Role dialogue method based on large model, intelligent agent, equipment and storage medium
By reconstructing the thinking link of the dialogue process in the big model, and using decision-making and dialogue big models to generate replies that meet the needs and role characteristics of the object, it solves the problem of lack of depth and temperature in the existing technology, and improves the authenticity of the dialogue sense and diversity.
Patent Information
- Application Number
- CN202411060989.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-02
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-08-02
AI Technical Summary
The existing role dialogue technology has problems such as insufficient realism and lack of depth and temperature in role-playing and dialogue replies, and the large language model lacks logical thinking ability in complex situations, resulting in the dialogue replies not meeting the scenario.
By adding functional modules to the big model, reconstructing the thinking link of the dialogue process, using the decision model to process perception information and memory information to generate decision information, and combining the dialogue model and role feature information to generate reply that meets the object's needs and role characteristics. .
It improves the authenticity and diversity of the conversation, makes the conversation more in line with the needs of the object and the character characteristics, and enhances the vividness and flexibility of the conversation.
Smart Images

Figure CN119047578A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of computer technology, and in particular to the field of chatbot technology, large model technology, and intelligent agent technology. Background Art
[0002] Large language models can handle a variety of natural language tasks, such as text classification, question-answering, and dialogue. The capabilities of large language models are gradually improving, and the multi-round dialogue capabilities based on large language models can realize role dialogue. Large language models have basic dialogue capabilities and role interpretation capabilities, but the roles are relatively fixed, and the chat content of the role replies is not particularly satisfactory. Summary of the invention
[0003] The present disclosure provides a role dialogue method, intelligent agent, device and storage medium based on a large model.
[0004] According to one aspect of the present disclosure, a character dialogue method based on a large model is provided, comprising:
[0005] The perception information and memory information are processed using a decision-making model to obtain decision information of the dialogue; wherein the perception information is obtained through perception processing based on the input information of the dialogue, and the memory information is obtained through information processing based on the object initiating the dialogue;
[0006] The decision information and role characteristic information of the dialogue are processed using the dialogue model to generate the response information of the dialogue.
[0007] According to another aspect of the present disclosure, there is provided an intelligent agent, comprising:
[0008] A decision system, used to process the perception information and memory information using a large decision model to obtain decision information of the dialogue; wherein the perception information is obtained through perception processing based on the input information of the dialogue, and the memory information is obtained through information processing based on the object initiating the dialogue;
[0009] The dialogue system is used to process the decision information and role characteristic information of the dialogue using the dialogue model to generate response information of the dialogue.
[0010] According to another aspect of the present disclosure, there is provided an electronic device, comprising:
[0011] at least one processor; and
[0012] a memory communicatively connected to the at least one processor; wherein,
[0013] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute any method in the embodiments of the present disclosure.
[0014] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute any method according to the embodiments of the present disclosure.
[0015] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, which implements any method according to the embodiments of the present disclosure when executed by a processor.
[0016] According to the present disclosure, a decision-making big model can be used to generate decision information based on perception information and memory information, and then a dialogue big model can be used to obtain a dialogue response that can meet the needs of the initiating object and is more in line with the characteristics of the role based on the role information in the decision information and memory information.
[0017] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.
[0019] Figure 1 is a flowchart of a large model-based role dialogue method according to an embodiment of the present disclosure;
[0020] Figure 2 is a flowchart of a large model-based character dialogue method according to another embodiment of the present disclosure;
[0021] Figure 3 is a flowchart of a large model-based character dialogue method according to another embodiment of the present disclosure;
[0022] Figure 4 A schematic diagram of an artificial intelligence agent according to an embodiment of the present disclosure;
[0023] Figure 5 A specific example diagram of the decision-making process according to an embodiment of the present disclosure;
[0024] Figure 6 is a schematic diagram of the structure of an intelligent agent according to an embodiment of the present disclosure;
[0025] Figure 7 is a schematic diagram of the structure of an intelligent agent according to another embodiment of the present disclosure;
[0026] Figure 8 is a block diagram of an electronic device for implementing an embodiment of the present disclosure. DETAILED DESCRIPTION
[0027] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.
[0028] Methods such as prompt engineering and thought chain alignment can improve the character consistency of large models and enhance the model's basic dialogue and character interpretation capabilities. Some early character dialogues were based on preset storylines, in which the interactive nodes and dialogue content were set in advance by the producer, and the user did not actually interact with the system.
[0029] Examples of existing technical solutions for character dialogue include:
[0030] (1) Role-playing products based on language models use prompt engineering to allow language models to perform role-playing. This technical solution is because the large language model has read many stories during training, coupled with the "intelligent" behavior exhibited by the larger language model. Users have found that the model can often demonstrate certain role-playing and conversational abilities under such prompts. However, the use of such prompts is highly dependent on the original memory of the large language model. If the large language model's memory of the role itself is vague, it will not be able to imitate a specific role. The definition of using prompts to restrict role-related knowledge is also vague, and it cannot effectively prevent the generation of the large language model's hallucination effect. Using prompts to restrict, the chatbot's conversational style will be greatly affected by the language model. Adjusting the prompt may alleviate this problem, but the prompt must be adjusted very finely for each specific role. These shortcomings obviously limit the use of this type of role-playing chatbot.
[0031] (2) Using the rich knowledge of the big model, fine-tune the character's dialogue ability into the model. This technical solution. Since the big model currently answers questions more as an assistant, its logical thinking ability in complex situations is not particularly satisfactory, and the dialogue often lacks practical meaning, requiring users to find topics. According to the theory that "the generalization ability of the big model is essentially a compression of the training data", after training on the dialogue data, the model learns the response terms of the character dialogue, but does not learn the motivation and thinking method in the speech. It is easy to give responses that do not fit the scene, and it cannot be as deep and warm as a real-life chat.
