Information processing method, program, and information processing apparatus

Abstract

To output a response conversation having features similar to those of a target input conversation.SOLUTION: An information processing method according to an embodiment of the present disclosure causes a computer to execute a process of extracting a feature of a target on the basis of a dialogue including an input conversation by the target and a response conversation by a language model, and when it is determined that the extracted feature satisfies a predetermined condition, outputting a feature prompt based on the feature of the target at a time when the predetermined condition is satisfied. In the information processing method according to an embodiment of the present disclosure, when the input conversation is input, the feature of the target is output together with the response conversation.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

[0002] Conventionally, dialogue systems for dialogue with users have been proposed. For example, the dialogue system described in Patent Document 1 includes a speech recognition processing unit that recognizes the content of utterances from the user, and a dialogue engine that generates utterances to the user in accordance with an application based on the recognition result. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2003-036095 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the technology described in Patent Document 1 does not take into consideration the point of outputting a response conversation that has similar characteristics to the target input conversation.

[0005] The present disclosure has been made in consideration of the above circumstances, and aims to provide an information processing method and the like for outputting characteristic prompts that can construct a pseudo-personality model that behaves in a similar manner to a target. [Means for solving the problem]

[0006] An information processing method according to one embodiment of the present disclosure extracts features of a target based on a dialogue including an input conversation by the target and a response conversation by a language model, and, if it is determined that the extracted features satisfy a predetermined condition, causes a computer to execute a process of outputting a feature prompt based on the features of the target at the time the predetermined condition is satisfied. [Effects of the Invention]

[0007] In an information processing method according to an embodiment of the present disclosure, it is possible to output characteristic prompts that can construct a pseudo-personality model that behaves in a similar manner to the target. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is an explanatory diagram illustrating an overview of an information processing system. [Figure 2] FIG. 2 is a block diagram illustrating an example of the configuration of a user terminal. [Figure 3] FIG. 2 is an explanatory diagram showing system prompts input into a language model. [Figure 4] FIG. 2 is an explanatory diagram showing input and output of a language model in a dialogue between a target and the language model. [Figure 5] FIG. 10 is an explanatory diagram illustrating output of a feature prompt by a language model. [Figure 6] 10A and 10B are explanatory diagrams showing examples of dialogues and feature prompts displayed on a user terminal. [Figure 7] FIG. 10 is an explanatory diagram illustrating an example of a feature prompt table. [Figure 8] 10 is a flowchart illustrating a feature prompt output process performed by a terminal processing unit of a user terminal. [Figure 9] FIG. 10 is an explanatory diagram showing the output of an imitation sentence by a language model. [Figure 10] 10 shows an example of a display of an imitation sentence on a user terminal. [Figure 11] 10 is a flowchart showing an imitation sentence output process performed by a terminal processing unit of a user terminal. [Figure 12] FIG. 10 is an explanatory diagram showing an overview of an information processing system according to a second embodiment. [Figure 13] FIG. 1 is a block diagram illustrating an example of the configuration of an information processing device. [Figure 14] FIG. 10 is an explanatory diagram illustrating an example of a feature prompt table. [Figure 15] 10 is a flowchart illustrating an example of a characteristic prompt output process performed by a processing unit of the information processing device. [Figure 16]10 shows an example of a display of an imitation sentence on a user terminal according to the second embodiment. [Figure 17] 10 is a flowchart showing an imitation sentence output process performed by a processing unit of the information processing device. DETAILED DESCRIPTION OF THE INVENTION

[0009] (Embodiment 1) The language model M according to the first embodiment extracts features of a target based on a dialogue including an acquired input conversation of the target and a response conversation by the language model M, and outputs a feature prompt based on the extracted features when the extracted features satisfy a predetermined condition. Hereinafter, the first embodiment will be described with reference to the drawings.

[0010] FIG. 1 is an explanatory diagram showing an overview of an information processing system S. The information processing system S includes a language model server 1 and a user terminal 2. The language model server 1 is, for example, a cloud server. The user terminal 2 is a smartphone, tablet terminal, personal computer, or the like owned by a user to whom a language model M is to be similarized. In this embodiment, the user terminal 2 is described as a personal computer. The language model server 1 and the user terminal 2 can communicate with each other via a network N. The language model server 1 stores the language model M.

[0011] In this embodiment, the target to which the language model M is to be similar is a user, but is not limited to this. The target to which the language model M is to be similar may be, for example, a character appearing in a work such as an anime, game, or movie, a virtual personality such as a VTuber or avatar, or an AI conversation bot character such as the AI ​​(artificial intelligence) bot "Rinna" developed by Microsoft Japan. In the former case, the user inputs an input conversation while pretending to be the character for which the user wishes to generate a feature prompt. In the latter case, the user inputs a response sentence from the AI ​​bot obtained through an app in which the AI ​​bot operates as the input conversation.

