history between the user and the DT of the dialogue partner, including the user's utterance candidates, and information indicating the personality of the dialogue partner, and outputs the evaluation value for each of the user's utterance candidates.
[0006] According to the present invention, it is possible to reduce the psychological burden that occurs when communicating with someone you meet for the first time.
[0007] FIG. 1 is a diagram illustrating an overview of an information processing system according to a first embodiment. FIG. 2 is a diagram illustrating an example of processing performed by the information processing system according to the first embodiment. FIG. 3 is a diagram illustrating an example of the configuration of the information processing system according to the first embodiment. FIG. 4 is a diagram illustrating information indicating the individuality of a dialogue partner input to the utterance generation unit of FIG. 3 and examples of utterance candidates for the dialogue partner output by the utterance generation unit. FIG. 5 is a diagram illustrating an example of utterance candidates generated by the utterance generation unit of FIG. 3. FIG. 6 is a diagram illustrating an example of data input to a deep learning model. FIG. 7 is a diagram illustrating an example of the hardware configuration of the information processing system according to the first embodiment. FIG. 8 is a diagram illustrating an overview of processing performed by the information processing system according to the first embodiment. FIG. 9 is a flowchart illustrating an example of processing procedures performed by the information processing system according to the first embodiment. FIG. 10 is a diagram illustrating an example of processing performed by the information processing system according to a second embodiment. FIG. 11 is a diagram illustrating an example of the configuration of an information processing system according to the second embodiment. FIG. 12 is a flowchart illustrating an example of processing procedures performed by the information processing system according to the second embodiment. FIG. 13 is a diagram illustrating an example of processing performed by an information processing system according to a third embodiment. Fig. 14 is a diagram illustrating an example of the configuration of an information processing system according to the third embodiment. Fig. 15 is a flowchart illustrating an example of a processing procedure executed by the information processing system according to the third embodiment. Fig. 16 is a diagram illustrating an example of the configuration of a computer that executes an information processing program.
[0008] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, first to third embodiments of the present invention will be described with reference to the accompanying drawings. The present invention is not limited to the respective embodiments.
[0009] First Embodiment First, an overview of an information processing system (information processing device) according to the first embodiment will be described using Fig. 1. When a speaker (a living person) converses with a dialogue partner DT (Digital Twin), the information processing system according to the first embodiment outputs an evaluation value that quantifies the impression that the dialogue partner has of the speaker's utterance candidates.
[0010] For example, the information processing system calculates an evaluation value that quantifies the impression that the dialogue partner has of a speaker's utterance candidate, based on information indicating the dialogue partner's personality and the dialogue history between the dialogue partner and the speaker. Below, an example will be explained where the speaker is person A and the speaker's dialogue partner is person B. Note that the information processing system is equipped with an utterance generation model that generates the dialogue partner's utterance (DT utterance of the dialogue partner) based on information indicating the dialogue partner's personality.
[0011] First, the information processing system accepts input of information indicating the personality of the dialogue partner (person B) (for example, a profile, psychological scale values, etc.). Then, the information processing system generates and outputs a DT utterance for person B by inputting the input information indicating the personality of person B into an utterance generation model. After that, person A inputs utterance candidates A, B, and C (reply candidates) in response to the utterance output from the utterance generation model into the information processing system.
[0012] The information processing system receives input of information indicating the individuality of person B, the dialogue history between person A and the model, and person A's utterance candidates A, B, and C. Using the deep learning model, the system calculates an evaluation value (score) for each of the utterance candidates A, B, and C, and outputs the evaluation value (utterance evaluation).
[0013] This allows person A to check the evaluation values of each of the utterance candidates A, B, and C. For example, if the scores of utterance candidates A, B, and C are the values shown by reference numeral 101, person A can know that utterance candidate A is the utterance that is most likely to give person B the best impression. This allows person A to know which utterance to select in a conversation with person B. As a result, the psychological burden on person A when communicating with person B can be reduced.
[0014] Next, an example of processing executed by the information processing system will be described with reference to Fig. 2. Note that the information processing system will be described here as an example realized by a server and client terminals.
[0015] First, the client terminal receives input from person A, for example, information indicating person A's personality, such as name, hometown, hobbies, occupation, and answers to a questionnaire about personality traits, as well as search conditions for a conversation partner. The above input may be manual or by voice.
[0016] The server uses information input from the client terminal to search for a conversation partner for person A. Then, using information indicating the personality of the searched conversation partner (person B), the server simulates the speech of the conversation partner (person B) and outputs it to the client terminal.
[0017] Person A checks the utterance of the conversation partner (person B) via the client terminal, and inputs utterance candidates (reply candidates) to the utterance into the client terminal. Note that input here can also be done by voice.