[0032] In summary, it is necessary to migrate the model dialogue style without degrading the basic capabilities, and to improve the realism of the dialogue experience of the dialogue model. In order to improve the role-playing ability of the model and to increase the dialogue time between the user and the model, the embodiment of the present disclosure designs the role dialogue model based on the paradigm of artificial intelligence (AI) agent (also called intelligent body), adds functional modules to the large language model, reconstructs the thinking link of the dialogue process, and enables the large model to dialogue like a "human".
[0033] Figure 1 1 is a flowchart of a large model-based role dialogue method according to an embodiment of the present disclosure. The method may include:
[0034] S101, using a large decision model to process the perception information and memory information to obtain decision information of the dialogue; wherein the perception information is obtained by perceiving and processing the input information of the dialogue, and the memory information is obtained by processing the information of the dialogue initiator;
[0035] S102: Use the dialogue macro model to process the decision information and role feature information of the dialogue to generate dialogue response information.
[0036] In the disclosed embodiment, the role dialogue method based on the large model can be executed by an intelligent agent, which can also be called an agent or a chat robot, etc. The large model can include a large language model (LLM).
[0037] In some examples, the intelligent agent may include a perception system, a memory system, a decision system, and a dialogue system. The perception system may include a perception large model, and the perception large model may be obtained by training an initial large model using a training sample of the perception large model. The input information of the dialogue is input into the perception large model of the perception system for perception, and perception information may be obtained. The input information of the dialogue may include one or more information such as text, voice, image, video, etc. input by the dialogue initiator. The dialogue initiator may include a user and / or a device used by the user, etc.
[0038] In the disclosed embodiment, the memory system may store one or more of the memory information of an object, the memory information of a role, the context of a conversation, etc. The name, identifier, etc. of the conversation initiating object is input into the memory system, and the memory information of the conversation initiating object can be searched in the memory system. The input information and / or perception information of the conversation is input into the memory system, and the conversation context corresponding to the input information and / or perception information of the conversation can be searched in the memory system. The name, identifier, etc. of a role is input into the memory system, and the memory information of the role (which may be referred to as role characteristic information) can be searched in the memory system.
[0039] In the disclosed embodiment, the decision system may include a decision big model, and the decision big model may be obtained by training the initial big model using the training samples of the decision big model. The perception information output by the perception big model and the memory information output by the memory system are input into the decision system, and the decision big model of the decision system may be used to make a decision based on the perception information and the memory information to obtain the decision information of the dialogue.
[0040] In the disclosed embodiment, the dialogue system may include a dialogue model, which may be obtained by training an initial model using training samples of the dialogue model. The decision information of the dialogue output by the decision model and the memory information of the role output by the memory system are input into the dialogue system, and the dialogue model of the dialogue system may be used to generate reply information for the input information of the dialogue.
[0041] According to the embodiments of the present disclosure, a decision-making big model can be used to generate decision information based on perception information and memory information, and then a dialogue big model can be used to obtain a dialogue response that can meet the needs of the initiating object and is more in line with the characteristics of the role based on the role information in the decision information and memory information.
[0042] In one implementation, the perception information includes one or more of emotion information, intention information, and dialogue state information of the dialogue initiator.
[0043] In the disclosed embodiment, the emotion information may indicate the emotion state of the conversation initiator, such as depression, excitement, calmness, etc. Emotion information may also be referred to as affective information. Intent information may indicate the conversation initiator's expectations, inclinations, or desires regarding the content of the conversation, such as the expectation that the topic of the chat is related to the weather, the expectation to learn about animation, etc. The conversation state information may be analyzed based on the emotion information and the intent information. The conversation state information may indicate the current state of the conversation, such as the state at the beginning of the conversation, the state at the end of the conversation, the state at the need to switch topics, etc.
[0044] According to the embodiments of the present disclosure, the needs of the object initiating the conversation can be better understood based on the emotional information and intention information of the object initiating the conversation, and more accurate conversation status information can be obtained to facilitate the large model to make more reasonable decisions based on emotional information, intention information, conversation status information, etc., and obtain a response that better meets the needs of the object.
[0045] In one embodiment, the memory information includes long-term memory and short-term memory, wherein the long-term memory includes one or more of object characteristics, interaction history, and character characteristic information; and the short-term memory includes one or more of conversation context, emotion information, and intention information.
[0046] The memory system can include long-term memory and short-term memory. Long-term memory can be used to store the interaction history between the agent and the user, the user's object features, and the role knowledge base. Short-term memory can be used to store the emotional information, intention information, and dialogue status output by the perception system.
[0047] In the disclosed embodiment, the memory system can use long-term memory and short-term memory to store the information that the agent needs to use during the dialogue process. The long-term memory can store the characteristics of the object that has a dialogue with the agent. The long-term memory can also store the interaction history between the object and the agent, such as the historical input information and historical reply information of the dialogue. The long-term memory can also store role feature information, such as the knowledge base of the role (including the knowledge information of different roles), the character of the role, the character of the role, the language habits of the role, etc. The short-term memory can store the context information of the dialogue, and the content of the previous dialogues of the current dialogue and the content of the previous topics can be found according to the identifier of the current dialogue. The short-term memory can also store the perception information output by the perception system, such as emotional information, intention information, etc. The long-term memory and short-term memory of the same object can have a corresponding relationship. For example, based on the identifier of an object, the characteristics, interaction history, role feature information, dialogue context, emotional information and intention information of the object can be found.