[0012] The user terminal 2 transmits a system prompt indicating an instruction to the language model M and an input conversation input by the user to the language model server 1. The system prompt and input conversation received by the language model server 1 are input to the language model M stored in the language model server 1. In the following description, it is assumed that transmitting a system prompt or an input conversation to the language model server 1 and inputting a system prompt or an input conversation to the language model M are synonymous.

[0013] The language model M outputs a response conversation to the input conversation in accordance with the instructions indicated by the input system prompt. The language model M also extracts user features from the input conversation and outputs the extracted features along with the response conversation. If the extracted features satisfy a predetermined condition, the language model M outputs a feature prompt based on the extracted features to the user terminal 2. The language model server 1 transmits the response conversation, user features, or feature prompt output by the language model M to the user terminal 2. Hereinafter, the output of the response conversation, user features, or feature prompt by the language model M and the transmission of the response conversation, user features, or feature prompt to the user terminal 2 will be described as synonymous. The feature prompt is a sentence that indicates the user's personality and speech characteristics and is generated based on the user features extracted from the user's input conversation. The output feature prompt does not necessarily need to be displayed on a screen in a user-recognizable format; it may be stored in the feature prompt table 221 and used confidentially as a prompt to be transmitted to the language model M. The feature prompt may not only be output when a predetermined condition is satisfied, but also be output at the user's request or automatically updated as needed.

[0014] FIG. 2 is a block diagram showing an example configuration of the user terminal 2. The user terminal 2 includes a device processing unit 21, a storage unit 22, a communication unit 23, a display unit 24, and an input unit 25. The device processing unit 21 is configured with a central processing unit (CPU), a micro processing unit (MPU), a graphics processing unit (GPU), a neural processing unit (NPU), a tensor processing unit (TPU), or the like, and performs various control processes, arithmetic processes, and the like. Note that the functions of the user terminal 2 may be realized by multiple devices, or may be a device corresponding to a node on a blockchain. Furthermore, the user terminal 2 may store a language model M and perform some or all of the processes executed by the language model server 1.

[0015] The storage unit 22 of the user terminal 2 stores an application program Pa, which is an application that receives input conversation from a user and displays the input conversation and a response conversation output by the language model M, and a feature prompt table 221. The application program Pa is provided to the user terminal 2 using, for example, a recording medium 22a. The recording medium 22a is, for example, a portable memory. Examples of the portable memory include a CD-ROM, a USB (Universal Serial Bus) memory, an SD card, a micro SD card, and a Compact Flash Memory (registered trademark). When the recording medium 22a is a portable memory, the processing element of the device processing unit 21 may read the program P from the recording medium 22a using a reading device (not shown). The device processing unit 21 of the user terminal 2 may obtain the application program Pa via the Internet and store it in the storage unit 22. Details of the feature prompt table 221 will be described later.

[0016] The communication unit 23 of the user terminal 2 is a communication module or a communication interface for wirelessly communicating with the language model server 1. The terminal processing unit 21 communicates with the language model server 1 via the communication unit 23 and an external network N.

[0017] The display unit 24 of the user terminal 2 is composed of a display and displays a screen for accepting input of an input conversation from the user, or a screen showing a dialogue including an input conversation input by the user and a response conversation received from the language model server 1.

[0018] The input unit 25 of the user terminal 2 is configured with, for example, a keyboard or a mouse, and accepts input of conversation from the user. The input unit 25 may also accept information input via an input device such as a touch panel, a camera, a microphone, or various sensors.

