Program, information processing method, and computer device

Abstract

This program causes a computer device to realize: a first determination function for determining, on the basis of user input, first prompt character information from character information set in advance as information pertaining to the personality and action guidelines of a character; a first generation function for generating a first prompt on the basis of the user input and the first prompt character information; a first acquisition function for acquiring, by using the first prompt, answer data pertaining to an answer of the character to the user input from a prescribed language model; and a first output function for outputting an answer to the user via the character on the basis of the answer data.

Description

Program, information processing method, and computer device

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

[0002] Recent advances in AI technology have led to the emergence of various language models with advanced natural language processing capabilities. In particular, large-scale language models, which have increased the amount of calculation, data, calculations, and parameters, have become capable of responding to questions and engaging in dialogue at a level that is less likely to cause discomfort to humans.

[0003] Many chatbots have appeared that use language models such as large-scale language models. Non-Patent Document 1 discloses that chatbots can be given character by editing the prompts sent to them. Specifically, it discloses that by setting the character's profile and behavioral rules in advance, responses can be obtained that are in line with the set content.

[0004] “Easy Guide to Creating Conversational AI: 2. Customize Your AI’s Personality with Instructions!” [online] [Retrieved March 25, 2024], Internet <URL: https: / / note.com / tomohoga / n / nfe233a3f7c19>

[0005] According to the technology in Non-Patent Document 1, it is possible to give a chatbot a certain degree of character. However, since the prompts for expressing the character are the same ones that are preset, it is difficult to increase the variety of response patterns, and there is room for improvement in terms of expressing the character.

[0006] An object of at least one embodiment of the present disclosure is to provide a new technique for enhancing the expressiveness of character in responses using a predetermined language model.

[0007] From a non-limiting perspective, a program according to one embodiment of the present disclosure is a program executed on a computer device, and causes the computer device to realize the following: a reception function for receiving user input from a user; a first determination function for determining character information for a first prompt from character information that is preset as information regarding the personality and guidelines of action of a specified character based on at least the user input; a first generation function for generating a first prompt based on the user input and the character information for the first prompt; a first acquisition function for using the first prompt to acquire answer data regarding the character's answer to the user input from a specified language model; and a first output function for outputting an answer to the user via the character based on the answer data.

[0008] From a non-limiting perspective, an information processing method according to one embodiment of the present disclosure is an information processing method including causing a computer device to execute the following steps: a receiving step of receiving user input from a user; a first determination step of determining character information for a first prompt from character information that is preset as information related to a character's personality and guidelines of action based on at least the user input; a first generation step of generating a first prompt based on the user input and the character information for the first prompt; a first acquisition step of acquiring answer data related to the character's answer to the user input from a predetermined language model using the first prompt; and a first output step of outputting an answer to the user via the character based on the answer data.

[0009] From a non-limiting perspective, a computer device according to one embodiment of the present disclosure is a computer device comprising: a receiving means for receiving user input from a user; a first determination means for determining character information for a first prompt from character information that is preset as information regarding a character's personality and guidelines for action based on at least the user input; a first generation means for generating a first prompt based on the user input and the character information for the first prompt; a first acquisition means for using the first prompt to acquire answer data regarding the character's answer to the user input from a predetermined language model; and a first output means for outputting an answer to the user via the character based on the answer data.

[0010] Each embodiment of the present application addresses one or more of the deficiencies.

[0011] 1 is a diagram showing an example of the configuration of a computer device corresponding to at least one embodiment of the present disclosure. FIG. 2 is a flowchart showing an example of an information processing method corresponding to at least one embodiment of the present disclosure. FIG. 3 is a diagram showing an example of the configuration of a computer device corresponding to at least one embodiment of the present disclosure. FIG. 4 is a flowchart showing an example of an information processing method corresponding to at least one embodiment of the present disclosure. FIG. 5 is a flowchart showing an example of a filler effect-related process corresponding to at least one embodiment of the present disclosure. FIG. 6 is a schematic diagram showing an example of an output screen corresponding to at least one embodiment of the present disclosure. FIG. 7 is a flowchart showing an example of a first output-related process corresponding to at least one embodiment of the present disclosure. FIG. 8 is a schematic diagram showing an example of an output screen corresponding to at least one embodiment of the present disclosure. FIG. 9 is a diagram showing an example of the configuration of a computer device corresponding to at least one embodiment of the present disclosure. FIG. 10 is a flowchart showing an example of an information processing method corresponding to at least one embodiment of the present disclosure.

[0012] Examples of embodiments of the present disclosure will be described below with reference to the drawings. Note that the various components in the examples of the embodiments described below can be combined as appropriate to the extent that no inconsistencies or contradictions occur. Furthermore, the content described in one embodiment (e.g., components with the same reference numerals or names as components already described) may be omitted in other embodiments. Furthermore, the content of operations and processes unrelated to the characteristic features of each embodiment may be omitted. Furthermore, the order of the various processes constituting the various flows described below may be random and may be executed in parallel to the extent that no inconsistencies or contradictions occur in the process content. Note that in this specification, the terms "first" and "second" are used merely for distinction and do not indicate the order of processes or the like.

[0013] First Embodiment First, a first embodiment of the present disclosure will be described. As the first embodiment, a program executed on a computer device will be exemplified below.

[0014] FIG. 1 is a diagram illustrating an example of the configuration of a computer device 1 corresponding to at least one embodiment of the present disclosure. The computer device 1 and a server device 5 can be connected to each other via a network 4, for example, wirelessly or via a wired connection. The wireless communication method is not particularly limited, and any conventionally known technology, such as a wireless LAN (Local Area Network) connection such as Wi-Fi, a Bluetooth (registered trademark) connection, or an infrared connection, can be used as appropriate. The wired communication method is not particularly limited, and any conventionally known technology, such as a USB (Universal Serial Bus) connection, can be used as appropriate. The communication may be short-range communication or long-range communication via a communication network such as the Internet. The communication network may be a so-called blockchain network.

[0015] The computer device 1 includes at least a storage unit 10 and a control unit 20. The storage unit 10 is configured by a storage device such as a read-only memory (ROM), a hard disk drive (HDD), or a solid state drive (SDD). The storage unit 10 stores various programs and data for operating the computer device 1. The storage unit 10 may include an external storage device accessible by the computer device 1.

[0016] The storage unit 210 can store, for example, a program 11 and character information 12. The program 11 is, for example, a computer program for causing the computer device 1 to realize various functions described below. The character information 12 is, for example, information set in advance as information relating to the personality and guidelines of a specific character.

[0017] The control unit 20 is configured with, for example, a processor and a memory such as a RAM (Random Access Memory). By executing the program 11, the control unit 20 of the computer device 1 functions as, for example, a receiving unit 21, a first determining unit 22, a first generating unit 23, a first acquiring unit 24, and a first output unit 25.

[0018] The receiving unit 21 has a function of receiving a user input from a user. The first determining unit 22 has a function of determining character information for a first prompt from character information 12 that is preset as information related to the personality and guidelines of a predetermined character, based on at least the user input.