[0018] When the server receives reply candidates for person A from the client terminal, it evaluates the reply candidates for person A. The client terminal then displays, for example, the utterance of the conversation partner (person B), person A's reply candidates in response to the utterance, and the score (evaluation value) of each reply candidate. This allows person A to confirm person B's impression of each reply candidate (for example, A, B, and C shown in Figure 2).
[0019] [Configuration Example] Next, a configuration example of an information processing system will be described with reference to Fig. 3. The information processing system includes a personality acquisition unit 11, a search condition acquisition unit 12, an utterance acquisition unit 13, a personality DB (database) 14, a dialogue partner search unit 15, an utterance generation unit 16, an utterance evaluation unit 17, and a result display unit 18.
[0020] The individuality acquisition unit 11 acquires information indicating the individuality of the user (e.g., profile information entered by the user, responses to a questionnaire). The questionnaire may be, for example, a questionnaire that diagnoses the user's personality traits using Bigfive diagnosis.
[0021] [Search Condition Acquisition Unit] The search condition acquisition unit 12 acquires search conditions for a conversation partner of the user. The search conditions are, for example, conditions such as an extroverted person, a person with similar hobbies, a person in the same occupation, etc.
[0022] [Utterance Acquisition Unit] The utterance acquisition unit 13 acquires the user's utterance. The user's utterance may be voice or text.
[0023] [Personality DB] The personality DB 14 stores information indicating the personalities of various users (e.g., profiles, personality traits, etc.). For example, the personality DB 14 stores information indicating the personalities of users acquired by the personality acquisition unit 11. The personality DB 14 also stores information indicating the personalities of other users (e.g., conversation partners of the user) acquired in the past. The personality DB 14 is realized, for example, by a storage device within the information processing system.
[0024] [Dialogue Partner Search Unit] The dialogue partner search unit 15 acquires, based on the search conditions received from the search condition acquisition unit 12, information indicating the personalities of dialogue partners that match the search conditions from the personality DB 14.
[0025] [Utterance Generation Unit] The utterance generation unit 16 generates DT utterances (candidate utterances) of the user's dialogue partner. The utterance generation unit 16 also stores pairs of the user's utterances and the dialogue partner's DT utterances generated in response to the user's utterances as a dialogue history between the two.
[0026] For example, the utterance generation unit 16 includes a natural language processing model that generates utterance candidates for the DT of the dialogue partner in response to the user's utterance based on information indicating the individuality of the dialogue partner and the dialogue history between the user and the DT of the dialogue partner.The utterance generation unit 16 then generates utterance candidates for the DT of the dialogue partner using the natural language processing model.
[0027] For example, the information processing system first trains a large-scale natural language processing model on a combination of information indicating the individuality of the user who conducted the dialogue and a dialogue log of the dialogue, and then constructs in advance a natural language generation model that can generate a response (utterance) to the utterance based on the input information indicating the individuality and the utterance of the dialogue partner. The construction of the natural language generation model uses, for example, the technology described in the following literature 1.
[0028] Reference 1: https: / / ai.googleblog.com / 2020 / 02 / exploring-transfer-learning-with-t5.html, https: / / huggingface.co / rinna / japanese-gpt-1b
[0029] For example, the utterance generation unit 16 generates utterance candidates for the DT (dialogue partner) of the dialogue partner by inputting the user's utterance text acquired by the utterance acquisition unit 13, information indicating the personality of the dialogue partner acquired by the dialogue partner search unit 15, and the dialogue history between the user and the DT of the dialogue partner into the natural language generation model.The utterance generation unit 16 then adds the generated utterance candidates for the dialogue partner to the dialogue history and outputs it.
[0030] 4 shows examples of information indicating the individuality of the dialogue partner input to the utterance generation unit 16 and utterance candidates for the dialogue partner output by the utterance generation unit 16. Note that <extra_id_0> and the like in FIG. 4 indicate special tokens for separating profiles and utterances.
[0031] Here, it is shown that the speech generation unit 16 generated and output "I stayed in Sapporo! I recommend it!", "I recommend a hotel in Sapporo!", etc. as candidate utterances (reply candidates) for the conversation partner in response to <extra_id_2> "Are there any recommended accommodations in Hokkaido?"
[0032] In addition, the utterance generation unit 16 may, for example, calculate the likelihood that each generated utterance candidate will be the next utterance of the user's utterance, and output the utterance candidate with the highest likelihood value (for example, "I stayed in Sapporo! I recommend it!").