[0019] FIG. 3 is an explanatory diagram showing a system prompt input to a language model M. The system prompt input to the language model M includes an instruction to extract user characteristics, including tone of voice, vocabulary, thoughts, values, and emotions, from the user's input conversation (for example, the sentence shown in FIG. 3, "Please extract user characteristics, including tone of voice, vocabulary, thoughts, values, and emotions"). The system prompt includes an instruction to output a response conversation that mirrors the user based on the user's input conversation (for example, the sentence shown in FIG. 3, "Please output a response conversation that mirrors the user, based on the user's input conversation"). The system prompt includes a sentence indicating a rule for extracting user characteristics. The sentence indicating the rule includes an instruction to ask the user about values, frequently used vocabulary, personality, beliefs, detailed profile, current feelings, etc., through the response conversation (for example, the sentence shown in FIG. 3, "Please ask the user about values, frequently used vocabulary, current feelings, etc., through the response conversation"). The sentence indicating the rule includes an instruction to output the user's understanding of the feature (for example, the sentence in FIG. 3, "When outputting a response conversation, please output your understanding of the user's feature."). The sentence indicating the rule includes an instruction to calculate the similarity between the input conversation and the response conversation (for example, the sentence in FIG. 3, "Please output the similarity between the input conversation and the response conversation."). The sentence indicating the rule includes an output example of the response conversation (for example, the sentence in FIG. 3, "Please output the response conversation, for example, as the following sentence (reflecting the features extracted from the user's input conversation)." Output example: That's good. I like ○○○○. And I think ○○○○ is important. Now, my next question is, do you have any words that I like or phrases that I use often?"). The sentence indicating the rule includes an output format and an output example of the feature (for example, the sentence in FIG. 3, "Please output the understanding, similarity, and extracted user features in the following format, for example," and the format of 10 lines following the sentence). The sentence indicating the rule includes an instruction to accept correction of the similarity output from the user (for example, an instruction to output the sentence "Please correct the similarity" shown in FIG. 3).The system prompt also includes a predetermined condition for outputting a feature prompt (for example, the sentence in FIG. 3, "When the similarity entered (modified) by the user exceeds 80 or when the input conversation entered by the user includes an instruction to output a feature prompt"). The system prompt also includes an instruction for outputting a feature prompt when the predetermined condition is met (for example, the sentence in FIG. 3, "(When the following conditions are met, please output a feature prompt.)"). The trigger for generating a feature prompt may be an automatic determination or a response to a direct or indirect request by the user, such as pressing a button, and the process executed at that time may be a separate program or prompt specialized for generating the feature prompt. The language model M may also receive input of external information collected by functions such as past conversation history and RAG (Retrieval-Augmented Generation).

[0020] FIG. 4 is an explanatory diagram showing input and output of the language model M in a dialogue between a target and the language model M. The language model M is configured, for example, by a GPT (Generative Pretrained Transformer). The language model M outputs a response conversation to an input conversation or a predetermined item based on an instruction indicated by a system prompt (see FIG. 3) input in advance.

[0021] After receiving a system prompt, the language model M receives an input of a user's input conversation. The language model M outputs a response conversation and extracted features according to the instructions included in the system prompt. Figure 4 shows the first input conversation and response conversation (first dialogue) and the second input conversation and response conversation (second dialogue).

[0022] The initial input conversation includes, for example, an instruction to output a response conversation by imitating (mirroring) the user. The language model M outputs the response conversation mirroring the input conversation and features extracted from the input conversation in a format instructed by the system prompt. In the output including the initial response conversation (first output), for example, a response conversation including a question asking about the user's values ​​and user features extracted from the initial input conversation are output. In the first output, the language model M also outputs a comprehension and a similarity. The comprehension is calculated, for example, based on the amount of information in the extracted features. The similarity indicates the similarity of the initial response conversation to the initial input conversation, calculated, for example, by cosine similarity or Euclidean similarity. The output of the language model M includes a prompt for the user to rate (correct) the similarity.

[0023] The second input conversation includes, for example, a score obtained by scoring (correcting) the similarity in the first output and the user's answer to the question included in the first response conversation. When the second input conversation is input, the language model M outputs a response conversation mirroring the input conversation and features extracted from the input conversation in a format instructed by the system prompt, just as when the first input conversation was input. The second response conversation is generated by mirroring the first and second input conversations. The user features included in the second output are features extracted from the first and second input conversations. That is, the language model M outputs a response conversation to the input conversation and improves itself by repeating dialogue with the user, thereby increasing the accuracy of mirroring the user. The understanding level output together with the response conversation to the second or subsequent input conversation is calculated, for example, based on the amount of information on features extracted from all input conversations input up to that point. The similarity level output together with the response conversation to the second or subsequent input conversation indicates the similarity between the input conversation input immediately before and the response conversation output for that input conversation. The similarity may be the average value of the similarity between each of the multiple input conversations that have been input so far and the latest response conversation, or the similarity between all of the multiple input conversations that make up the history and the latest response conversation.