[0019] The first generating unit 23 has a function of generating a first prompt based on the user input received by the receiving unit 21 and the character information for the first prompt determined by the first determining unit 22. The first acquiring unit 24 has a function of acquiring, from a predetermined language model 51, response data regarding the response of the character to the user input, using the first prompt generated by the first generating unit 23.

[0020] The first output unit 25 has a function of outputting an answer to the user via the character based on the answer data acquired by the first acquisition unit 24 .

[0021] The server device 5 includes at least a predetermined language model 51. The server device 5 uses the language model 51 to create various response data in response to various prompts sent from the computer device 1. The created various response data are transmitted to the computer device 1 as appropriate. The server device 5 may be configured by multiple computers.

[0022] The predetermined language model 51 is not particularly limited as long as it can generate response data for various prompts received from the computer device 1, but is preferably, for example, a large-scale language model. The language model 51 may be provided in the computer device 1.

[0023] Next, a description will be given of an information processing method using the computer device 1 according to the first embodiment. Fig. 2 is a flowchart showing an example of the information processing method according to at least one embodiment of the present disclosure.

[0024] In step S1, the computer device 1 accepts a user input from a user. In step S2, the computer device 1 determines character information for a first prompt from character information 12 that is preset as information related to the personality and guidelines of a predetermined character, based on at least the user input accepted in step S1 (first determination).

[0025] In step S3, the computer device 1 generates a first prompt based on the user input received in step S1 and the first prompt character information determined in step S2 (first generation).

[0026] In step S4, the computer device 1 uses the first prompt generated in step S3 (for example, by sending the first prompt to the server device 5) to acquire answer data regarding the character's answer to the user input from the predetermined language model 51 (first acquisition). In step S5, the computer device 1 outputs an answer to the user via the character based on the answer data acquired in step S4 (first output), and then ends the process.

[0027] As one aspect of the first embodiment, a new technique can be provided for enhancing the expressiveness of a character in a response using a predetermined language model. Specifically, instead of always generating the same prompt using preset character information, information to be used for prompt generation is determined from preset character information based on user input, etc., making it possible to dynamically generate prompts. As a result, for example, while having the advantage of being able to respond to any user input, it is possible to provide a response that is appropriate for the situation and is characteristic of the character.

[0028] In the first embodiment, the term "computer device" is not particularly limited, but examples thereof include a stationary game console, a portable game console, a wearable terminal, a desktop or notebook personal computer, a tablet computer, a portable terminal such as a PDA (Personal Digital Assistant) or a smartphone, or a server computer.

[0029] Furthermore, in the first embodiment, "user input" is not particularly limited, but examples thereof include voice input by the user's speech and text input using a keyboard, controller, etc. In addition to or instead of these, input may be made by user actions such as gestures and facial expressions. Note that if the "computer device" is a server-type computer, the user input is temporarily accepted by a terminal or the like held by the user and transmitted to the "computer device," whereby the user input is accepted by the "computer device."

[0030] Furthermore, in the first embodiment, the "predetermined character" is not particularly limited, but may be, for example, a fictional character or a character based on a real person, animal, plant, etc.

[0031] In the first embodiment, "character information" includes information about the individuality and guidelines of the character. The "character information" preferably includes at least one, and preferably multiple, of the following: the character's personality, profile, hobbies, preferences, speaking style, current goal, potential topics, prohibited topics, what to do next, memories of experiences, knowledge, guidelines of action, and conditions for changing the guidelines of action.

[0032] In the first embodiment, the "first prompt" refers to a prompt that causes the language model to generate a response to a user input. The "first prompt" is dynamically generated, for example, in response to receipt of a user input. The "first prompt" may include, for example, the content of the user input as well as a portion of the "character information" that is selected based on the content of the user input.

[0033] Furthermore, in the first embodiment, "outputting an answer to the user via a character" is not particularly limited, but may include, for example, displaying text in a chat format together with an image such as an icon representing the character, and / or outputting the answer as the voice of the character using voice synthesis, etc. Furthermore, a motion image of the character may be output together with these outputs.

[0034] Second Embodiment Next, a second embodiment will be described below, in which a program executed by a computer device will be exemplified as the second embodiment.

[0035] 3 is a diagram illustrating an example of the configuration of a computer device 101 corresponding to at least one embodiment of the present disclosure. The computer device 101 and the server device 105 can be connected to each other for communication via a network 4, for example, wirelessly or via a wired connection.

[0036] The computer device 101 includes at least a storage unit 110 and a control unit 120. The storage unit 110 can store, for example, a program 111, character information 112, history information 113, user information 114, filler effect information 115, audio information 116, and model information 117.

[0037] The program 111 is, for example, a computer program for causing the computer device 101 to realize various functions described below.

[0038] The character information 112 is, for example, information set in advance as information relating to the individuality and guidelines of a specific character. The "character information" includes, for example, the character's personality, profile, hobbies, preferences, speaking style, current goal, potential topics, prohibited topics, what to do next, memories of experiences (such as an action timetable listing what the character was doing before starting a conversation), knowledge, the character's guidelines of action, and conditions for changing the guidelines of action, and at least one of each is set, preferably multiple.

[0039] The history information 113 includes, for example, information about the history of a conversation between the user and a character. Specifically, the history information 113 may include the content of a predetermined amount of the most recent conversation, a summary of the conversation so far, and information input by the user that has been preset or determined to be important (for example, the user's name, hobbies, the user's emotions determined from the user's input, etc.). The history information 113 may also include the progress of the conversation between the user and the character.

[0040] The user information 114 is, for example, information about a user, other than the content of a conversation between the user and a character. The program according to this embodiment can be applied to, for example, chatbots used in various services, corporate activities, and the like, as well as NPCs (Non-Player Characters) placed in game spaces and other virtual spaces. When applied to these, the user information 114 may be, for example, usage history of various services and information about users registered therein, information about a user character operated by a user in a game space or other virtual space, or information about the progress of a game, etc.

[0041] The filler effect information 115 is information related to the filler effect output by the character. The filler effect information 115 may include, for example, information related to the output of text information, motion information, etc. output by the character in the filler effect, and information related to the decision conditions for determining which of multiple filler effect candidates to decide on (which to select).

[0042] The voice information 116 includes, for example, information related to the voice uttered by the character. For example, text information included in the filler effect information 115 and answer data generated by the language model 151 are output as the voice of the character by performing voice synthesis or the like with reference to the voice information 116.

[0043] The model information 117 includes, for example, information regarding the modeling and motion of a character. The model information 117 may also include, for example, information for generating a user character, a game space or other virtual spaces, virtual objects to be placed in these spaces, and the like.

[0044] When the program 111 is executed, the control unit 120 of the computer device 101 functions, for example, as a reception unit 121, a first determination unit 122, a first generation unit 123, a first acquisition unit 124, a conversion unit 125, a first output unit 126, a second determination unit 127, a second generation unit 128, a second acquisition unit 129, and a second output unit 130.

[0045] The reception unit 121 has a function of receiving user input from a user. The user input may be, for example, text input or voice input. If the user input is voice input, it may be converted into text using, for example, voice recognition technology. The reception unit 121 may also receive an image of the user as the user input.