[0048] In the disclosed embodiment, the memory system can use a preset prompt word template to splice one or more of the context information, emotional information, and intention information to obtain a short-term memory prompt word. One or more of the conversation context, emotional information, and intention information can be stored in the form of a short-term memory prompt word to obtain the short-term memory of the memory system. Furthermore, the memory system can input the short-term memory prompt word of a certain object into the decision-making system to participate in the decision-making.
[0049] According to the embodiments of the present disclosure, long-term memory and short-term memory related to the dialogue can be stored in the memory system, and the rich and diverse memory information provides a basis for the decision-making system to generate dialogues that meet the needs of the object and are consistent with the role.
[0050] Figure 2 1 is a flow chart of a method for character dialogue based on a large model according to another embodiment of the present disclosure, and the method may include one or more features of the method for character dialogue based on a large model. In one embodiment, S101 uses a decision large model to process perception information and memory information to obtain decision information of the dialogue, including:
[0051] S201, using the decision-making big model, executing the thinking process based on the dialogue state information, and obtaining the thinking result;
[0052] S202, obtaining an action direction based on the thinking result and one or more execution action processes of the dialogue context, emotion information, intention information, object characteristics, and interaction history;
[0053] S203, performing an observation process based on the action direction to obtain an observation result;
[0054] S204: Obtain decision information of the dialogue based on the action direction and / or the observation result.
[0055] For example, the character's thinking logic includes one or more of thinking results, action results, and observation results; the decision information may include one or more of content decision information, motivation decision information, emotion decision information, and tone decision information.
[0056] In the disclosed embodiment, the decision system may include a decision model, and the decision model may generate decision information of the dialogue based on the thought chain. The thought chain of the decision system may include meta-motivation, action mode, thinking mode, etc. The meta-motivation of the thought chain may be to keep the dialogue going; the action mode of the thought chain may adopt the mode of reason + action; the thinking mode of the thought chain may adopt the mode of thinking-action-observation.
[0057] In the disclosed embodiment, the decision-making model can make decisions based on the long-term memory and short-term memory sent by the memory system and the perception information output by the perception system to obtain the decision information of the dialogue. The decision information of the dialogue can also be called the optimal execution strategy, such as whether to change the topic, response strategy, dialogue motivation, dialogue emotion, dialogue tone and reply content.
[0058] In the disclosed embodiment, the decision system may first execute the thinking process in the thinking-action-observation mode according to the dialogue state information output by the perception system. For example, the decision model may execute the thinking process based on the dialogue state information to determine whether to continue the current topic or to change a new topic. After the thinking process is completed, the thinking result may be obtained, such as continuing the current topic or changing a new topic.
[0059] In the disclosed embodiment, the decision system can execute the action process in the thinking-action-observation mode according to the thinking results, dialogue context, emotional information, intention information, object characteristics, interaction history, etc., and decide the action direction to be taken. The action direction can also become a response strategy. For example, if the thinking result is to continue the current topic, the emotional information is that the mood is not good, and the action direction may include that the character needs to be more active and enthusiastic. For another example, if the thinking result is to change the topic, and the interaction history includes topics related to game A that have been asked, the action direction may include topics related to game A.
[0060] In the disclosed embodiment, the decision system can perform the observation process in the thinking-action-observation mode according to the action direction to obtain the observation result. The observation result may include the reaction that the decision system may encounter after taking action based on the action direction, the operation details that need to be paid attention to, etc. For example, after changing the topic related to game A, observe whether the conversation initiator has further response to the topic related to game A.
[0061] In the disclosed embodiment, the decision system can use the decision model to generate one or more decision information including content decision information, motivation decision information, emotion decision information and tone decision information according to the action direction and / or observation results. Content decision information may include text, voice, image or video content of the expected reply. For example, "Are you in a bad mood? Do you need to open game A to relax?" Emotion decision information may include the emotions of the character and / or the object. For example, the low mood of the object corresponds to the emotional enthusiasm of the character. Tone decision information may include the tone of the character's reply. For example, a gentle expected reply is adopted for the low mood of the object.
[0062] According to the embodiments of the present disclosure, through the thinking process, the action process and the observation process, a strategy for dialogue response that meets the object needs and role characteristics can be generated based on one or more of the dialogue state information, dialogue context, emotional information, intention information, object characteristics, interaction history, etc., which is conducive to generating more vivid, more flexible, more in line with the object needs and more character-specific responses in the dialogue system.
[0063] Figure 3 1 is a flow chart of a character dialogue method based on a big model according to another embodiment of the present disclosure, and the method may include one or more features of the character dialogue method based on a big model. In one embodiment, S102 uses the dialogue big model to process the decision information and character feature information of the dialogue to generate dialogue reply information, including:
[0064] S301. The decision information of the dialogue output by the decision macromodel and the target role characteristic information obtained from the memory system are input into the dialogue macromodel to generate reply information that conforms to the characteristic information of the target role and the dialogue status information.
[0065] In the disclosed embodiment, the dialogue system can receive decision information from the decision system and target role feature information from the memory system, and input this information into the dialogue model to generate feature information that conforms to the target role and reply information that conforms to the current dialogue scene. The current dialogue scene can be determined based on the dialogue state information. For example, if the dialogue state information is the initial dialogue state, the current dialogue scene is the start scene. If the dialogue state information includes a state where a topic needs to be switched, the current dialogue scene is a topic switching scene.