[0033] When the utterance generation unit 16 generates the above-mentioned utterance candidates, it may generate up to the top k utterance candidates with the highest likelihood of being the next utterance after the input utterance, for example, by selecting tokens in order from the first token, as shown in Figure 5.
[0034] [Utterance Evaluation Unit] Returning to the explanation of Fig. 3, the utterance evaluation unit 17 calculates an evaluation value that quantifies the impression that the dialogue partner has of the user's utterance candidate based on the dialogue history between the user and the dialogue partner's DT, including the user's utterance candidate, and information indicating the personalities of the user and the dialogue partner. For example, the utterance evaluation unit 17 calculates an evaluation value that quantifies the impression that the dialogue partner has of the user when the dialogue partner sees the user's utterance candidate.
[0035] For example, the utterance evaluation unit 17 acquires information indicating the individuality of the user from the individuality acquisition unit 11. The utterance evaluation unit 17 also acquires information indicating the individuality of the dialogue partner from the dialogue partner search unit 15. The utterance evaluation unit 17 also acquires the user's utterance candidates from the utterance acquisition unit 13, and acquires the dialogue history between the user and the dialogue partner DT from the utterance generation unit 16. The utterance evaluation unit 17 then inputs this acquired information into the deep learning model (deep learning model for utterance evaluation) described above, thereby calculating an evaluation value for each of the user's utterance candidates.
[0036] Then, the utterance evaluation unit 17 outputs the user's utterance candidates, the evaluation values of each utterance candidate, and the dialogue history between the user and the dialogue partner DT to the result display unit 18.
[0037] [Example of Building a Deep Learning Model for Speech Evaluation] An example of building a deep learning model for speech evaluation will be described. First, as a preliminary step, a large-scale natural language processing model is trained to output a relationship value between two speakers (e.g., a value indicating the rapport between the two speakers or Rubin's love scale) from a combination of information indicating the personalities of the two speakers and their dialogue history (see FIG. 6). For training, for example, the technology described in Reference 2 is used.
[0038] Reference 2: https: / / huggingface.co / docs / transformers / model_doc / longformer
[0039] The utterance evaluation unit 17 outputs the interpersonal relationship value between the user and the dialogue partner DT, obtained by inputting the dialogue history between the user and the dialogue partner DT, including the user's utterance candidates, into the above-mentioned large-scale natural language processing model, as an evaluation value for the user's utterance candidates.
[0040] For example, if the interpersonal relationship value output by the large-scale natural language processing model is Rubin's Love Scale, the evaluation value is the amount of change in the love scale before and after an utterance in a conversation. In this case, if the evaluation value is negative, it can be seen that the conversation partner thinks less of the person after the utterance than before. On the other hand, if the evaluation value is positive, it can be seen that the conversation partner thinks more of the person after the utterance than before.
[0041] Returning to the explanation of Fig. 3, the result display unit (output processing unit) 18 outputs an evaluation value for each utterance candidate. For example, the result display unit 18 displays on a screen the user's utterance candidates output from the utterance evaluation unit 17, the evaluation value of each utterance candidate, and the dialogue history between the user and the dialogue partner (DT of the dialogue partner). This allows the user to check, for example, the dialogue history with the dialogue partner, the utterance candidate that the user will utter next, and its evaluation value.
[0042] After the result display unit 18 displays the user's utterance candidates and their evaluation values, if the user inputs or selects a next utterance from the utterance candidates, the utterance acquisition unit 13 acquires the utterance. Then, the utterance generation unit 16 generates an utterance from the conversation partner in response to the utterance.
[0043] 7, among the components of the information processing system, the personality DB 14, the dialogue partner search unit 15, the utterance generation unit 16, and the utterance evaluation unit 17 are provided, for example, in a server. Also, among the components of the information processing system, the personality acquisition unit 11, the search condition acquisition unit 12, the utterance acquisition unit 13, and the result display unit 18 are provided, for example, in a client terminal.
[0044] [Processing Overview] Next, an overview of the processing executed by the information processing system will be described with reference to Fig. 8. First, the dialogue partner search unit 15 of the information processing system searches for information indicating the personality of a dialogue partner from the personality DB 14 using the dialogue partner search conditions acquired by the search condition acquisition unit 12 ((1) Search for dialogue partner).
[0045] Next, the utterance generation unit 16 generates a DT utterance of the dialogue partner using information indicating the individuality of the dialogue partner ((2) Simulated utterance of dialogue partner). Then, the result display unit 18 outputs the generated DT utterance of the dialogue partner.
[0046] The user then inputs their own utterance candidates (reply candidates) in response to the dialogue partner's DT utterance. The utterance evaluation unit 17 then calculates evaluation values for the user's utterance candidates ((3) User's Utterance Evaluation). The utterance evaluation unit 17 then outputs the user's utterance candidates, the evaluation values for each utterance candidate, and the dialogue history to the result display unit 18.