[0024] FIG. 5 is an explanatory diagram illustrating the output of a feature prompt by the language model M. When an input conversation input by a user satisfies a predetermined condition, the language model M outputs a feature prompt based on the extracted features. The predetermined condition may be, for example, that the input conversation includes a statement indicating that the similarity of the previous response conversation by the language model M is equal to or greater than a predetermined threshold (e.g., 80 points) (the input conversation on the left in FIG. 5), that a user request such as pressing a button has been made, or that an instruction to output a feature prompt is included (the input conversation on the right in FIG. 5). The language model M may automatically continue to update the latest feature prompt until the predetermined condition is satisfied, or may output a feature prompt when the number of times the input conversation is input from the user terminal 2 reaches a predetermined number or more, as the predetermined condition has been satisfied. The device processing unit 21 of the user terminal 2 may determine whether the features extracted based on the input conversation input by the user satisfy the predetermined condition, and if it determines that the predetermined condition is satisfied, may transmit information instructing the language model M to output a feature prompt. When an input conversation that satisfies predetermined conditions based on a previously input system prompt is input, the language model M outputs a characteristic prompt based on characteristics extracted from multiple previous input conversations. The characteristic prompt includes characteristics of the user's personality (thoughts, personality, values, memory, knowledge, or mood) and characteristics of the user's tone of voice (tone of voice, vocabulary, or personal expressions). Note that the characteristic prompt may also include non-verbal characteristics (appearance, facial expression, tone of voice, tempo of conversation, actions, gestures, or behavior). In other words, the characteristic prompt is output based on the user's characteristics output along with the immediately preceding response conversation. The characteristic prompt is a prompt that causes the language model M to output a sentence that imitates the user. In other words, when a characteristic prompt is input, the language model M outputs a sentence that imitates the user (see FIG. 6). The characteristic prompt may include information related to at least one item of the characteristics of the user's personality, including the user's thoughts, personality, values, or mood, and may also include information related to multiple items.Furthermore, the feature prompts may include information relating to at least one of the following features of speech characteristics: the user's speech tone, vocabulary, memory, knowledge, thoughts, personality, values, facial expressions, tone of voice, conversational tempo, actions, gestures, behavior, or speech tone characteristics including personal expressions, or may include information relating to multiple features. Furthermore, the feature prompts do not necessarily need to be displayed on a screen or the like in a format recognizable to the user, and may be stored in the feature prompt table 221 and used in secret as prompts to be sent to the language model.

[0025] FIG. 5 shows two examples of input conversations that satisfy a predetermined condition when the language model M outputs a feature prompt. The input conversation shown on the left is an input conversation that includes a notification that the similarity of the previous response conversation is equal to or greater than a predetermined threshold. The input conversation shown on the right is an input conversation that includes an instruction to output a feature prompt. When an input conversation that satisfies a predetermined condition is input, the language model M outputs a feature prompt that expresses features extracted from the previous input conversation in multiple sentences. The feature prompt may include user features extracted from the input conversation that satisfies the condition. The language model M may output a feature prompt when a button for generating a feature prompt is pressed or when the calculated level of understanding is equal to or greater than a predetermined value. The predetermined threshold value of similarity or the predetermined value of understanding, which are the predetermined conditions that cause the language model M to output a feature prompt, may be changeable as needed by including a value change instruction in a system prompt or the input conversation.

[0026] 6 is an explanatory diagram showing an example of the display of dialogues and feature prompts on the user terminal 2. The device processing unit 21 of the user terminal 2 displays, in chat format, the input conversation that has been received, and the response conversation and feature prompts that have been output by the language model M on the display unit 24. These outputs may be expressed as audio, video, or the movements of a 3D model or robot.

[0027] The device processing unit 21 of the user terminal 2 displays the input conversation input by the user via the input unit 25 on the display unit 24, and transmits the input conversation to the language model server 1 and inputs it to the language model M. The language model M outputs a response conversation that mirrors the user's input conversation in a format instructed by the system prompt, as well as user features extracted from the input conversation. The device processing unit 21 alternately displays the user's input conversation and the response conversation output by the language model M in response to the input conversation, thereby displaying a dialogue including the input conversation and the response conversation in a chat format. Furthermore, if the input conversation input to the language model M satisfies a predetermined condition (if the input conversation includes a statement indicating that the similarity of the previous response conversation based on the language model M is equal to or greater than a predetermined threshold, or if the input conversation includes an instruction to output a feature prompt), the language model M outputs a feature prompt that expresses the features extracted from the user's input conversation in sentences. Note that the example shown in FIG. 6 illustrates a case where the input conversation includes an instruction to output a feature prompt. The feature prompt is displayed next to (below) the input conversation that satisfies the predetermined condition. The device processing unit 21 can store (copy) the feature prompt in the feature prompt table 221 of the storage unit 22 by accepting the selection of a copy command displayed together with the output feature prompt. Also, as shown in Fig. 6, a model generation command instructing model generation may be displayed. When the model generation command is displayed, the device processing unit 21 inputs the output feature prompt into the language model M and generates a language model based on the feature prompt.