[0066] In the disclosed embodiment, before a dialogue with an intelligent agent, the role of the intelligent agent can be pre-selected by the dialogue initiator object, or the role of the intelligent agent can be determined based on the input information analysis. The role of the intelligent agent can include anime characters, film and television characters, literary characters, historical characters, scientific and technological characters, etc. The memory system can pre-store long-term memory and / or short-term memory of various characters. According to the role of the intelligent agent, the target role feature information can be retrieved from the long-term memory. After retrieving the target role feature information, the memory system can send the target role feature information to the dialogue system.
[0067] According to the embodiments of the present disclosure, a dialogue macro model can be used to generate responses that are more vivid, more flexible, more in line with the needs of the object, and more character-specific, based on decision information that meets the needs of the object and target role characteristic information.
[0068] In one embodiment, the training samples of the decision model include one or more of the following:
[0069] Perceive sample information;
[0070] Memorize sample information;
[0071] Using the large language model to infer the role's thinking logic and / or decision information to be processed based on the perception sample information and / or memory sample information;
[0072] Annotate the character thinking logic to be processed to obtain the annotated character thinking logic;
[0073] The decision information to be processed is labeled to obtain labeled decision information.
[0074] In the disclosed embodiment, a large language model can be used to reversely deduce the chain of thought based on the perceived sample information and / or the memory sample information to obtain the character thinking logic to be processed, such as derived observation results, derived action directions, and derived thinking results. Then, the derived observation results, derived action directions, and derived thinking results are annotated using machine annotation or manual annotation. Based on the derived observation results, derived action directions, and derived thinking results, the large language model can further reversely deduce to obtain the decision information to be processed, such as one or more of derived content decision information, derived motivation decision information, derived emotion decision information, and derived tone decision information.
[0075] During the training process of the decision-making big model, the perception sample information and / or memory sample information in the training samples can be used as the input features of the model, and the annotated role thinking logic and / or annotated decision information in the training samples can be used as the expected input features of the model. The initial decision-making big model can be fine-tuned to obtain the trained decision-making big model.
[0076] According to the embodiment of the present disclosure, the decision-making large model trained using the training samples of the decision-making large model can obtain decision information that is more in line with the perception information and memory information, and thus obtain response information that is more in line with the role characteristics and object needs.
[0077] In one embodiment, the training samples of the dialogue model include: based on the role dialogue data set, obtaining role feature sample information and role dialogue sample information, and the role feature sample information includes one or more of role knowledge, role background, role personality, role character and role language habits.
[0078] In the disclosed embodiment, the role dialogue data set can be understood as a role data source. For example, the role dialogue data set may include literary works, film and television works, historical documents, etc. By using a large model to process the role data source, role feature sample information such as role knowledge, role background, role personality, role character, and role language habits can be extracted from the role data source. These role feature sample information can be stored in the memory system or stored separately in the training sample set. Among them, role knowledge can include content that the role can understand. For example, the knowledge of the physicist role includes physical knowledge. The knowledge of the doctor role includes medical knowledge. For another example, the role background may include the role's life background, learning background, work background, etc. For another example, the role personality may include introversion, extroversion, etc. The role character may include bravery, intelligence, wit, loyalty, etc. Role personality and role character may also have the same or similar meanings. Role personality and / or role character may also be referred to as role personality. For another example, the role language habits may include the role's idioms, vividness, directness, dullness, etc.
[0079] By using the big model to process the character data source, character dialogue sample information can be extracted from the character data source. The character dialogue sample information can include dialogues that appear in literary works, film and television works, historical documents, etc. For example, a conversation involving character R in a certain literary work includes question Q1 and reply A1, question Q2 and reply A2, question Q3 and reply A3. The character feature sample information and the questions in the character dialogue sample information can be used as the input features of the model, and the replies in the character dialogue sample information can be used as the output features of the model. The initial dialogue big model can be fine-tuned to obtain a trained dialogue big model. The trained dialogue big model has the ability to play different roles and reply to dialogues.
[0080] According to the embodiment of the present disclosure, the dialogue big model trained using the training samples of the dialogue big model can generate reply information that is more in line with the role characteristics and object needs, so the reply information is more realistic and more flexible.
[0081] In one embodiment, the training samples of the perceptual large model include one or more of the following:
[0082] Enter sample information;
[0083] The perception information to be processed is obtained based on the perception sample information using the large language model;
[0084] The perception information to be processed is labeled to obtain labeled perception information.
[0085] In the disclosed embodiment, information such as text, voice, image or video can be input into a large language model for reverse deduction to obtain derived emotion information, derived intention information and derived dialog state perception information to be processed. Then, the derived emotion information, derived intention information and derived dialog state perception information to be processed are annotated by machine or manual annotation to obtain annotated emotion information, annotated intention information and annotated dialog state and other annotated perception information.
[0086] During the training process, the input sample information can be used as the input features of the initial perceptual big model, and the labeled perceptual information can be used as the output features of the initial perceptual big model. The initial perceptual big model can be fine-tuned to obtain a trained perceptual big model.
[0087] According to the embodiments of the present disclosure, the perception big model obtained by training using the training samples of the perception big model can perceive the perception information that is consistent with the emotions and intentions of the object, and then obtain the reply information of the conversation that is consistent with the emotions and intentions of the object.
[0088] In order to allow the dialogue initiator to truly experience the same role dialogue experience as human dialogue, the disclosed embodiment can build an artificial intelligence agent (AI agent) based on the dialogue ability of a large language model, with independent perception systems, memory systems and decision systems as auxiliary. When conducting a role dialogue, the perception system can analyze the current dialogue state based on the perceived emotions and desires of the dialogue initiator, and the decision system can plan the direction of dialogue development based on the results of the perception system and the role feature information in the memory system, the dialogue initiator's portrait and interaction history. This not only improves the model's understanding of the dialogue initiator, but also greatly improves the diversity of the model's responses, making the dialogue development have infinite possibilities.