[0047] The information processing system executes the above processes (2) and (3) until the dialogue between the user and the dialogue partner DT ends.
[0048] [Example of Processing Procedure] Next, an example of processing procedure executed by the information processing system will be described with reference to Fig. 9. First, the individuality acquisition unit 11 acquires information indicating the individuality of the user (S1). Then, the search condition acquisition unit 12 acquires search conditions for the user's conversation partner (S2). Then, the conversation partner search unit 15 acquires information indicating the individuality of the conversation partner from the individuality DB 14 using the conversation partner search conditions acquired in S2 (S3).
[0049] After S3, the utterance acquisition unit 13 acquires the user's utterance (S4). Thereafter, the utterance generation unit 16 generates an utterance from the dialogue partner in response to the user's utterance acquired in S4 (S5). Note that if there is already a dialogue history between the user and the dialogue partner, the utterance generation unit 16 adds the dialogue partner's utterance generated in S5 to the dialogue history and outputs it. Thereafter, the result display unit 18 outputs the dialogue partner's utterance generated in S5 (the dialogue history between the user and the dialogue partner including the utterance) (S6).
[0050] After S6, the utterance acquisition unit 13 acquires utterance candidates of the user (candidates for replies to the utterance of the conversation partner) (S7).
[0051] After S7, the utterance evaluation unit 17 evaluates each of the user's utterance candidates based on information indicating the personalities of the user and the conversation partner and the conversation history between the user and the conversation partner (S8).Then, the result display unit 18 outputs each of the user's utterance candidates and their evaluation values (S9).
[0052] After S9, if the utterance acquisition unit 13 has not acquired the user's utterance even after a predetermined time has elapsed, or if the user has input an instruction to end the dialogue, the information processing system determines that the dialogue has ended (Yes in S10) and ends the processing. On the other hand, if the utterance acquisition unit 13 has acquired the user's utterance within the predetermined time, the information processing system determines that the dialogue has not ended yet (No in S10) and returns to the processing of S5.
[0053] By having the information processing system execute the above-described process, the user can learn what kind of utterances to make a good impression on the conversation partner during a conversation with the conversation partner.
[0054] [Second embodiment] Next, an information processing system according to a second embodiment will be described. The information processing system according to the second embodiment performs a dialogue simulation between a user's DT and a dialogue partner's DT. Here, the information processing system calculates an evaluation value that quantifies the impression that each speaker's utterance candidate gives to the other party, and selects utterances that constitute the dialogue simulation based on the evaluation value.
[0055] Next, an example of processing executed by the information processing system will be described with reference to Fig. 10. Note that, again, the information processing system will be described as being realized by a server and client terminals.
[0056] First, as in the first embodiment, the client terminal accepts input of information indicating the individuality of person A, such as name, hometown, hobbies, occupation, and responses to a questionnaire regarding personality traits. The client device also accepts input of conditions for the dialogue simulation (details of which will be described later).
[0057] Next, the server searches for a dialogue partner using the information input from the client terminal. Then, the server simulates the dialogue partner's (person B's) utterances using the input dialogue simulation conditions and information indicating the individuality of the searched dialogue partner (person B). The server also simulates person A's response to person B's utterances using information indicating person A's individuality. In this way, the server simulates a dialogue between persons A and B. Then, the client terminal displays the results of the dialogue simulation by the server.
[0058] When the server selects an utterance for each speaker in the dialogue simulation, it uses an evaluation value that quantifies the impression that the dialogue partner has of each utterance candidate.
[0059] For example, the server selects the utterance with the highest evaluation value for person B from among the utterance candidates for person A, and also selects the utterance with the highest evaluation value for person A from among the utterance candidates for person B. This allows the server to perform a dialogue simulation in which the speakers give each other a good impression.
[0060] [Configuration Example] Next, a configuration example of an information processing system according to the second embodiment will be described with reference to Fig. 11. The same components as those in the first embodiment are denoted by the same reference numerals, and a description thereof will be omitted.
[0061] The information processing system includes a personality acquisition unit 11, a search condition acquisition unit 12, a personality DB 14, a dialogue partner search unit 15, an utterance generation unit 16a, an utterance evaluation unit 17a, a result display unit 18, a simulation condition acquisition unit 19, and an utterance selection unit 20.
[0062] [Utterance Generation Unit] The utterance generation unit 16a simulates a dialogue between a user (person A) and a dialogue partner (person B). For example, the utterance generation unit 16a generates utterance candidates for each of person A and person B in a dialogue between person A (DT of person A) and person B (DT of person A).