[0028] FIG. 7 is an explanatory diagram showing an example of the feature prompt table 221. The feature prompt table 221 stores feature prompts output by the language model M. The management items (fields) of the feature prompt table 221 include, for example, a user ID field and a feature prompt field. The user ID field stores a user ID assigned to a user whose feature is extracted through a dialogue with the language model M. That is, in the user terminal 2, multiple users can input input conversations and obtain their respective feature prompts through a dialogue with the language model M. The feature prompt field stores a feature prompt output by the language model M based on a feature extracted from the input conversation of the user corresponding to each user ID. The feature prompt table 221 may also store a history of the dialogue between each user and the language model M. If the target is a character appearing in a work or an AI conversation bot character, the feature prompt may be associated with a character ID and stored (recorded) in the feature prompt table 221. A test environment may be prepared to test the behavior of the generated feature prompt, or the generated feature prompt may be provided as a production bot service. Furthermore, the feature prompt may be associated with a similarity to the input conversation of the response conversation immediately before the language model M outputs the feature prompt, and the similarity may be stored (recorded) in the feature prompt table 221 together with a user ID or a character ID. At this time, the feature prompt and the similarity associated with the feature prompt may be read from the feature prompt table 221 and used for tuning (retraining) the language model M. That is, the device processing unit 21 reads the feature prompt temporarily stored by the interruption process based on the user ID at a later date. The device processing unit 21 may then provide the read feature prompt to the language model M together with the user's new input conversation and the system prompt, thereby continuing to regenerate the interrupted feature prompt until the similarity reaches a predetermined value or greater.

[0029] 8 is a flowchart showing a feature prompt output process by the device processing unit 21 of the user terminal 2. The device processing unit 21 of the user terminal 2 accepts input of a system prompt (S1). The device processing unit 21 inputs the system prompt to the language model M (sends it to the language model server 1) (S2). The device processing unit 21 accepts input of an input conversation (S3). The device processing unit 21 inputs the input conversation to the language model M (sends it to the language model server 1) (S4). The device processing unit 21 acquires the output of the language model M (S5). The device processing unit 21 determines whether the output of the language model M is a feature prompt (whether it is a response conversation) (S6). If the output of the language model M is a feature prompt (S6: YES), the device processing unit 21 displays the acquired feature prompt on the display unit 24 (S7) and ends the process. If the output of the language model M is not a feature prompt (S6: NO), that is, if the output of the language model M is a response conversation, the device processing unit 21 displays the acquired response conversation on the display unit 24 (S8) and returns the process to S3. Furthermore, the system prompt or program for processing to output a feature prompt based on the conversation history with the user and reference information may be implemented and called as a separate system independent of the system prompt that performs mirroring and similarity measurement to extract the target features.

[0030] 9 is an explanatory diagram showing the output of an imitation sentence by a language model M. When a query specifying a feature prompt output from the language model M, the role of the language model M, and the content of the sentence to be output is input, the device processing unit 21 of the user terminal 2 imitates the user based on the input feature prompt, and outputs a sentence with the content specified in the query.

[0031] As shown in FIG. 9 , a feature prompt indicating a user's characteristics read from the storage unit 22 in the user terminal 2 is input to the language model M. The feature prompt input to the language model M includes the sentence, "You are a user with the input characteristics. Please output a sentence as if you were the user," as an instruction to output a sentence by imitating the user indicated by the feature prompt. In the example shown in FIG. 9 , the sentence, "Please introduce yourself," which specifies the content of the sentence to be output, is input as a query. When such a feature prompt and query are input, the language model M outputs a sentence (imitated sentence) that imitates the user's speech characteristics (tone of speech, vocabulary, or personal expression) based on the user's characteristics included in the feature prompt. Furthermore, the language model M reflects the user's personality characteristics (thoughts, personality, values, or mood) included in the feature prompt into the content of the sentence specified in the query, and outputs the imitated sentence. As a result, the imitated sentence output by the language model M imitates the user. When generating and outputting an imitated sentence, reference information acquired using a past conversation history or a method such as RAG may be reflected in addition to the feature prompt. Additionally, instructions for generating and outputting these feature prompts may be contained in an external application rather than in system prompts, and may be transmitted in a manner that is transparent to the user.

[0032] FIG. 10 shows an example of the display of imitation sentences on the user terminal 2. The device processing unit 21 of the user terminal 2 displays the input feature prompt and query, as well as the imitation sentence, on the display unit 24, for example, in a chat format. The feature prompt and query may be input from a terminal device other than the user terminal 2, transmitted from the terminal device to the language model server 1, and input to the language model M. In this case, the terminal device may obtain the imitation sentence output by the language model M from the language model server 1 and display it. The device processing unit 21 of the user terminal 2 can transmit the feature prompts stored in the storage unit 22 to another terminal device via the network N or a portable memory. The display of the imitation sentences is not limited to text-based screen display, and may also be expressed using images, audio, video, 3D avatars, stereoscopic images, dolls, robots, etc. that resemble people.