[0089] The schematic diagram of the artificial intelligence agent (AI agent) of the present disclosure embodiment is as follows Figure 4 As shown, it can include four parts, including the perception system, memory system, decision system and dialogue model (or dialogue system). The artificial intelligence agent can conduct role dialogue according to the following steps: the first step is that the perception system perceives the emotions, intentions and dialogue status of the dialogue initiator based on the input of the dialogue initiator and the current dialogue; the second step is that the decision system comprehensively analyzes and decides on the optimal execution strategy (also called decision information) based on the results of the perception system and the object characteristics, interaction history (interaction history), role knowledge base and other data in the memory system. The third step is that the dialogue model, under the guidance of the optimal execution strategy of the decision system, based on the dialogue ability of the large model and the role-playing ability learned during the fine-tuning process, gives a reply with clear meaning, distinctive role style, and can drive the dialogue forward. The implementation of these four parts will use different trained large models or deep learning models. The details are as follows:
[0090] 1. Perception system: The perception system is mainly responsible for processing and understanding the input of the dialogue initiator. It perceives the situation of the dialogue initiator based on the input of the dialogue initiator and the current dialogue history, including the emotions, intentions, and dialogue status of the dialogue initiator. The significance of the perception system is equivalent to observing the words and expressions in real conversations. This ability is very important, as it determines whether the dialogue initiator has the desire to continue the conversation with the model. Since information such as emotions and desires belongs to relatively shallow semantics, we can use a smaller large model here to save computing costs and reduce response time.
[0091] 2. Memory system: The memory system is an important part of the agent solution. Memory mainly solves the problem of professional knowledge transmission, which can enable the agent to have long-term and short-term memory and make the agent more intelligent. Long-term memory mainly includes the characteristics of the dialogue initiator, interaction history (interaction history) and role-related knowledge graph (role knowledge base), etc. Short-term memory mainly includes the context dialogue of the dialogue initiator and the desire (intention), emotion (emotion) of the dialogue initiator and other knowledge that is strongly related to the current dialogue. The long-term memory part is mainly implemented in the form of vector database, because it is embedded (Embedding) storage, which can be more conducive to information compression and efficient retrieval under the premise of large-scale data. Short-term memory is generally transmitted through prompt words (Prompt). After being obtained by the memory system, it is transmitted to the decision-making system through prompt words. At the same time, this information can be used in the decision-making system to help the system make better decisions.
[0092] 3. Decision-making system: The decision-making system uses the sustainability of the conversation as a meta-motivation, and uses reasons + actions to solve problems in a thinking-action-observation mode. For example, the thinking process can consider whether to continue the current topic or find a new topic; the action process can determine the action direction such as what strategy the character responds with, what motivation the character has for the conversation, and what emotion the character has for the conversation; for each action direction of the action process, the corresponding observation results can be given in the observation link. The decision-making system can eventually output what tone the character should use to say what. Based on the results of the perception system and the data in the memory system, the decision-making system analyzes the emotions, intentions, historical behaviors and other information of the initiator of the conversation, and comprehensively considers and decides on the optimal execution strategy in terms of content, motivation, emotion, tone and other aspects.
[0093] For example, Figure 5 As shown. In the historical dialogue and input information, the object's input information 1 includes "Who are you?". The character's (such as AI) reply 1 includes "Don't you know me? I'm Y!". The object's input information 2 includes "Oh". The character's reply 2 includes "Have you seen any new movies recently?". The object's input information 3 includes "No". The decision-making system uses reason + action mode to execute the thinking-action-observation process.
[0094] For example, the thinking process could include whether the character should continue with the current topic or find a new topic (thinking):
[0095] 1. The current topic is that the character asks the subject if he has seen a new movie, and the subject says no;
[0096] 2. There is no more content to be advanced in the current topic, and a new topic needs to be found;
[0097] For example, the action process and observation process can include: What strategies does the character need to respond (action):
[0098] 1. The subject’s characteristics show that the subject likes to play games. You can start by asking the subject whether he has played any new games recently.
[0099] Observation: Subject did not respond to topic of movie, and likely would not respond to game;
[0100] 2. The interaction history with the subject shows that the subject is in a low mood. You can start from this perspective and pay attention to the subject's emotional state;
[0101] Observation: Don’t touch the target’s sad spot;
[0102] 3. The subject is currently in a low mood, and the character can be more enthusiastic and positive;
[0103] Observation: Be careful not to put pressure on the subject;
[0104] For example, decision information may include: What does the character need to say? In what tone?
[0105] Tone: Gentle, caring, and enthusiastic;
[0106] Content: "You don't seem to be in a good mood. Is it because you are too tired recently? If you want to talk to someone, I am always here. If you don't want to talk about this, I can still play games with you!"
[0107] 4. Dialogue system: The dialogue model is mainly responsible for generating responses to the initiator of the dialogue according to the optimal execution strategy of the decision-making system. The implementation of the dialogue model is based on the fine-tuning of the large language model. The model is trained to learn the ability of dialogue and role-playing to generate responses with clear meaning, distinctive role style, and the ability to drive the dialogue forward.