[0063] For example, the utterance generation unit 16a inputs a question on a topic common to persons A and B and information indicating the individuality of person B into the natural language processing model (Process 1). Then, the utterance generation unit 16a generates multiple utterance candidates for person B using the natural language processing model (Process 2).
[0064] Next, the utterance generation unit 16a inputs the dialogue history between person A (DT of person A) and person B (DT of person A), the utterance of person B selected from the utterance candidates obtained in process 2, and information indicating the individuality of person A into the natural language processing model, and generates multiple utterance candidates for person A (process 3).
[0065] Thereafter, the utterance generation unit 16a inputs the dialogue history between person A and person B, the utterance of person A selected from the utterance candidates obtained in process 3, and information indicating the individuality of person B into the natural language processing model, and generates multiple utterance candidates for person B (process 4).
[0066] The utterance generating unit 16a simulates a conversation with the persons A and B by repeating the above processes 3 and 4.
[0067] [Utterance Evaluation Unit] The utterance evaluation unit 17a calculates an evaluation value of the utterance candidate generated by the utterance generation unit 16a. For example, the utterance evaluation unit 17a calculates an evaluation value indicating the impression that the utterance candidate of person A or person B gives to the conversation partner, based on the dialogue history between person A (DT of person A) and person B (DT of person A) and information indicating the personalities of persons A and B.
[0068] For example, the utterance evaluation unit 17a calculates an evaluation value for each utterance candidate by inputting information indicating the user's individuality acquired by the individuality acquisition unit 11, information indicating the individuality of the dialogue partner acquired by the dialogue partner search unit 15, the dialogue history between person A (DT of person A) and person B (DT of person A) generated by the utterance generation unit 16a, and utterance candidates for the next turn into the above-mentioned deep learning model.
[0069] For example, the utterance evaluation unit 17a calculates an evaluation value that quantifies the impression that an utterance candidate of person A gives to person B. The utterance evaluation unit 17a also calculates an evaluation value that indicates the impression that an utterance candidate of person B gives to person A.
[0070] The utterance evaluation unit 17a outputs the dialogue history and utterance candidates of persons A and B output from the utterance generation unit 16a, and the evaluation value of each utterance candidate of person A or person B, to the utterance selection unit 20.
[0071] The simulation condition acquisition unit 19 acquires the conditions of the dialogue simulation (dialogue simulation conditions). The conditions of the dialogue simulation include, for example, the number of dialogue turns in the dialogue simulation, and a selection criterion for utterances constituting the dialogue (for example, selecting the utterance candidate with the highest evaluation value).
[0072] [Utterance Selector] The utterance selector 20 selects an utterance candidate that satisfies the conditions of the dialogue simulation acquired by the simulation condition acquirer 19 from among the utterance candidates generated by the utterance generator 16a.
[0073] For example, the utterance selection unit 20 repeats the process of selecting an utterance candidate that satisfies the conditions of the dialogue simulation (e.g., has the highest evaluation value) from each utterance candidate and its evaluation value output from the utterance evaluation unit 17a, for the number of dialogue turns indicated in the conditions of the dialogue simulation.
[0074] For example, consider a case where the evaluation values of utterance candidates A, B, and C received by the utterance selection unit 20 are 0.5, 0.3, and -1.1. In this case, if the condition of the dialogue simulation is "select the utterance candidate with the highest evaluation value," the utterance selection unit 20 selects utterance candidate A as an utterance that constitutes the dialogue.
[0075] If the number of turns in the dialogue including utterance candidate A has not yet reached the above number of turns, the utterance selection unit 20 outputs utterance candidate A to the utterance generation unit 16a. On the other hand, if the number of turns in the dialogue including utterance candidate A has already reached the above number of turns, the utterance selection unit 20 outputs a dialogue history (results of the dialogue simulation) including utterance candidate A to the result display unit 18.
[0076] [Example of Processing Procedure] Next, an example of processing procedure executed by the information processing system will be described with reference to Fig. 12. The processing of S21 to S23 in Fig. 12 is similar to the processing of S1 to S3 in Fig. 9, so the description will be omitted and the description will start from S24 in Fig. 12.
[0077] The simulation condition acquisition unit 19 acquires the simulation conditions of the dialogue (conditions for dialogue simulation) (S24). After that, the utterance generation unit 16a generates utterance candidates for the user or the dialogue partner (S25). The utterance generation unit 16a then outputs the generated utterance candidates and the dialogue history between the user and the dialogue partner, if any, to the utterance evaluation unit 17a.