[0033] 11 is a flowchart showing an imitation sentence output process by the device processing unit 21 of the user terminal 2. The device processing unit 21 of the user terminal 2 reads out a feature prompt from the storage unit 22 (S11) and accepts a query input (S12). The device processing unit 21 displays the read out feature prompt and the input query on the display unit 24 (S13). The device processing unit 21 inputs the read out feature prompt and the input query to the language model M (sends them to the language model server 1) (S14). Note that in S14, the user's past conversation history or reference information may also be input to the language model M. The device processing unit 21 acquires the imitation sentence output by the language model M from the language model server 1 (S15). The device processing unit 21 displays the acquired imitation sentence on the display unit 24 (S16), and ends the process. Note that some or all of the information, such as the feature prompt, conversation history, reference information, query, and imitation sentence, may or may not be displayed in a manner that is visible to the user.

[0034] According to the above configuration and processing, it is possible to extract user characteristics through a dialogue including a user's input conversation and a response conversation output by the language model M by mirroring the user's input conversation, and obtain a feature prompt based on the extracted characteristics. Furthermore, by inputting a feature prompt into the language model M, it is possible to output an imitation sentence that imitates the user. Note that the target mirrored by the language model M is not limited to the user. As described above, the target mirrored by the language model M may be a conversation performed by a user imitating the characteristics of a specific person, or a fictional character or AI conversation bot character that appears in a work. That is, the user terminal 2 may input, for example, a specific person's quotes, notes, profile information, or video, lines, settings, and scenarios of a character that appears in a work, an input conversation input by a user imitating a character in a work, or an input conversation by an AI conversation bot character into the language model M, and the language model M may output a response conversation that mirrors the lines of a character in a work, an input conversation input by a user imitating a character in a work, or an input conversation by an AI conversation bot character.

[0035] (Embodiment 2) FIG. 12 is an explanatory diagram showing an overview of an information processing system S according to the second embodiment. The information processing system S according to the second embodiment includes an information processing device 3. The information processing device 3 is, for example, a server device owned by a provider that provides users with imitation sentences that mirror the user using a language model M in response to a user's request. The information processing device 3 inputs pre-stored system prompts and an input conversation input at a user terminal 2 into the language model M of the language model server 1. The information processing device 3 transmits the response conversation and the user's features output by the language model M to the user terminal 2. The information processing device 3 also stores the feature prompts output by the language model M.

[0036] 13 is a block diagram showing an example configuration of an information processing device 3. The information processing device 3 includes a processing unit 31, a storage unit 32, and a communication unit 33. Note that the functions of the information processing device 3 may be realized by multiple server devices or computers, or may be a device corresponding to a node on a blockchain. Furthermore, the information processing device 3 may execute some or all of the processes executed by the language model server 1 or the user terminal 2.

[0037] The processing unit 31 of the information processing device 3 is composed of a CPU (Central Processing Unit), an MPU (Micro Processing Unit), a GPU (Graphics Processing Unit), an NPU (Neural Processing Unit), or a TPU (Tensor Processing Unit), and performs various control processes, arithmetic processes, etc. by reading and executing a program P (program product) etc. that is pre-stored in the memory unit 32.

[0038] The storage unit 32 of the information processing device 3 is, for example, a volatile memory or a non-volatile memory. The storage unit 32 stores a program P, system prompt data SP, and a feature prompt table 321. The program P may be provided to the information processing device 3 using a computer-readable storage medium 32a. The storage medium 32a may be, for example, a portable memory. Examples of the portable memory include a CD-ROM, a USB memory, an SD card, a micro SD card, and a Compact Flash Memory (registered trademark). When the storage medium 32a is a portable memory, the processing element of the processing unit 31 may read the program P from the storage medium 32a using a reading device (not shown). The read program P is written to the storage unit 32. Furthermore, the program P may be provided to the information processing device 3 by the communication unit 33 communicating with an external device.

[0039] The system prompt data SP stored in the storage unit 32 of the information processing device 3 is text data of a system prompt similar to the system prompt shown in FIG. 3. Note that, in the following description, reading the system prompt data SP and reading a system prompt are considered to be synonymous. The storage unit 32 of the information processing device 3 stores the system prompt data SP in advance. The system prompt data SP may be provided to the information processing device 3, for example, using a computer-readable storage medium 32b. The storage medium 32b may be, for example, a portable memory. If the storage medium 32b is a portable memory, the processing element of the processing unit 31 may read the program P from the storage medium 32b using a reading device (not shown). The read system prompt data SP is written to the storage unit 32. Furthermore, the system prompt data SP may be provided to the information processing device 3 by the communication unit 33 communicating with an external device.

[0040] The communication unit 33 of the information processing device 3 is a communication module or communication interface for communicating with other devices such as the language model server 1 or the user terminal 2 via wired or wireless means, and is, for example, a wide-area wireless communication module such as LTE (registered trademark), 4G, or 5G. The processing unit 31 communicates with the language model server 1 or the user terminal 2 via the communication unit 33 through an external network N such as the Internet.