[0108] The construction of training samples of the large models of the perception system, decision-making system and dialogue system of the embodiment of the present disclosure, as well as the construction of long-term and short-term memory data of the memory system are shown in Table 1, and are as follows:
[0109] Table 1
[0110]
[0111] The following are data examples for several systems:
[0112] 1. Perception system: Emotions can include the thoughts, feelings, and psychological states of the initiator of the conversation. Emotions and desires of the conversation. The disclosure allows users to input and add annotations. When annotating, the context will be used as prior knowledge, and annotations can be made according to the conversation context. The annotated content can include the emotions of the initiator of the conversation, the desires of the initiator of the conversation, and the state of the conversation.
[0113] 2. Memory system: Characterize the characteristics of the dialogue initiator. The characteristics of the dialogue initiator are divided into basic characteristics and time-effect characteristics. Basic characteristics include characteristics formed by the long-term behavior of the dialogue initiator. Time-effect characteristics can be extracted based on the behavior of the dialogue initiator in the past period of time, and can represent the behavioral characteristics of the dialogue initiator in a period of time.
[0114] 3. Decision-making system: Use the big model to reversely deduce the dialogue data to obtain the psychological thinking logic of the character when answering, and then make annotations that conform to the general thinking logic and finally modify it. Since the thinking process of the dialogue is diverse, specific restrictions can be made on the expression of the thinking paradigm. After trial standard adjustment, this disclosure enriches the guiding principles of the big model reverse deduction to construct a more comprehensive and accurate analysis dimension, and present the final strategy in the form of a thinking chain.
[0115] 4. Dialogue system: The character dialogue dataset can be obtained based on novels and scripts, and is constructed in combination with high-quality interesting dialogues obtained from the Internet. The character dialogue dataset is mainly constructed around the following three dimensions: the character's knowledge and background, the character's personality or character, and the character's language habits. Each character has its own background, so in the construction of the chatbot, the present disclosure enables the chatbot to understand the setting of the corresponding story.
[0116] Personality or character: The personality and character setting of a character is a very important part of animation, film and television, and even game works. The setting of personality and character needs to be consistent throughout the work. When creating some literary works, the personality setting of the character is even defined first, and then the subsequent writing work is carried out. The present disclosure trains the personality and character reflected by the agent to be consistent with the original setting of the work.
[0117] Language habits: Language habits are the easiest to imitate by language models. If appropriate examples are given in the text (context), the language model can imitate and output. The dialogue model disclosed in this disclosure can enable fans of literary and film works to "reproduce" classic scenes of literary and film works when interacting with the products disclosed in this disclosure, so that fans of these works can have a better experience.
[0118] The large models of the perception system, decision system and dialogue system of the disclosed embodiment can use the constructed data to train the model to improve the perception system's emotional desire perception ability, planning ability, dialogue model role dialogue ability, etc. The memory system can store the information that the agent needs to use when conducting role dialogue. The specific training methods of each model and the use of the memory system are as follows:
[0119] 1. Perception system: The perception model can be fine-tuned based on a small-scale model, such as a 7B to 14B language model, and the prompt words are used to limit the output of the perception model to formatted output emotions, desires, and dialogue status.
[0120] 2. Memory system: By recommending knowledge graph inference algorithms and vector representation methods of graph models, we build a large-scale heterogeneous graph data system for the portraits of conversation initiators, and use graph vector representation to calculate similarities to match the personalized content of a single conversation initiator. Use the big model to extract key information from the conversation history of the conversation initiator, and build a retrieval knowledge base based on the importance and chronological order of the key information. When replying, the model builds a more appropriate reply based on the conversation history and the current conversation topic, combining the big model with knowledge graph technology and applying it to role conversations. A breakthrough is achieved in the long-term memory and specificity of the model to enhance the personalization of role conversations. The conversation model predicts the next reply, and can predict the next reply based on the existing multiple rounds of historical conversations.
[0121] 3. Dialogue system: The dialogue model can be fine-tuned based on the large language model, and it pays more attention to the improvement of multi-round dialogue capabilities on the basis of general knowledge capabilities. Based on the large language model, the self-focused transformer (SFT) and differential policy optimization (DPO) model training are carried out to continuously optimize the dialogue model capabilities, enhance the dialogue model's dialogue capabilities in many aspects such as the character's background, personality setting, dialogue interactivity, and emotional intelligence, and improve the generalization of the dialogue model and the quality rate of responses.
[0122] The character dialogue model needs to have some basic capabilities, including the response content is consistent with the scene, consistent with the character setting, and provides similar emotional values. In character dialogue, distinct personality is crucial. Each character should show unique personality traits and language style, so that the dialogue initiator can clearly distinguish the dialogue between different characters, thereby increasing the interest of the dialogue. At the same time, emotional expression is also indispensable. The dialogue should truly show the unique emotions of the characters, such as love, hate, friendship and betrayal. By allowing the dialogue initiator to deeply feel the emotional fluctuations of the characters during the dialogue, the dialogue initiator can resonate with the dialogue initiator. In order to maintain the authenticity and natural fluency of the dialogue, the content of the dialogue should be consistent with the character setting and situation, fit the interests and desires of the dialogue initiator, promote natural and smooth communication and interaction between the dialogue initiator and the model, and avoid the model's responses being too stiff or unreal. In addition, the dialogue model can use more vivid language, combined with some interesting and stalky expressions, to increase the interest of the dialogue initiator and make the dialogue more vivid and interesting.
[0123] 4. Decision-making system: The decision-making model can be trained or fine-tuned with a model that has the same or similar base as the dialogue model as the initial model. The disclosed embodiment can provide rich dialogue scene corpus training to help the decision-making model understand the dialogue paradigm and infer the emotional expression of the dialogue party.