[0078] After S25, the utterance evaluation unit 17a evaluates each utterance candidate based on information indicating the personalities of the user and the dialogue partner and the dialogue history between the user and the dialogue partner (S26). Then, the utterance selection unit 20 selects utterances that constitute the dialogue from the utterance candidates generated in S25 based on the dialogue simulation conditions acquired in S24 (S27). For example, the utterance selection unit 20 selects the utterance candidate with the highest evaluation value from the utterance candidates generated in S25.
[0079] After S27, if the utterance selection unit 20 determines that the number of dialogue turns including the utterance selected in S27 has reached the number of dialogue turns indicated in the dialogue simulation conditions (Yes in S28), it outputs a dialogue history including the utterance selected in S27 (results of the dialogue simulation) to the result display unit 18. Then, the result display unit 18 outputs the dialogue history (S29).
[0080] On the other hand, if the utterance selection unit 20 determines that the number of dialogue turns including the utterance selected in S27 has not reached the number of dialogue turns indicated in the dialogue simulation conditions (No in S28), the process returns to S25.
[0081] By executing the above process, the information processing system can output the results of a dialogue simulation in which the speakers leave a good impression on each other, allowing the user to know what kind of dialogue with the dialogue partner will help them build a good relationship with each other.
[0082] In the second embodiment, when selecting an utterance that constitutes a dialogue from among the utterance candidates of the user's dialogue partner (person B), the evaluation value of each utterance candidate is used, but this is not limiting. For example, the information processing system may select, from among the utterance candidates of person B, an utterance that is most likely to be the next utterance after the utterance of the user (person A).
[0083] [Third Embodiment] Next, an information processing system according to a third embodiment will be described. The information processing system according to the third embodiment creates multiple patterns of dialogue simulation between a user's DT and a dialogue partner's DT. The information processing system then calculates and outputs evaluation values that quantify the impressions received by the user and the dialogue partner for each dialogue simulation pattern.
[0084] Next, an example of processing executed by the information processing system will be described with reference to Fig. 13. Note that, again, the information processing system will be described as being realized by a server and client terminals.
[0085] As in the second embodiment, the server searches for a dialogue partner using information input from the client terminal. Then, the server simulates the speech of the dialogue partner (person B) using the conditions for the dialogue simulation and information indicating the personality of the searched dialogue partner (person B). The server also simulates person A's response to person B's speech using information indicating person A's personality. In this way, the server simulates a dialogue between persons A and B. The server creates multiple patterns of the results of the above dialogue simulation.
[0086] The server then calculates an evaluation value for each result of the dialogue simulation. The server then outputs the results of the dialogue simulations and their evaluation values to the client terminal. The client terminal displays the results of the dialogue simulations for the dialogue between persons A and B together with their evaluation values (e.g., 0.8, -1.2, 0).
[0087] This allows the user (person A) to know what conversation patterns there are with the conversation partner (person B) and the evaluation value of each conversation pattern.
[0088] [Configuration Example] Next, a configuration example of an information processing system according to the third embodiment will be described with reference to Fig. 14. The same components as those in the first and second embodiments are denoted by the same reference numerals, and the description thereof will be omitted.
[0089] The information processing system includes a personality acquisition unit 11, a search condition acquisition unit 12, a personality DB 14, a dialogue partner search unit 15, an utterance generation unit 16b, an utterance evaluation unit 17b, a result display unit 18, and a simulation condition acquisition unit 19a.
[0090] [Simulation Condition Acquisition Unit] The simulation condition acquisition unit 19a acquires the conditions of the dialogue simulation (dialogue simulation conditions). The dialogue simulation conditions here include, for example, the number of dialogue turns in the dialogue simulation and the number of dialogue simulation patterns.
[0091] [Utterance Generation Unit] The utterance generation unit 16b simulates a dialogue between a user (person A) and a dialogue partner (person B). For example, the utterance generation unit 16b inputs a question on a topic common to persons A and B and information indicating the personality of person A into the natural language processing model (Process 1). Then, the utterance generation unit 16b generates a reply that reflects the personality of person A using the natural language processing model (Process 2).
[0092] Next, the utterance generation unit 16b generates a reply that reflects the individuality of person B by inputting the utterance of person A obtained in process 2 and information indicating the individuality of person B into the natural language processing model (process 3).
[0093] The utterance generation unit 16b performs a dialogue simulation by repeating the above processes 2 and 3 for a specified number of dialogue turns (process 4). Then, the utterance generation unit 16b generates dialogue simulation results for multiple patterns by executing processes 2 to 4 while changing the conditions for the dialogue simulation (process 5). For example, in pattern 1, the utterance generation unit 16b generates a dialogue simulation result in which the most likely utterance from person A's utterance candidates is selected as an answer to the first question. In pattern 2, the utterance generation unit 16b generates a dialogue simulation result in which the next most likely utterance from person A's utterance candidates is selected as the answer. Furthermore, in pattern N, the utterance generation unit 16b generates a dialogue simulation result in which the Nth most likely utterance from person A's utterance candidates is selected as the answer.