[0046] The first determination unit 122 has a function of determining character information for a first prompt from character information 112 that is preset as information relating to the personality and guidelines of action of a predetermined character, based at least on the above-mentioned user input.

[0047] In addition, when the user input includes a topic related to the hobbies and / or preferences of the character set in the character information 112, the first determination unit 122 can include instructions in the character information for the first prompt to generate another topic related to the hobbies and / or preferences.

[0048] Furthermore, when the user input includes a topic that corresponds to a prohibited topic (a topic that is set as being prohibited for the character to talk about) set in the character information 112, the first determination unit 122 can include, in the character information for the first prompt, an instruction to generate a topic different from the prohibited topic. Note that there are no particular limitations on the method for determining whether the user input includes hobbies, preferences, prohibited topics, or topics related to these, and various natural language processing methods such as Word2vec can be used, for example.

[0049] Furthermore, the first determination unit 122 can further determine first prompt character information from the character information 112 based on the character's predetermined action guideline and the change condition for the action guideline.

[0050] The first generating unit 123 has a function of generating a first prompt based on the user input received by the receiving unit 121 and the first prompt character information determined by the first determining unit 122 .

[0051] The first generation unit 123 may also generate the first prompt based on a history of user inputs and character responses. Specifically, the first generation unit 123 may extract information stored in the history information 113, such as the content of a predetermined amount of the most recent conversation between the user and the character, a summary of the conversation, and the like, and use this information to generate the first prompt.

[0052] Furthermore, the first generation unit 123 can generate a first prompt including content instructing the language model 151 to indicate a point at which a character should express a predetermined emotion in the text of the answer data generated by the language model 151. Furthermore, when a point at which an emotion should be expressed is present, the first generation unit 123 can generate a first prompt including content instructing the language model 151 to indicate the beginning of the sentence in which the point is present, in addition to the point at which the emotion is present. Note that the predetermined emotion may be, for example, joy, anger, sadness, or happiness, but is not limited to these, and may be set appropriately by the program administrator depending on the application mode of the program according to this embodiment.

[0053] First acquisition unit 124 has a function of acquiring answer data regarding a character's answer to a user input from language model 151, using the first prompt generated by first generation unit 123. First acquisition unit 124, for example, transmits the first prompt to server device 105 and receives the answer data generated by language model 151 from server device 105.

[0054] The conversion unit 125 has a function of converting text included in the answer data acquired by the first acquisition unit 124 into a phrase appropriate for the character. The conversion unit 125 may perform the above conversion using, for example, a trained model that has learned the character's tone of voice through machine learning. Note that, without providing the conversion unit 125, the character's tone of voice may be expressed by setting information related to the character's tone of voice in the character information 112 and including an instruction related to the tone of voice in the first prompt.

[0055] The first output unit 126 has a function of outputting an answer to the user via a character based on the answer data acquired by the first acquisition unit 124. The first output unit 126 outputs an answer based on, for example, text converted by the conversion unit 125.

[0056] It is also preferable that the output by the first output unit 126 includes the voice and / or motion images of the character. In this case, it is preferable that the first output unit 126 adds tag information to the text data, which tag information indicates the tone of voice, intonation, pauses, etc. when outputting the text aloud, and the actions and expressions, etc. of the character when outputting the motion images, based on conditions preset by the program administrator, for example, and outputs the text data in accordance with the tag information.

[0057] Furthermore, when the user input includes a topic related to the character's hobbies and / or preferences, the first output unit 126 preferably outputs a voice and / or the motion image differently from when the topic is not included. Specifically, when the user input includes a topic related to a favorite hobby or preference, the first output unit 126 preferably outputs a voice and / or a facial expression that sounds happier than usual. When the user input includes a topic related to a disliked hobby or preference, the first output unit 126 preferably outputs a voice and a facial expression that sounds displeased.

[0058] Furthermore, when the answer data indicates the above-mentioned emotional expression point, the first output unit 126 preferably outputs the answer while having the character express the emotional expression by voice and / or motion images at least at the emotional expression point. Furthermore, when the answer data indicates the beginning of a sentence in which an emotional expression point exists, the first output unit 126 preferably outputs the answer while having the character express the emotional expression by voice and / or motion images at least from the beginning of the sentence to the emotional expression point. Specifically, it is preferable to control the change of emotion gradually from the beginning of the sentence.

[0059] Furthermore, when an expression point is indicated in the response data, the first output unit 126 preferably terminates the expression related to the expression of emotion in response to a predetermined termination condition being satisfied after the emotion is expressed at the expression point. The predetermined termination condition may be set appropriately by the program administrator depending on the application mode of the program according to this embodiment, and may be, for example, the passage of a predetermined time or the appearance of the next expression point.

[0060] The second determination unit 127 has a function of determining a filler effect based on the user input received by the reception unit 121. Preferably, the second determination unit 127 analyzes the content of the user input and determines the filler effect based on the analysis result by referring to the filler effect information 115. This analysis may include, for example, an analysis of whether the user input includes a question or not, or an analysis of the user's emotions based on the text, voice, or facial expression entered by the user. Furthermore, in addition to or instead of the above analysis, the second determination unit 127 may determine the filler effect based on determination data acquired by the second acquisition unit 129 described below.

[0061] The second generation unit 128 has a function of generating a second prompt including an instruction sentence for instructing the second generation unit 128 to determine the content of the user input. There are no particular limitations on the type of determination that can be made, but examples of the determination that can be made include whether the user input is positive or negative, or whether the user input is different from the previous topic.

[0062] Second acquisition unit 129 has a function of using the second prompt to acquire determination data regarding the result of the above determination from language model 151. Second acquisition unit 129, for example, transmits the second prompt to server device 105 and receives answer data generated by language model 151 from server device 105.

[0063] The second output unit 130 has a function of outputting, via a character, a filler effect determined by the second determination unit 127 during the period from when the receiving unit 121 receives a user input until when the first output unit 126 outputs an answer. The output by the second output unit 130 preferably includes the voice and / or motion image of the character.

[0064] The server device 105 includes at least a predetermined language model 151. The server device 105 uses the language model 151 to create various types of response data in response to various prompts sent from the computer device 101. The language model 151 is not particularly limited as long as it can create response data in response to various prompts received from the computer device 101, but it is preferable that the language model 151 be, for example, a large-scale language model.

[0065] Next, an information processing method using the computer device 101 according to the second embodiment will be described. Fig. 4 is a flowchart showing an example of the information processing method corresponding to at least one of the embodiments of the present disclosure.

[0066] In step S10, the computer device 101 accepts user input from the user. In step S20, the computer device 101 executes filler effect-related processing. In step S30, the computer device 101 executes first output-related processing. The processes from step S10 to step S30 are repeated until a conversation end condition is met, such as the user performing an end operation.

[0067] 4 shows an overview of the information processing method in this embodiment. The first output-related process in step S30 takes several seconds for the language model 151 to generate answer data, which means the user has to wait a certain amount of time before the computer device 101 outputs an answer to the user input. Therefore, in this embodiment, filler effect-related process in step S20 is performed during this time, thereby reducing the impression that a wait is occurring. The filler effect-related process and the first output-related process are initiated in response to the acceptance of a user input in step S10, and some of the processes may be executed in parallel.