[0041] FIG. 14 is an explanatory diagram showing an example of the feature prompt table 321. The record layout of the feature prompt table 321 is the same as that of the feature prompt table 221 (see FIG. 7) stored in the user terminal 2 in embodiment 1. The feature prompt table 321 may also include information related to reference information such as past conversation history and RAG. The user ID may be an ID corresponding to the user terminal 2 that accepted the input of the input conversation and transmitted it to the information processing device 3 to cause the language model M to output a feature prompt. That is, the information processing device 3 acquires input conversations from each of the multiple user terminals 2, and uses the language model M to acquire and store feature prompts that indicate the personality and tone of voice characteristics of the user who owns each user terminal 2. The processing unit 31 of the information processing device 3 reads out the feature prompt corresponding to the user ID specified in the user terminal 2 from the feature prompt table 321, and inputs the read feature prompt into the language model M, thereby causing the language model M to output the user's imitation sentence.

[0042] 15 is a flowchart showing an example of a feature prompt output process by the processing unit 31 of the information processing device 3. The processing unit 31 of the information processing device 3 acquires an input conversation input at the user terminal 2 from the user terminal 2 (S21). The processing unit 31 reads a system prompt from the storage unit 32 (S22). The processing unit 31 inputs the read system prompt into the language model M (transmits it to the language model server 1) (S23). The processing unit 31 inputs the input conversation into the language model M (transmits it to the language model server 1) (S24). Note that a past conversation history or reference information of the user may be input to the language model M in S24. The processing unit 31 acquires a response conversation output by the language model M (S25). The processing unit 31 outputs the acquired response conversation to the user terminal 2 (S26). During these processes, history information, reference information, input information from a microphone, camera, sensor, etc., or variable parameters such as the current date and time and information of the logged-in user may be transmitted to the language model M along with the prompt and query.

[0043] The processing unit 31 calculates the similarity between the response conversation output to the user terminal 2 and the input conversation input immediately before outputting the response conversation (S27). The processing unit 31 determines whether the calculated similarity between the input conversation and the response conversation is equal to or greater than a predetermined threshold (S28). Note that the processing unit 31 may also determine whether the similarity output by the language model M in S28 or the similarity input (modified) by the user at the user terminal 2 is equal to or greater than a predetermined threshold.

[0044] If the similarity is equal to or greater than a predetermined threshold (S28: YES), or if a generation request is received from the user, the processing unit 31 generates a feature prompt based on the user's features output by the language model M along with the response conversation (S29). The processing unit 31 stores the generated feature prompt in the feature prompt table 321 in association with the user ID of the user associated with the user terminal 2 to which the input conversation was input (S30), and terminates the processing. If the similarity is less than the predetermined threshold (S28: NO), the processing unit 31 returns the processing to S21. Note that, in the processing after the processing is returned from S28, the reading of the system prompt according to S22 and the input of the system prompt according to S23 to the language model M may be skipped. Also, if the input conversation input in S21 includes an instruction to output a feature prompt, the processing unit 31 may generate a feature prompt based on the user's features output along with the most recently output response conversation. Also, the feature prompt may be automatically generated, updated, or displayed whenever a change occurs in the extracted feature information, even if there is no generation request from the user or a predetermined condition is not satisfied.

[0045] FIG. 16 shows an example of the display of an imitation sentence on a user terminal 2 according to the second embodiment. The device processing unit 21 of the user terminal 2 according to the second embodiment accepts input of a user ID (the user ID of the user to be imitated by the language model M) and a query (for example, the sentence "Please introduce yourself" shown in FIG. 16) corresponding to a feature prompt for outputting an imitation sentence. Some or all of these input contents may be transmitted in a form that is not visible to the user. The processing unit 31 of the information processing device 3 reads out the feature prompt corresponding to the input user ID from the feature prompt table 321, and inputs the read feature prompt and the query input on the user terminal 2 into the language model M, thereby causing the language model M to output the imitation sentence.

[0046] The device processing unit 21 of the user terminal 2 displays the input user ID and query, and the imitation sentence on the display unit 24, for example, in a chat format. The display of the imitation sentence is not limited to a text-based screen display, and may be expressed using images, sounds, videos, 3D avatars, stereoscopic images, dolls, robots, etc. that resemble people. The device processing unit 21 may acquire a feature prompt (feature prompt input into the language model M) corresponding to the input user ID from the information processing device 3, and display it on the same screen as the imitation sentence. The feature prompt may be provided in advance from the information processing device 3 to the user terminal 2 via communication or a recording medium, and may be input into the language model M by an operation on the user terminal 2, as in the first embodiment.