[0124] The disclosed embodiment constructs an artificial intelligence agent (AI Agent) to provide a perception system, a memory system, and a decision-making system for the dialogue model obtained by fine-tuning the large model, thereby significantly improving the shortcomings of the dialogue model such as poor logic, lack of knowledge, and insufficient dialogue analysis capabilities, which can bring great help and advantages to the product. The role dialogue achieved through the present disclosure can bring a more vivid and realistic dialogue experience to the dialogue initiator. The perception system, memory system, and decision-making system possessed by the artificial intelligence agent (AI Agent) can make its own dialogue model more intelligent and flexible, so as to better understand the needs of the dialogue initiator, remember the preferences of the dialogue initiator, and make more reasonable responses and decisions. This personalized dialogue communication method can not only increase the participation and interactivity of the dialogue initiator, but also greatly enhance the attractiveness and competitiveness of the product.
[0125] The role dialogue realized through the present disclosure will provide a more intelligent and personalized document editing experience for the dialogue initiator, while also opening up a broader development space for the product and creating more value and convenience for the dialogue initiator.
[0126] Figure 6 is a schematic diagram of the structure of an agent according to an embodiment of the present disclosure. In one implementation, the agent may include:
[0127] The decision system 601 is used to process the perception information and the memory information using the decision model to obtain the decision information of the dialogue; wherein the perception information is obtained by the perception processing based on the input information of the dialogue, and the memory information is obtained by the information processing based on the dialogue initiator;
[0128] The dialogue system 602 is used to process the decision information and role characteristic information of the dialogue using the dialogue macro model to generate dialogue reply information.
[0129] Figure 7 is a schematic diagram of the structure of an agent according to another embodiment of the present disclosure, and the agent may include one or more features of the above-mentioned agent. In one embodiment, the agent further includes:
[0130] The perception system 603 is used to obtain the perception information based on the input information of the dialogue through perception processing, and the perception information includes one or more of the emotion information, intention information and dialogue state information of the dialogue initiator.
[0131] In one embodiment, the intelligent agent also includes: a memory system 604 for extracting and storing memory information, which includes long-term memory and short-term memory, wherein the long-term memory includes one or more of object characteristics, interaction history, and character characteristic information; and the short-term memory includes one or more of dialogue context, emotional information, and intention information.
[0132] In one implementation, the decision system 601 includes:
[0133] A thinking module 701 is used to use the decision-making model to execute a thinking process based on the dialogue state information to obtain a thinking result;
[0134] An action module 702 is used to execute an action process based on the thinking result and one or more of the conversation context, emotion information, intention information, object characteristics, and interaction history to obtain an action direction;
[0135] An observation module 703 is used to perform an observation process based on the action direction to obtain an observation result;
[0136] The decision module 704 is used to obtain decision information of the dialogue based on the action direction and / or the observation result.
[0137] In one embodiment, the dialogue system 602 is also used to input the decision information of the dialogue output by the decision model and the target role feature information obtained from the memory system into the dialogue model to generate reply information that conforms to the feature information of the target role and the dialogue status information.
[0138] In one embodiment, the training samples of the decision model include one or more of the following:
[0139] Perceive sample information;
[0140] Memorize sample information;
[0141] Using the large language model to infer the role's thinking logic and / or decision information to be processed based on the perception sample information and / or memory sample information;
[0142] Annotate the character thinking logic to be processed to obtain the annotated character thinking logic;
[0143] The decision information to be processed is labeled to obtain labeled decision information.
[0144] In one embodiment, the training samples of the large dialogue model include:
[0145] Based on the role dialogue data set, role feature sample information and role dialogue sample information are obtained, where the role feature sample information includes one or more of role knowledge, role background, role personality, role character, and role language habits.
[0146] In one embodiment, the training samples of the perceptual large model include one or more of the following:
[0147] Enter sample information;
[0148] The perception information to be processed is obtained based on the perception sample information using the large language model;
[0149] The perception information to be processed is labeled to obtain labeled perception information.
[0150] For the description of specific functions and examples of various systems and modules of the intelligent body in the embodiments of the present disclosure, please refer to the relevant description of the corresponding steps in the above method embodiments, which will not be repeated here.
[0151] In the technical solution disclosed herein, the acquisition, storage and application of user personal information involved are in compliance with the provisions of relevant laws and regulations, are authorized by the user, and do not violate public order and good morals.
[0152] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.
[0153] Figure 8A schematic block diagram of an example electronic device 800 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.
[0154] like Figure 8 As shown, the device 800 includes a computing unit 801, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.
[0155] A number of components in the device 800 are connected to the I / O interface 805, including: an input unit 806, such as a keyboard, a mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a disk, an optical disk, etc.; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 809 allows the device 800 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.
[0156] The computing unit 801 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 801 performs the various methods and processes described above, such as a role dialogue method based on a large model. For example, in some embodiments, the role dialogue method based on a large model may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 808. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 800 via ROM 802 and / or communication unit 809. When the computer program is loaded into RAM 803 and executed by the computing unit 801, one or more steps of the role dialogue method based on the large model described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to perform a role dialogue method based on a large model in any other appropriate manner (e.g., by means of firmware).
[0157] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.
[0158] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.
[0159] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
[0160] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).
[0161] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.
[0162] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.
[0163] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.
[0164] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the principles of the present disclosure shall be included in the protection scope of the present disclosure.
Claims
1. A role dialogue method based on a large model, comprising: The perception information and memory information are processed using a decision-making large model to obtain decision information of the dialogue; wherein the perception information is obtained by perceiving and processing the input information of the dialogue, and the memory information is obtained by processing the information of the dialogue initiator; The decision information and role characteristic information of the dialogue are processed using the dialogue macro model to generate dialogue response information.