[0068] The filler effect-related processing in step S20 will now be described in detail with reference to Figures 5 and 6. Figure 5 is a flowchart showing an example of the filler effect-related processing corresponding to at least one of the embodiments of the present disclosure.

[0069] In step S21, the computer device 101 generates a second prompt and transmits it to the server device 105. The second prompt includes, for example, an instruction to determine whether the content of the user input is positive or negative, or whether the content of the user input differs from the immediately preceding topic. The instruction in the second prompt may be an instruction for binary classification as described above, or may be an instruction for multi-valued classification. The instruction may also be an instruction to output the proportions of various parameters (e.g., joy, anger, sadness, happiness, etc.) in the content of the user input, or to extract keywords from the content of the user input. The extracted keywords can be used when repeating what is said (e.g., repeating the extracted keywords, not repeating certain keywords, etc.).

[0070] In step S22, server device 105 creates answer data to the second prompt using language model 151. Specifically, language model 151 performs the above-described binary classification of the content of the user input and transmits the determination result to computer device 101. In step S23, computer device 101 obtains an answer to the second prompt from server device 105.

[0071] In step S24, the computer device 101 analyzes the content of the user input. The process of step S24 is preferably executed in parallel with the processes of steps S21 to S23. Alternatively, the computer device 101 may be configured to execute only one of the processes of step S24 and steps S21 to S23.

[0072] In step S24, for example, an analysis may be performed to determine whether the user input includes a question, or to analyze the user's emotions based on the text, voice, or facial expression entered by the user. These analyses may be performed using, for example, conventionally known text analysis or voice analysis techniques, or trained models that have undergone machine learning to enable these analyses.

[0073] In step S25, the computer device 101 determines a filler effect based on the answer data acquired in step S23 and the analysis results in step S25. The filler effect information 115 stores, for example, information about multiple types of filler effects and selection conditions for selecting each effect. The computer device 101 determines which selection conditions the results of the above judgment and analysis satisfy, and then determines the filler effect.

[0074] For example, selectable filler effects are limited depending on whether the user input is positive or negative, and also depending on whether the content of the user input differs from the immediately preceding topic. Similarly, selectable filler effects may be limited depending on whether the user input includes a question or the type and intensity of the user's emotion obtained as an analysis result. If there are multiple effects that ultimately become selectable, they may be selected randomly. Furthermore, priorities may be assigned to the results of the judgment and analysis, and the filler effect may be determined based on the result with the highest priority.

[0075] In step S26, the computer device 101 outputs the filler effect determined in step S25 via the character. Specifically, the filler effect is output by displaying text on the display screen, outputting audio generated for the character, and motion images representing the character's movements. The output of step S26 is executed between accepting the user input in step S10 and the first output in step S41, which will be described later.

[0076] 6 is a schematic diagram illustrating an example of an output screen corresponding to at least one embodiment of the present disclosure. Specifically, FIG. 6 is an example of an output screen in step S26.

[0077] 6A, a character image 162, a user icon 163, a character icon 164, and areas 165a and 165b are displayed on an output screen 161. Area 165a is displayed in association with the user icon 163, and displays the user's utterance (user input content) in text. Area 165b is displayed in association with the character icon 164, and displays the character's utterance (output content in step S26) in text.

[0078] 6A, the content of the user input includes a question, as shown in area 165a. If such an analysis result is obtained in step S24, a filler effect may be selected, in which at least a part of the user input is output as text and / or voice via a character, as shown in area 165b. In addition, the character's tone of voice, facial expression, movements, etc. may be changed to suggest that the character is thinking about the answer.

[0079] Note that the filler effect shown in Fig. 6(a) in which at least a part of the user input is output as text and / or audio, or simply, an effect that repeats part of the user input, may be selected not only in the case shown in Fig. 6(a) but also in other cases. For example, it may be selected in the case shown in Fig. 6(b) described below, or may be an effect that can always be selected randomly regardless of the above-mentioned determination results and analysis results.

[0080] In the case of (b) in FIG. 6, the content of the user input differs from that in the case of (a) in FIG. 6. In the case of (b) in FIG. 6, the language model 151 determines that the content of the user input displayed in the text area 165b is positive. In this case, for example, one piece of text is randomly selected from texts set to be selected in positive cases, and text and audio are output based on that text. The character's tone of voice, facial expression, movements, and the like may also be designed to give a positive impression. Regarding facial expressions, for example, a smiling face may be selected, as in the character image 162 in (b) in FIG. 6.

[0081] Next, the first output-related process in step S30 will be described in detail with reference to Figures 7 and 8. Figure 7 is a flowchart showing an example of the first output-related process corresponding to at least one of the embodiments of the present disclosure.

[0082] In step S31, the computer device 101 determines first prompt character information based on at least the content of the user input. Specifically, based on the content of the user input, the computer device 101 selects first prompt character information from the character information 112. For some items included in the character information 112, some or all of them may be randomly selected and always included in the first prompt character information.

[0083] In step S31, the first prompt character information may be selected based on the content of the user input as well as information stored in the history information 113, information stored in the user information 114, etc.

[0084] In step S32, the computer device 101 determines whether the topic input by the user is related to the character's hobbies or preferences. If it is determined to be a related topic (Yes in step S32), in step S33, the computer device 101 includes information related to the hobbies or preferences from the character information 112 in the character information for the first prompt. If it is determined not to be a related topic (No in step S32), for example, the process proceeds to step S34. Note that, with regard to the information related to the hobbies or preferences, some or all of the information selected randomly may always be included in the character information for the first prompt.

[0085] In step S34, the computer device 101 determines whether the topic input by the user includes a prohibited topic for the character. If it is determined that the prohibited topic is included (Yes in step S34), in step S35, the computer device 101 includes information related to the prohibited topic from the character information 112 in the character information for the first prompt. If it is determined that the prohibited topic is not a related topic (No in step S34), for example, the computer device 101 proceeds to step S36. Note that, with regard to the information related to the prohibited topic, some or all of the information selected randomly may always be included in the character information for the first prompt. Furthermore, the order in which the processes of steps S31 to S35 are performed is not limited to the order described above.

[0086] In step S36, the computer device 101 generates a first prompt based on the content of the user input in step S10 and the first prompt character information determined in steps S31 to S35. The first prompt may further include information stored in the history information 113, information stored in the user information 114, etc. Table 1 below shows a specific example of the configuration of the first prompt. Note that each item in Table 1 below may include multiple contents as long as there is no unattended area.

[0087]

[0088] "What to do next" in Table 1 can be changed, for example, according to the content of the user input or a guideline of action selected in accordance with the change conditions for the guideline of action stored in character information 112. Furthermore, with regard to "information related to the previous topic" in Table 1, for example, if the most recent user input is about noodles and knowledge or experience about noodles is stored in character information 112, this fact will be included in the first prompt, and as a result, the answer generated by language model 151 may include content that talks about that knowledge or experience.

[0089] Furthermore, by including a summary of the conversation up to that point in the first prompt, such as "Summary of the most recent topic" in Table 1, it becomes possible to include the content of the past conversation in the answer generated by the language model 151. For example, when a user asks, "What were we talking about just now?", it is possible to generate an answer that reflects a past memory.