[0047] 17 is a flowchart showing an imitation sentence output process by the processing unit 31 of the information processing device 3. The processing unit 31 of the information processing device 3 acquires, from the user terminal 2, the user ID and query of the user to be imitated by the language model M, which have been input to the user terminal 2 (S41). The processing unit 31 reads out a feature prompt corresponding to the acquired user ID from the feature prompt table 321 (S42). The processing unit 31 inputs the read feature prompt and query into the language model M (S43). Note that past conversation history or reference information of the user may also be input into the language model M in S43. The processing unit 31 acquires the imitation sentence output by the language model M (S44). The processing unit 31 outputs the acquired imitation sentence to the user terminal 2 (S45), and ends the process.

[0048] According to the above configuration and processing, the information processing device 3 can extract user features through a dialogue including a user's input conversation and a response conversation output by the language model M by mirroring the user's input conversation, obtain feature prompts based on the extracted features, and store the feature prompts for each user. The information processing device 3 can also output imitation sentences that imitate the user by inputting feature prompts corresponding to the user into the language model M. In this embodiment, the processing unit 31 of the information processing device 3 may output the language model M's level of understanding of the user based on the input conversation and the response conversation. In this case, for example, when the input conversation or the response conversation is input, the processing unit 31 inputs the input conversation or the response conversation into a learning model trained to output feature elements (such as tone of voice, vocabulary, thoughts, values, or emotions) included in the input conversation or the response conversation, outputs the feature elements included in the input conversation or the response conversation, and outputs the level of understanding according to the number of elements output from the input conversation or the response conversation during the dialogue. In addition, the processing unit 31 of the information processing device 3 may input the input conversation and the response conversation into a learning model that has been trained to output the similarity between the input conversation and the response conversation when the input conversation and the response conversation are input, and output the similarity between the input conversation and the response conversation.

[0049] The embodiments disclosed herein are illustrative in all respects and should not be considered limiting. The technical features described in each embodiment may be combined with one another, and the scope of the present invention is intended to include all modifications within the scope of the claims and equivalents thereto. Furthermore, independent and dependent claims described in the claims may be combined with one another in any and all combinations, regardless of the reference format. Furthermore, while the claims use a format in which a claim references two or more other claims (multiple claim format), this is not limiting. Multiple claims (multiple multiple claims) that reference at least one other multiple claim may also be used. [Explanation of symbols]

[0050] 1: Language model server 2: User terminal 21: Terminal processing section 22: Storage section 221: Feature prompt table 22a: Recording medium 23: Communications Department 24:Display section 25: Input section 3: Information processing equipment 31: Processing section 32: Storage section 321: Feature prompt table 32a: Recording medium 32b: Recording medium 33: Communications Department M: Language model N: Network P: Program Pa: Application program SP: System prompt data S: Information Processing System

Claims

1. extracting features of the target based on a dialogue including an input conversation by the target and a response conversation by a language model; If it is determined that the extracted feature satisfies a predetermined condition, a feature prompt based on the feature of the target at the time when the predetermined condition is satisfied is output. An information processing method that causes a computer to execute a process.

2. When the input conversation is input, the response conversation and the characteristics of the target are output. The information processing method according to claim 1 .

3. When the input conversation is input, the response conversation is output; Obtaining a similarity between the input conversation and the output response conversation 3. The information processing method according to claim 1.

4. If the similarity is equal to or greater than a predetermined threshold, output the feature prompt. The information processing method according to claim 3 .

5. The characteristics include the subject's tone of voice, vocabulary, personal expressions, memory, knowledge, thoughts, personality, values, or mood.

3. The information processing method according to claim 1.

6. providing system prompts to the language model that instruct the extraction of the features, output of the response conversation that mirrors the subject, and generation of the feature prompts; 3. The information processing method according to claim 1.

7. The system prompt includes rules for extracting the features and instructions for asking the subject about the degree of similarity between the input conversation and the response conversation. The information processing method according to claim 6.

8. The system prompt includes the predetermined condition and an instruction to output the characteristic prompt when the predetermined condition is met. The information processing method according to claim 6.

9. The system prompts include examples of questions to ask the subject and a format for outputting the characteristics. The information processing method according to claim 6.

10. acquiring a feature prompt output based on a feature of the object extracted based on a dialogue between the object and a language model when a predetermined condition is satisfied; Get the user's query, The acquired feature prompt and the query are provided to a language model, thereby outputting a sentence. A program that causes a computer to perform a process.

11. extracting features of the target based on a dialogue including an input conversation by the target and a response conversation by a language model; If it is determined that the extracted feature satisfies a predetermined condition, a feature prompt based on the feature of the target at the time when the predetermined condition is satisfied is output. A program that causes a computer to perform a process.

12. extracting features of the target based on a dialogue including an input conversation by the target and a response conversation by a language model; If it is determined that the extracted feature satisfies a predetermined condition, a feature prompt based on the feature of the target at the time when the predetermined condition is satisfied is output. Processing section An information processing device comprising:

Images