2. The method according to claim 1, wherein: The perception information includes one or more of the emotion information, intention information and dialogue state information of the dialogue initiator.
3. The method according to claim 1, wherein: The memory information includes long-term memory and short-term memory. The long-term memory includes one or more of object characteristics, interaction history, and role characteristic information; the short-term memory includes one or more of dialogue context, emotion information, and intention information.
4. The method according to any one of claims 1 to 3, wherein: Use the decision-making model to process the perception information and memory information to obtain the decision information of the dialogue, including: Use the big decision model to execute the thinking process based on the dialogue state information and obtain the thinking results; Based on the thinking result and one or more execution action processes of the dialogue context, emotion information, intention information, object characteristics, and interaction history, an action direction is obtained; Performing an observation process based on the action direction to obtain an observation result; Based on the action direction and / or the observation result, decision information of the dialogue is obtained.
5. The method according to claim 1, wherein: The decision information and role feature information of the dialogue are processed using the dialogue macro model to generate dialogue reply information, including: The decision information of the dialogue output by the decision big model and the characteristic information of the target role obtained from the memory system are input into the dialogue big model to generate reply information that conforms to the characteristic information of the target role and the dialogue status information.
6. The method according to claim 1, wherein: The training samples of the decision-making model include one or more of the following: Perceive sample information; Memorize sample information; Using the large language model to infer the role's thinking logic and / or decision information to be processed based on the perception sample information and / or memory sample information; Annotate the character thinking logic to be processed to obtain the annotated character thinking logic; The decision information to be processed is labeled to obtain labeled decision information.
7. The method according to claim 1, wherein: The training samples of the large dialogue model include: Based on the role dialogue data set, role feature sample information and role dialogue sample information are obtained, wherein the role feature sample information includes one or more of role knowledge, role background, role personality, role character and role language habits.
8. The method according to claim 1, wherein: The training samples of the perceptual large model include one or more of the following: Enter sample information; The perception information to be processed is obtained based on the perception sample information using the large language model; The perception information to be processed is labeled to obtain labeled perception information.
9. An intelligent agent, comprising: A decision system, used to process the perception information and memory information using a large decision model to obtain decision information of the dialogue; wherein the perception information is obtained by perceiving and processing the input information of the dialogue, and the memory information is obtained by processing the information of the dialogue initiator; The dialogue system is used to process the decision information and role characteristic information of the dialogue using the dialogue macro model to generate dialogue reply information.
10. The intelligent agent according to claim 9, wherein: The intelligent agent also includes: The perception system is used to obtain the perception information through perception processing based on the input information of the dialogue, and the perception information includes one or more of the emotion information, intention information and dialogue state information of the dialogue initiator.
11. The intelligent agent according to claim 9, wherein: The intelligent agent also includes: A memory system is used to extract and store memory information, wherein the memory information includes long-term memory and short-term memory. The long-term memory includes one or more of object characteristics, interaction history, and character characteristic information; the short-term memory includes one or more of dialogue context, emotional information, and intention information.
12. An agent according to any one of claims 9 to 11, wherein: The decision-making system comprises: The thinking module is used to use the decision-making model to execute the thinking process based on the dialogue state information and obtain the thinking results; An action module, configured to execute an action process based on the thinking result and one or more of the conversation context, emotion information, intention information, object characteristics, and interaction history to obtain an action direction; An observation module, used to perform an observation process based on the action direction to obtain an observation result; A decision module is used to obtain decision information of the dialogue based on the action direction and / or the observation result.
13. The intelligent agent according to claim 9, wherein: The dialogue system is also used to input the decision information of the dialogue output by the decision model and the target role feature information obtained from the memory system into the dialogue model to generate reply information that conforms to the feature information of the target role and the dialogue status information.
14. The intelligent agent according to claim 9, wherein: The training samples of the decision-making model include one or more of the following: Perceive sample information; Memorize sample information; Using the large language model to infer the role's thinking logic and / or decision information to be processed based on the perception sample information and / or memory sample information; Annotate the character thinking logic to be processed to obtain the annotated character thinking logic; The decision information to be processed is labeled to obtain labeled decision information.
15. The intelligent agent according to claim 9, wherein: The training samples of the dialogue model include: based on the role dialogue data set, role feature sample information and role dialogue sample information are obtained, and the role feature sample information includes one or more of role knowledge, role background, role personality, role character and role language habits.
16. The intelligent agent according to claim 9, wherein: The training samples of the perceptual large model include one or more of the following: Enter sample information; The perception information to be processed is obtained based on the perception sample information using the large language model; The perception information to be processed is labeled to obtain labeled perception information.
17. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 8.
18. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-8.
19. A computer program product, comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 8.
Citation Information
Patent Citations
Implementation method and device for enhancing virtual digital representation in meta universe
CN117008718A
Dialogue sample generation method, chat conversation large model training method and related devices
CN117573946A
Intelligent question answering method and device based on large model, electronic equipment and storage medium
CN117648408A
Emotion support dialogue generation method, system and device based on thinking chain reasoning
CN117932041A
Interaction method and device, medium and electronic equipment
CN117959715A
Cited By
Method for enhancing role perception of large language model
CN119760095A
A method for enhancing role awareness of a large language model
CN119760095B
Intelligent question learning method and system based on large language model
CN120541193A
Dialogue method and device based on large model, equipment and medium
CN120561234A
Multi-agent cooperative emotion support dialogue generation method and system
CN120561255A