[0090] Returning to the description of Fig. 7, the first prompt generated in step S36 is transmitted to server device 105. In step S37, server device 105 generates answer data to the first prompt using language model 151. Server device 105 transmits the answer data to the first prompt to computer device 101.

[0091] In step S38, the computer device 101 acquires answer data for the first prompt from the server device 105. Note that, in order to facilitate processing related to tone conversion, addition of effect tags, and output via a character, which will be described later, it is preferable to acquire the answer data in a format that can be directly used by the program 111 (for example, JSON format, etc.). The format to be used is determined depending on the program 111. Furthermore, obtaining answer data in such a format can be made possible, for example, by including content instructing that effect in the first prompt.

[0092] In step S39, the computer device 101 converts the text indicated by the answer data acquired in step S38 into a phrase (tone of voice) that suits the character.

[0093] In step S40, the computer device 101 adds various effect tags to the text data converted in step S39. Table 2 below shows a specific example of the processing from step S37 to step S40.

[0094]

[0095] In Table 2 above, in step S40, performance tags are added in a format that can be directly used by program 111. In addition to those shown in Table 2 above, it is also preferable to add performance tags that instruct the tone of voice, intonation, pauses, etc. when outputting text aloud, and the movements and expressions of characters when outputting motion images.

[0096] Returning to the explanation of Fig. 7, in step S41, the computer device 101 performs output (first output) via the character based on the data to which the effect tag was added in step S40. Specifically, output based on the answer generated by the language model 151 is performed by displaying text on the display screen, outputting voice generated for the character, and using motion images representing the character's movements.

[0097] In step S42, the computer device 101 updates the information included in the character information 112, history information 113, user information 114, etc. as necessary based on the results of the series of processes performed up to this point. This updating may be performed appropriately when data to be updated is generated.

[0098] 8 is a schematic diagram illustrating an example of an output screen corresponding to at least one embodiment of the present disclosure. Specifically, FIG. 8 is an example of an output screen in step S41.

[0099] 8A shows an example in which the character information 112 specifies watching F1 races and baseball games as the character's hobbies. In FIG. 8A, as shown in area 166a, the user input includes the topic "I went to see cars." In this case, for example, in step S32, it is determined that a topic related to the above-mentioned hobby is included, and information related to the above-mentioned hobby is included in the first prompt. As a result, the character's response shown in area 166b includes a topic related to F1. Similarly, in area 166d, the topic of baseball is raised as a result of determining that the topics "I went to see cars" and "F1" are related to the topic "watching baseball."

[0100] Note that "A car..." in area 166b and "Oh, I see" in area 166d are texts output as a result of filler effect-related processing. Also, because the conversation is about a topic the character likes, the character's tone of voice, facial expression, movements, etc. can be designed to give the impression of fun. Regarding facial expressions, for example, a smiling face can be selected, as in character image 162 in FIG. 8(a).

[0101] FIG. 8B shows an example in which weapon crafting is set as a prohibited topic in the character information 112, and herbal cultivation is set as the character's favorite hobby. In FIG. 8B, as shown in area 166e, the user input includes the topic "how to craft weapons." In this case, for example, in step S34, it is determined that a topic related to the prohibited topic is included, and information related to this prohibited topic is included in the first prompt. As a result, the character's response shown in area 166f rejects the topic of weapon crafting. Furthermore, for example, because "how to craft weapons" and "how to craft herbs" are related, "how to craft herbs" is presented as a topic in area 166f. The "Um, well..." in area 166f is text output as a result of the filler effect-related processing.

[0102] As one aspect of the second embodiment, a new technique can be provided for enhancing the expressiveness of a character in a response using a predetermined language model. Specifically, instead of always generating the same prompt using preset character information, the information to be used for prompt generation is determined from preset character information based on user input, etc., making it possible to dynamically generate prompts. As a result, for example, while having the advantage of being able to respond to any user input, it is possible to provide a response that is appropriate for the situation and is characteristic of the character.

[0103] As one aspect of the second embodiment, the first output may include voice and / or motion images of the character, which may enhance the expressiveness of the character and improve the user experience, for example.

[0104] As one aspect of the second embodiment, the character information may include information about the character's hobbies and / or preferences, thereby enhancing the expressiveness of the character, for example.

[0105] As one aspect of the second embodiment, when a user input includes a topic related to the character's hobbies and / or preferences, the character information for the first prompt may include instructions for generating another topic related to the hobbies and / or preferences, thereby enhancing the character's expressiveness. Specifically, humans sometimes shift the topic from the topic at hand to something associated with the current topic, and this makes it possible to reproduce such human behavior with a character. In addition, by using different voice and / or motion images for the character than when the topic is not included, the character's expressiveness can be further enhanced.

[0106] As one aspect of the second embodiment, the character information includes information regarding prohibited topics that the character is prohibited from talking about, and if the user input includes a topic that falls under the prohibited topic, the character information for the first prompt can include instructions for generating a different topic that is different from the prohibited topic, thereby, for example, preventing the character from having a conversation that is inconsistent with its character, and reducing the risk of damaging the character's image, etc.

[0107] As one aspect of the second embodiment, by generating a first prompt based on a history of user input and responses, it becomes possible for the character to reproduce human behavior, such as responding based on recognition of what has already been said or reflecting on past topics.

[0108] As one aspect of the second embodiment, by determining character information for the first prompt from character information based on pre-set conditions for changing the guidelines of action, it is possible, for example, to make the character's answers and behavioral changes closer to those of a human, or to give the changes a story-like quality.

[0109] As one aspect of the second embodiment, by converting the text included in the response data generated by the language model into wording appropriate for the character and then outputting it, it is possible to enhance the expressiveness of the character, for example.

[0110] In one aspect of the second embodiment, a first prompt is generated that includes an instruction to indicate a point in the text of the answer data where a character should express a predetermined emotion, and an answer is output while the character expresses the emotion at least at that point through audio and / or motion images, thereby making it easy to achieve, for example, a rich and natural expression of emotion by the character. Furthermore, by generating a first prompt that includes an instruction to indicate not only that point but also the beginning of the sentence in which that point exists, it is possible to, for example, gradually change the emotion from the beginning of the sentence to that point, thereby reducing the risk of the user feeling uncomfortable due to an abrupt change in emotion at that point.

[0111] As one aspect of the second embodiment, by terminating the expression related to the expression of emotion in response to a predetermined termination condition being met after the emotion is expressed at the expression point, it is possible to reduce the risk of the user feeling uncomfortable due to a sudden change in emotion, for example by allowing the intensity of the emotion to naturally diminish.

[0112] As one aspect of the second embodiment, by determining a filler effect based on a user input and outputting the filler effect via a character between the reception of the user input and the output of a response to that input, a new technique can be provided for enhancing the expressiveness of a character, for example, in responses using a predetermined language model. Specifically, it becomes easy to realize natural expressions that make the character appear as if they are actually thinking of a response in response to the user input. This also reduces the impression that the user may have of a long wait time until the response is output.