[0094] [Utterance Evaluation Unit] The utterance evaluation unit 17b calculates an evaluation value of the result (dialogue history) of the dialogue simulation between the user (person A) and the dialogue partner (person B) generated by the utterance generation unit 16b.
[0095] For example, the utterance evaluation unit 17b inputs information indicating the user's individuality acquired by the individuality acquisition unit 11, information indicating the individuality of the dialogue partner acquired by the dialogue partner search unit 15, and multiple dialogue histories generated by the utterance generation unit 16b into the deep learning model, and calculates evaluation values of the user (person A) and dialogue partner (person B) for each dialogue history.
[0096] Then, the utterance evaluation unit 17b outputs the dialogue histories generated by the utterance generation unit 16b and the evaluation value for each dialogue history to the result display unit 18.
[0097] By executing the above-described process, the information processing system can output the results of a plurality of patterns of dialogue simulation to the user along with their evaluation values.
[0098] [Example of Processing Procedure] Next, an example of processing procedure executed by the information processing system will be described with reference to Fig. 15. The processing of S31 to S33 in Fig. 15 is the same as the processing of S1 to S3 in Fig. 9, so the description will be omitted and the description will start from S34 in Fig. 15.
[0099] The simulation condition acquisition unit 19a acquires the simulation conditions of the dialogue (S34). The simulation conditions of the dialogue include, for example, the number of dialogue simulation patterns.
[0100] After S34, the utterance generation unit 16b generates dialogue histories (results of dialogue simulations between persons A and B) for the number of patterns specified in the dialogue simulation conditions (S35). Then, the utterance evaluation unit 17b evaluates each dialogue history generated in S35 (S36). Thereafter, the utterance evaluation unit 17b outputs each dialogue history and its evaluation value to the result display unit 18. Then, the result display unit 18 outputs each dialogue history and its evaluation value (S37).
[0101] By executing the above process, the information processing system can output multiple patterns of dialogue simulations and their evaluation values, allowing the user to check various possible dialogue patterns with the dialogue partner and learn what kind of dialogue with the dialogue partner will help them build a good relationship with each other.
[0102] [System Configuration, etc.] The components of each unit shown in the figure are conceptual functional units and do not necessarily have to be physically configured as shown. In other words, the specific form of distribution and integration of each device is not limited to that shown, and all or part of them can be functionally or physically distributed and integrated in any unit depending on various loads, usage conditions, etc. Furthermore, all or any part of the processing functions performed by each device can be realized by a CPU and a program executed by the CPU, or can be realized as hardware using wired logic.
[0103] Furthermore, among the processes described in the above embodiments, all or part of the processes described as being performed automatically can be performed manually, or all or part of the processes described as being performed manually can be performed automatically using a known method.In addition, the information including the processing procedures, control procedures, specific names, various data and parameters shown in the above documents and drawings can be changed as desired unless otherwise specified.
[0104] [Program] The information processing system described above can be implemented by installing a program (information processing program) as package software or online software on a desired computer. For example, by executing the program on an information processing device, the information processing device can function as an information processing system. The information processing device referred to here includes mobile communication terminals such as smartphones, mobile phones, and PHS (Personal Handyphone Systems), as well as terminals such as PDAs (Personal Digital Assistants).
[0105] 16 is a diagram showing an example of a computer that executes an information processing program. The computer 1000 includes, for example, a memory 1010 and a CPU 1020. The computer 1000 also includes a hard disk drive interface 1030, a disk drive interface 1040, a serial port interface 1050, a video adapter 1060, and a network interface 1070. These components are connected by a bus 1080.
[0106] The memory 1010 includes a read-only memory (ROM) 1011 and a random access memory (RAM) 1012. The ROM 1011 stores a boot program such as a basic input / output system (BIOS). The hard disk drive interface 1030 is connected to a hard disk drive 1090. The disk drive interface 1040 is connected to a disk drive 1100. A removable storage medium such as a magnetic disk or optical disk is inserted into the disk drive 1100. The serial port interface 1050 is connected to a mouse 1110 and a keyboard 1120, for example. The video adapter 1060 is connected to a display 1130, for example.