[0113] As one aspect of the second embodiment, by analyzing whether a user input includes a question or not and determining the filler presentation based on the analysis results, it becomes possible to output, for example, a more natural filler presentation.

[0114] As one aspect of the second embodiment, by determining the filler effect based on the results of analyzing the user's emotions based on the text or voice input by the user, it becomes easier to select a filler effect that is appropriate for the atmosphere of the situation, for example.

[0115] In one aspect of the second embodiment, a second prompt including an instruction sentence for determining the content input by the user is generated, determination data relating to the result of the determination is acquired from a language model, and a filler effect is determined based on the determination data. This makes it easy to select a filler effect that is appropriate for the atmosphere of the situation. Furthermore, by having the language model perform the determination, the processing load on the computer device that outputs to the user can be reduced.

[0116] As one aspect of the second embodiment, the above determination includes determining whether the user input is positive or negative, and / or determining whether the user input is different from the previous topic, which makes it easier to select a filler effect that is appropriate for the atmosphere of the situation, for example.

[0117] As one aspect of the second embodiment, by providing a filler effect in which at least a portion of a user input is output as text and / or audio via a character, it becomes possible to reproduce human behavior such as, for example, repeating what is being said in a conversation by a character.

[0118] In the second embodiment, the contents described in the first embodiment can be adopted to the extent necessary for the "computer device," "user input," "specified character," "character information," "first prompt," and "outputting an answer to the user via a character."

[0119] Furthermore, in the second embodiment, the "conditions for changing a guideline of action" define the conditions under which a character's guideline of action changes. The "conditions for changing a guideline of action" may be set as appropriate by a program administrator according to the application mode of the program according to this embodiment, the personality of the character, and the like. Examples of the "conditions for changing a guideline of action" include: fulfilling a predetermined condition, such as a specific topic coming up in a conversation between a user and a character; the conversation lasting longer than a predetermined time; a predetermined emotion obtained by analyzing the user's emotion or an emotion set for a character reaching a certain intensity or higher; or the user's game progress exceeding a predetermined level, and other conditions that information stored as user information fulfills a predetermined condition.

[0120] In the second embodiment, the "second prompt" refers to a prompt that causes the language model to determine the content input by the user based on a predetermined criterion, such as binary classification or multi-value classification.

[0121] In addition, in the second embodiment, a "filler effect" is an effect that is used, for example, to fill the time it takes for a language model to generate answer data between the time a user input is received and the time an answer to that input is output.

[0122] Third Embodiment Next, a third embodiment will be described with reference to Figures 9 and 10. In the following, a program executed by a computer device will be exemplified as the third embodiment.

[0123] 9 is a diagram illustrating an example of the configuration of a computer device 201 corresponding to at least one embodiment of the present disclosure. The computer device 201 and a server device 205 can be connected to each other for communication via a network 4, for example, wirelessly or via a wired connection.

[0124] The computer device 201 includes at least a storage unit 210 and a control unit 220. The storage unit 210 can store, for example, a program 211 and character information 212. When the program 211 is executed, the control unit 220 of the computer device 201 functions as, for example, a reception unit 221, a first generation unit 222, a first acquisition unit 223, and a first output unit 224.

[0125] The reception unit 221 has a function of receiving user input from the user.

[0126] The first generating unit 222 has a function of generating a first prompt based on at least the user input received by the receiving unit 221. Furthermore, the first generating unit 222 can generate a first prompt including an instruction sentence that instructs the character to indicate a point at which the character should express a predetermined emotion in the text of answer data, which will be described later.

[0127] The first acquisition unit 223 has a function of acquiring, from a predetermined language model 251, response data relating to the character's response to the user input, using the first prompt generated by the first generation unit 222.

[0128] The first output unit 224 has a function of outputting an answer to the user via the character based on the answer data acquired by the first acquisition unit 223. The output includes outputting audio and / or motion images of the character. Furthermore, if the expression point is indicated in the answer data acquired by the first acquisition unit 223, the first output unit 224 can output an answer while having the character express an emotional expression by audio and / or motion images at least at the expression point.

[0129] The server device 205 includes at least a predetermined language model 251. The server device 205 uses the language model 251 to create various types of response data in response to various prompts sent from the computer device 201. The language model 251 is not particularly limited as long as it can create response data in response to various prompts received from the computer device 201, but it is preferable that the language model 251 be, for example, a large-scale language model.

[0130] Next, an information processing method using the computer device 201 according to the third embodiment will be described. Fig. 10 is a flowchart showing an example of the information processing method corresponding to at least one of the embodiments of the present disclosure.

[0131] In step S51, the computer device 201 accepts a user input from the user.

[0132] In step S52, the computer device 201 generates a first prompt (first generation) based on at least the user input received in step S51. In step S52, for example, the computer device 201 can generate a first prompt including an instruction sentence instructing the character to indicate a point in the text of the answer data generated by the language model 251 where the character should express a predetermined emotion.

[0133] In step S53, the computer device 201 uses the first prompt generated in step S52 to acquire response data regarding the character's response to the user input from the language model 251 (first acquisition).

[0134] In step S54, the computer device 201 outputs an answer to the user via the character based on the answer data acquired in step S53 (first output), and then ends the process. Here, if the answer data acquired in S53 indicates the above-mentioned expression point, in step S54, the answer is output while the character expresses the expression of emotion by voice and / or motion image at least at the above-mentioned expression point.

[0135] As one aspect of the third embodiment, a new method for enhancing the expressiveness of characters in responses using a predetermined language model can be provided. Specifically, it becomes easier to realize rich and natural emotional expression by characters.

[0136] In the third embodiment, the "computer device," "user input," "predetermined character," "character information," "first prompt," and "outputting an answer to the user via a character" may each adopt the contents described in the first or second embodiment to the extent necessary.

[0137] Fourth Embodiment Next, a fourth embodiment will be described with reference to Figures 11 and 12. In the following, a program executed by a computer device will be exemplified as the fourth embodiment.

[0138] 11 is a diagram illustrating an example of the configuration of a computer device 301 corresponding to at least one embodiment of the present disclosure. The computer device 301 and a server device 305 can be connected to each other for communication via a network 4, for example, wirelessly or via a wired connection.

[0139] The computer device 301 includes at least a storage unit 310 and a control unit 320. The storage unit 310 can store, for example, a program 311 and character information 312. When the program 311 is executed, the control unit 320 of the computer device 301 functions as, for example, a reception unit 321, a first generation unit 322, a first acquisition unit 323, a first output unit 324, a second determination unit 325, and a second output unit 326.

[0140] The receiving unit 321 has a function of receiving a user input from a user. The first generating unit 322 has a function of generating a first prompt based on at least the user input received by the receiving unit 321.

[0141] The first acquisition unit 323 has a function of acquiring answer data regarding the character's answer to the user input from a predetermined language model 351, using the first prompt generated by the first generation unit 322. The first output unit 324 has a function of outputting an answer to the user via the character, based on the answer data acquired by the first acquisition unit 323.

[0142] The second determination unit 325 has a function of determining a filler effect based on the user input received by the reception unit 321. The second output unit 326 has a function of outputting the filler effect determined by the second determination unit 325 via the character during the period from when the reception unit 321 receives the user input until when the first output unit 324 outputs an answer.

[0143] Server device 305 includes at least a predetermined language model 351. Server device 305 uses language model 351 to create various types of response data in response to various prompts sent from computer device 301. There are no particular limitations on language model 351 as long as it is capable of creating response data in response to various prompts received from computer device 301, but it is preferable that language model 351 be, for example, a large-scale language model.

[0144] Next, an information processing method using the computer device 301 according to the fourth embodiment will be described. Fig. 12 is a flowchart showing an example of the information processing method according to at least one of the embodiments of the present disclosure.

[0145] In step S61, the computer device 301 accepts a user input from the user. In step S62, the computer device 301 determines a filler effect based on the user input accepted in step S61 (second determination).

[0146] In step S62, the computer device 301 outputs the filler effect determined in step S62 via the character (second output) between the time when the user input is received in step S61 and the time when the answer is output in step S66 described later.

[0147] In step S64, the computer device 301 generates a first prompt based on at least the user input received in step S61 (first generation). In step S65, the computer device 301 uses the first prompt generated in step S64 to obtain response data regarding the character's response to the user input from the language model 351 (first acquisition).

[0148] In step S66, the computer device 301 outputs an answer to the user via the character based on the answer data acquired in step S65 (first output), and then ends the process.

[0149] As one aspect of the fourth embodiment, a new technique can be provided for enhancing the expressiveness of a character in a response using a predetermined language model. Specifically, it becomes easier to realize natural expressions that make it appear as if the character is actually thinking of an answer in response to a user input. This also reduces the impression that the user may have of a long wait time before an answer is output.

[0150] In the fourth embodiment, the terms "computer device," "user input," "predetermined character," "character information," "first prompt," and "outputting an answer to the user via a character" can each adopt the contents described in the first or second embodiment to the extent necessary.

[0151] In addition, in the fourth embodiment, the "second prompt" and the "filler effect" may adopt the contents described in the second embodiment to the extent necessary.

[0152] Although the embodiments of the present disclosure have been described above in detail, the present invention is not limited to the above-described embodiments and can be appropriately modified, improved, etc. The present invention is defined by the claims and includes all modifications within the meaning and scope of the claims.

[0153] [Notes] The above-mentioned embodiments have been described in such a manner that a person having ordinary skill in the art to which the invention pertains can carry out the following invention.

[0154] [1] A program executed on a computer device, causing the computer device to realize the following: a reception function for receiving user input from a user; a first determination function for determining character information for a first prompt from character information preset as information relating to the personality and guidelines of action of a predetermined character based on at least the user input; a first generation function for generating a first prompt based on the user input and the character information for the first prompt; a first acquisition function for acquiring answer data relating to the character's answer to the user input from a predetermined language model using the first prompt; and a first output function for outputting an answer to the user via the character based on the answer data.

[0155] [2] The program according to [1] above, wherein the output includes audio and / or motion images of the character.

[0156] [3] The program according to [1] or [2] above, wherein the character information includes information about the character's hobbies and / or preferences.

[0157] [4] The program described in [3] above, wherein the first decision function is capable of, when the user input includes a topic related to the character's hobbies and / or preferences, including instructions for generating another topic related to the hobbies and / or preferences in the character information for the first prompt.

[0158] [5] The program described in [3] or [4] above, wherein the output includes audio and / or motion images of the character, and the first output function, when the user input includes a topic related to the character's hobbies and / or preferences, causes the audio and / or motion images to be different from when the topic is not included.

[0159] [6] The program described in any of [1] to [5] above, wherein the character information includes information regarding prohibited topics that the character is prohibited from talking about, and the first decision function is capable of, when the user input includes a topic that falls under the prohibited topic, including instructions for generating a different topic different from the prohibited topic in the character information for the first prompt.

[0160] [7] The program according to any one of [1] to [6] above, wherein the first generating function further generates the first prompt based on a history of the user input and the answers.

[0161] [8] The program according to any one of [1] to [7] above, wherein the first determination function further determines the character information for the first prompt from the character information based on a change condition of a predetermined action guideline.

[0162] [9] The program according to any one of [1] to [8] above, further causing the computer device to realize a conversion function that converts text included in the answer data acquired by the first acquisition function into wording appropriate to the character, and the first output function outputs the answer based on the text converted by the conversion function.

[0163]

[10] The program described in any of [1] to [9] above, wherein the output includes voice and / or motion images of the character, the first generation function is capable of generating a first prompt including an instruction sentence instructing the character to indicate an expression point in the text of the answer data where the character should express a predetermined emotion, if such an expression point is indicated in the answer data, and when the expression point is indicated in the answer data, the first output function outputs the answer while having the character express the expression of the emotion at least at the expression point through the voice and / or the motion image.

[0164]

[11] The program described in

[10] above, wherein the first generation function is capable of generating a first prompt including an instruction to indicate the beginning of the sentence in which a predetermined emotion should be expressed in addition to the point in the text of the answer data where the character should express that emotion, when that point is present, and when the expression point is indicated in the answer data, the first output function outputs the answer while having the character express the emotion through the voice and / or the motion image at least from the beginning of the sentence to the expression point.

[0165]

[12] The program described in

[10] or

[11] above, wherein, when the expression point is indicated in the response data, the first output function terminates the expression related to the expression of the emotion in response to a predetermined termination condition being satisfied after the expression of the emotion at the expression point.

[0166]

[13] The program according to any one of [1] to

[12] above, further causing the computer device to realize a second determination function that determines a filler effect based on the user input, and a second output function that outputs the filler effect determined by the second determination function via the character during the period from when the user input is received until the answer is output by the first output function.

[0167]

[14] The program described in

[13] above, wherein the second determination function analyzes the content input by the user and determines the filler effect based on the analysis results, and the analysis includes an analysis of whether the user input includes a question sentence.

[0168]

[15] The program described in

[13] or

[14] above, wherein the second determination function analyzes the content input by the user and determines the filler effect based on the analysis results, and the analysis includes an analysis of the user's emotions based on the text or voice of the user input.

[0169]

[16] The program according to any one of

[13] to

[15] above, further comprising: a second generation function that generates a second prompt including an instruction sentence that instructs the content input by the user to be judged; and a second acquisition function that uses the second prompt to acquire judgment data regarding the result of the judgment from the language model, wherein the second determination function determines the filler presentation based on the judgment data.

[0170]

[17] The program according to

[16] above, wherein the determination includes determining whether the user input has positive content or negative content.

[0171]

[18] The program according to

[16] or

[17] , wherein the determination includes determining whether the user input is different from the immediately preceding topic.

[0172]

[19] The program according to any one of

[13] to

[18] above, wherein the filler effect includes an effect of outputting at least a part of the user input as text and / or audio via the character.

[0173]

[20] An information processing method including causing a computer device to execute the following steps: a receiving step of receiving user input from a user; a first determination step of determining character information for a first prompt from character information preset as information relating to a character's personality and guidelines of action based on at least the user input; a first generation step of generating a first prompt based on the user input and the character information for the first prompt; a first acquisition step of acquiring answer data relating to the character's answer to the user input from a predetermined language model using the first prompt; and a first output step of outputting an answer to the user via the character based on the answer data.