[0107] The hard disk drive 1090 stores, for example, an OS 1091, an application program 1092, a program module 1093, and program data 1094. That is, the programs that define the processes executed by the information processing system are implemented as program modules 1093 in which computer-executable code is written. The program modules 1093 are stored, for example, in the hard disk drive 1090. For example, the program modules 1093 for executing processes similar to those of the functional configuration in the information processing system are stored in the hard disk drive 1090. Note that the hard disk drive 1090 may be replaced with an SSD (Solid State Drive).
[0108] Data used in the processing of the above-described embodiment is stored as program data 1094, for example, in the memory 1010 or the hard disk drive 1090. The CPU 1020 then reads the program module 1093 or the program data 1094 stored in the memory 1010 or the hard disk drive 1090 into the RAM 1012 as necessary and executes them.
[0109] The program module 1093 and program data 1094 may not necessarily be stored in the hard disk drive 1090, but may also be stored in a removable storage medium and read by the CPU 1020 via the disk drive 1100 or the like. Alternatively, the program module 1093 and program data 1094 may be stored in another computer connected via a network (such as a local area network (LAN) or a wide area network (WAN)). The program module 1093 and program data 1094 may then be read by the CPU 1020 from the other computer via the network interface 1070.
[0110] REFERENCE SIGNS LIST 11 Personality acquisition unit 12 Search condition acquisition unit 13 Utterance acquisition unit 14 Personality DB 15 Dialogue partner search unit 16, 16a, 16b Utterance generation unit 17, 17a, 17b Utterance evaluation unit 18 Result display unit 19, 19a Simulation condition acquisition unit 20 Utterance selection unit
Claims
1. An information processing device comprising: a speech generation unit that generates speech candidates for a DT (Digital Twin) of a dialogue partner in a dialogue between a user and the DT of the dialogue partner; and a speech evaluation unit that calculates an evaluation value that quantifies the impression that the dialogue partner has of the user's speech candidates based on a dialogue history between the user and the DT of the dialogue partner, including the user's utterance candidates, and information indicating the individuality of the dialogue partner, and outputs the evaluation value for each of the user's utterance candidates.
2. The information processing device according to claim 1, characterized in that the utterance generation unit generates utterance candidates for the DT of the dialogue partner using a natural language processing model that generates utterance candidates for the DT of the dialogue partner in response to an utterance of the user based on information indicating the individuality of the dialogue partner and a dialogue history between the user and the DT of the dialogue partner.
3. The information processing device according to claim 1, further comprising: an utterance selection unit that selects utterance candidates constituting the dialogue between the user's DT and the dialogue partner's DT based on the evaluation value for each utterance candidate of the user's DT and the dialogue partner's DT based on the evaluation value for each utterance candidate of the user's DT and the dialogue partner's DT, and an output processing unit that outputs the dialogue between the user's DT and the dialogue partner's DT including the selected utterance candidate.
4. The information processing device according to claim 3, further comprising a simulated condition acquisition unit that acquires simulated conditions which are conditions for selecting utterance candidates that constitute a dialogue between the user's DT and the dialogue partner's DT, and the utterance selection unit further selects the utterance candidates based on the simulated conditions.
5. The information processing device according to claim 3, characterized in that the utterance generation unit creates a plurality of dialogue patterns between the DT of the user and the DT of the dialogue partner using utterance candidates for each of the DT of the user and the DT of the dialogue partner, the utterance evaluation unit calculates, for each of the dialogue patterns, an evaluation value that quantifies the impression that the user and the dialogue partner have of the dialogue pattern based on information indicating the individuality of the user and the dialogue partner, and the output processing unit further outputs, for each of the dialogue patterns, an evaluation value that quantifies the impression that the user and the dialogue partner have of the dialogue of the pattern.
6. An information processing method executed by an information processing device, comprising: a step of generating utterance candidates for a DT (Digital Twin) of a dialogue partner in a dialogue between a user and the DT of the dialogue partner; a step of calculating an evaluation value that quantifies the impression that the dialogue partner has of the user's utterance candidates based on a dialogue history between the user and the DT of the dialogue partner, including the user's utterance candidates, and information indicating the individuality of the dialogue partner, and outputting the evaluation value for each of the user's utterance candidates.
7. An information processing program for causing a computer to execute the steps of: generating speech candidates for a DT (Digital Twin) of a dialogue partner in a dialogue between a user and the DT of the dialogue partner; calculating an evaluation value that quantifies the impression that the dialogue partner has of the user's speech candidates based on a dialogue history between the user and the DT of the dialogue partner, including the user's speech candidates, and information indicating the individuality of the dialogue partner, and outputting the evaluation value for each of the user's speech candidates.
Citation Information
Patent Citations
Interactive business support system and interactive business support method
JP2021157419A