[0174]

[21] A computer device comprising: a receiving means for receiving a user input from a user; a first determining means for determining character information for a first prompt from character information preset as information relating to a character's personality and guidelines of action based on at least the user input; a first generating means for generating a first prompt based on the user input and the character information for the first prompt; a first obtaining means for obtaining answer data relating to the character's answer to the user input from a predetermined language model using the first prompt; and a first output means for outputting an answer to the user via the character based on the answer data.

[0175]

[22] A program executed on a computer device, causing the computer device to realize: a reception function for receiving user input from a user; a first generation function for generating a first prompt based on at least the user input; a first acquisition function for using the first prompt to obtain answer data regarding an answer to the user input from a predetermined language model; and a first output function for outputting an answer to the user via a predetermined character based on the answer data, wherein the output includes audio and / or motion images of the character, and the first generation function is capable of generating a first prompt including an instruction sentence instructing the character to indicate, if there is an expression point in the text of the answer data where the character should express a predetermined emotion, and when the expression point is indicated in the answer data, the first output function outputs the answer while having the character express the emotion at least at the expression point through the audio and / or the motion image.

[0176]

[23] The program according to

[22] above, comprising one or more of the configurations according to [1] to

[19] above.

[0177]

[24] An information processing method including causing a computer device to execute the following steps: a receiving step of receiving user input from a user; a first generating step of generating a first prompt based on at least the user input; a first obtaining step of obtaining answer data regarding an answer to the user input from a predetermined language model using the first prompt; and a first output step of outputting an answer to the user via a predetermined character based on the answer data, wherein the output includes a voice and / or a motion image of the character, and the first generating step is capable of generating a first prompt including an instruction sentence instructing the character to indicate, if there is an expression point in the text of the answer data where the character should express a predetermined emotion, the expression point if indicated in the answer data, and wherein, in the first output step, the answer is output while the character expresses the expression of the emotion at least at the expression point through the voice and / or the motion image.

[0178]

[25] A computer device comprising: a receiving means for receiving user input from a user; a first generating means for generating a first prompt based on at least the user input; a first obtaining means for obtaining answer data relating to an answer to the user input from a predetermined language model using the first prompt; and a first output means for outputting an answer to the user via a predetermined character based on the answer data, wherein the output includes a voice and / or a motion image of the character, and the first generating means is capable of generating a first prompt including an instruction sentence for instructing the character to indicate, if there is an expression point in the text of the answer data where the character should express a predetermined emotion, the expression point if indicated in the answer data, and when the expression point is indicated in the answer data, the first output means outputs the answer while having the character express the emotion at least at the expression point through the voice and / or the motion image.

[0179]

[26] A program executed on a computer device, causing the computer device to realize: a reception function for receiving user input from a user; a first generation function for generating a first prompt based on at least the user input; a first acquisition function for acquiring answer data regarding an answer to the user input from a predetermined language model using the first prompt; and a first output function for outputting an answer to the user via a predetermined character based on the answer data; and further causing the computer device to realize: a second determination function for determining a filler effect based on the user input; and a second output function for outputting the filler effect determined by the second determination function via the character during the period from when the user input is received to when the answer is output by the first output function.

[0180]

[27] The program according to

[26] above, comprising one or more of the configurations according to [1] to

[19] above.

[0181]

[28] An information processing method comprising causing a computer device to execute: a receiving step of receiving a user input from a user; a first generating step of generating a first prompt based on at least the user input; a first obtaining step of obtaining answer data regarding an answer to the user input from a predetermined language model using the first prompt; and a first output step of outputting an answer to the user via a predetermined character based on the answer data; and further causing the computer device to execute: a second determining step of determining a filler effect based on the user input; and a second output step of outputting the filler effect determined in the second determining step via the character during the period from after the user input is received to before the answer is output in the first output step.

[0182]

[29] A computer device comprising: a receiving means for receiving a user input from a user; a first generating means for generating a first prompt based on at least the user input; a first obtaining means for obtaining answer data regarding an answer to the user input from a predetermined language model using the first prompt; and a first output means for outputting an answer to the user via a predetermined character based on the answer data; and further comprising: a second determining means for determining a filler effect based on the user input; and a second output means for outputting the filler effect determined by the second determining means via the character during the period from the acceptance of the user input to the output of the answer by the first output means.

[0183] 1, 101, 201, 301: Computer devices 5, 105, 205, 305: Server devices 51, 151, 251, 351: Language models

Claims

1. A program executed on a computer device, causing the computer device to realize the following: a reception function for receiving user input from a user; a first determination function for determining character information for a first prompt from character information previously set as information relating to the personality and guidelines of action of a specified character based on at least the user input; a first generation function for generating a first prompt based on the user input and the character information for the first prompt; a first acquisition function for using the first prompt to acquire answer data relating to the character's answer to the user input from a specified language model; and a first output function for outputting an answer to the user via the character based on the answer data.

2. The program of claim 1, wherein the character information includes information regarding the character's hobbies and / or preferences, and the first decision function is capable of, when the user input includes a topic related to the character's hobbies and / or preferences, including in the character information for the first prompt an instruction for generating another topic related to the hobbies and / or preferences.

3. The program described in claim 1 or 2, wherein the character information includes information regarding prohibited topics that the character is prohibited from talking about, and the first decision function is capable of including instructions in the character information for the first prompt to generate a different topic different from the prohibited topic when the user input includes a topic that falls under the prohibited topic.

4. The program according to claim 1 or 2, wherein the first predetermined function further determines character information for the first prompt from the character information based on a predetermined change condition for a course of action.

5. The program of claim 1 or 2, wherein the output includes audio and / or motion images of the character, the first generation function is capable of generating a first prompt including an instruction sentence instructing the character to indicate, if there is an expression point in the text of the answer data where the character should express a specified emotion, and when the expression point is indicated in the answer data, the first output function outputs the answer while having the character express the expression of the emotion through the audio and / or motion images at least at the expression point.

6. The program according to claim 1 or 2, further causing the computer device to realize a second determination function that determines a filler effect based on the user input, and a second output function that outputs the filler effect determined by the second determination function via the character during the period from when the user input is received until the answer is output by the first output function.

7. An information processing method comprising causing a computer device to execute the following steps: a receiving step of receiving user input from a user; a first determination step of determining character information for a first prompt from character information previously set as information relating to a character's personality and guidelines for action based on at least the user input; a first generation step of generating a first prompt based on the user input and the character information for the first prompt; a first acquisition step of acquiring answer data relating to the character's answer to the user input from a predetermined language model using the first prompt; and a first output step of outputting an answer to the user via the character based on the answer data.

8. A computer device comprising: a receiving means for receiving user input from a user; a first determining means for determining character information for a first prompt from character information previously set as information relating to a character's personality and guidelines for action based on at least the user input; a first generating means for generating a first prompt based on the user input and the character information for the first prompt; a first obtaining means for obtaining answer data relating to the character's answer to the user input from a predetermined language model using the first prompt; and a first output means for outputting an answer to the user via the character based on the answer data.

Images