Condition assessment method, device, equipment and medium

Abstract

The present invention provides a state evaluation method, device, equipment and medium, wherein the method includes: receiving a user's reply to the current round of evaluation questions; determining a dialogue jump strategy and the user's emotional intention in the scene based on the scene where the evaluation question is located, the reply, and the user's historical dialogue in the state evaluation; determining a target dialogue model from each candidate dialogue model based on the dialogue jump strategy, and generating the next round of evaluation questions based on the target dialogue model, and returning to receive the user's reply to the next round of evaluation questions until the interaction ends; performing state evaluation based on the user's emotional intention in the scene where the evaluation question is located in each round. The state evaluation method, device, equipment and medium provided by the present invention realize flexible switching of dialogue modes through a dialogue jump strategy, improve the natural fluency of the dialogue, and enable users to have a more intimate experience in interaction.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a state assessment method, device, equipment and medium. Background Art

[0002] Status assessment is crucial in areas such as mental health, education, and healthcare. It not only helps to identify potential psychological problems, but also improves service quality and enhances user experience. In modern society, accurate status assessment can help various fields better understand user needs and provide personalized support.

[0003] Existing state assessment technologies mainly rely on task-based dialogue and open-domain dialogue, but both have their own shortcomings. Although task-based dialogue is highly structured and can effectively collect information, it lacks emotional resonance and is difficult to meet the emotional needs of users; while open-domain dialogue is natural and smooth and can establish an emotional connection with users, it is difficult to focus on specific tasks, resulting in incomplete collection of assessment information and failure to complete the state assessment task. Summary of the invention

[0004] The present invention provides a state assessment method, device, equipment and medium, which are used to solve the defects in related technologies that task-based dialogues are difficult to meet emotional needs and open-domain dialogues are difficult to complete assessment tasks.

[0005] The present invention provides a state assessment method, comprising the following steps: Receive user responses to the current round of evaluation questions; Based on the scenario where the evaluation question is located, the reply, and the historical dialogue of the user in the status evaluation, determine the dialogue jump strategy and the emotional intention of the user in the scenario; Based on the dialogue jump strategy, determine the target dialogue model from the candidate dialogue models, generate the next round of evaluation questions based on the target dialogue model, and return to receive the user's reply to the next round of evaluation questions until the interaction ends; Based on the emotional intention of the user in the scenario of the evaluation questions in each round, a state evaluation is performed.

[0006] According to a state assessment method provided by the present invention, the method determines the dialogue jump strategy and the emotional intention of the user in the scenario based on the scenario where the assessment question is located, the reply, and the historical dialogue of the user in the state assessment, including: Generate input text based on the scenario where the assessment question is located, the reply, the historical dialogue of the user in the state assessment, and an input text template, wherein the input text template includes task description information of dialogue jump and emotion intention recognition; The input text is input into a large language model to obtain a dialogue jump strategy output by the large language model and the emotional intention of the user in the scenario.

[0007] According to a state evaluation method provided by the present invention, the input text template also includes strategy selection description information of the dialogue jump strategy.

[0008] According to a state assessment method provided by the present invention, each candidate dialogue model includes a chat dialogue model and a task-based dialogue model oriented to a state assessment task; The step of determining a target dialogue model from each candidate dialogue model based on the dialogue jump strategy includes: When the dialogue jump strategy is empathy response, determining the chat dialogue model as the target dialogue model; When the dialogue jump strategy is not the empathic response, the task-based dialogue model is determined to be the target dialogue model.

[0009] According to a state evaluation method provided by the present invention, when the dialogue jump strategy is not the empathic response, generating the next round of evaluation questions based on the target dialogue model includes: Based on the emotional intention in the scenario, updating the scenario state of each scenario used for the status evaluation; Determine the scene to be answered based on the scene status of each scene, the scene question and answer process of the state evaluation, and the dialogue jump strategy; Based on the target dialogue model and the question-and-answer scenario, the evaluation questions for the next round are generated.

[0010] According to a state evaluation method provided by the present invention, generating the next round of evaluation questions based on the target dialogue model and the question-answering scenario includes: Based on the reply, updating the scenario dialogue stack of each scenario used for the state evaluation; Based on the target dialogue model, the scenario to be asked and answered, and the scenario dialogue stack of the scenario to be asked and answered, the evaluation questions for the next round are generated.

[0011] According to a state evaluation method provided by the present invention, generating the next round of evaluation questions based on the target dialogue model, the scenario to be asked and answered, and the scenario dialogue stack of the scenario to be asked and answered includes: Generate a question and answer prompt based on the scenario to be asked and answered, the scenario dialogue stack of the scenario to be asked and answered, and the scenario question and answer process; Based on the target dialogue model, an evaluation question corresponding to the question and answer prompt is generated as the evaluation question for the next round.

[0012] The present invention also provides a state assessment device, comprising the following modules: A response receiving unit, used to receive the user's response to the evaluation question of the current round; A jump strategy and emotional intention determination unit, used to determine a dialogue jump strategy and the emotional intention of the user in the scenario based on the scenario where the evaluation question is located, the reply, and the historical dialogue of the user in the status evaluation; A model selection and evaluation question generation unit, configured to determine a target dialogue model from the candidate dialogue models based on the dialogue jump strategy, generate evaluation questions for the next round based on the target dialogue model, and return to receive the user's reply to the evaluation questions for the next round until the interaction ends; The state evaluation unit is used to perform state evaluation based on the emotional intention of the user in the scenario of the evaluation questions in each round.

[0013] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, any of the above-mentioned state assessment methods is implemented.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the state assessment method described in any one of the above is implemented.

[0015] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the state assessment method described above is implemented.

[0016] The state assessment method, device, equipment and medium provided by the present invention realize flexible switching of dialogue modes through dialogue jump strategies, which can achieve accurate assessment and effectively provide emotional support. This flexible switching improves the natural fluency of the dialogue and enables users to have a more intimate experience in the interaction. At the same time, by dynamically generating evaluation questions in each round of dialogue, the user's emotional expression and state change trend are gradually accumulated, thereby realizing accurate analysis and real-time adjustment of the user's emotional state, effectively improving the accuracy and personalization level of emotional assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 It is one of the flow charts of the state assessment method provided by the present invention.

[0019] Figure 2 It is a flow chart of generating the next round of evaluation questions based on the question and answer prompts provided by the present invention.

[0020] Figure 3 This is the second flow chart of the state assessment method provided by the present invention.

[0021] Figure 4 It is a structural schematic diagram of the state assessment device provided by the present invention.

[0022] Figure 5 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] Status assessment has become an important part of many fields such as mental health, education, medical care and social interaction. As the pace of society accelerates and the pressure of life increases, the emotional problems people face are becoming increasingly complex. Status assessment is not only used to discover potential mental health problems, but also helps various industries better understand users' emotional needs and provide personalized services. For example, status assessment in the field of education can improve teaching methods, while in the medical field, accurate status assessment is the key to identifying mental health problems (such as depression and anxiety). The advancement of status assessment technology is of great significance to improving user experience, promoting health management and optimizing service quality.

[0025] Current state assessment technologies mainly rely on task-based dialogues and open-domain dialogues. Task-based dialogues have a structured process and can effectively guide users to complete information collection and specific tasks, but they often lack natural emotional resonance and are difficult to truly touch users' emotions. Open-domain dialogues allow users to express themselves freely and can provide more natural and intimate emotional support, but due to the lack of a structured process, they are insufficient in task focus and information collection. Therefore, there is an urgent need for a method that can both meet users' emotional needs and complete state assessment tasks.

[0026] In view of the above problems, an embodiment of the present invention provides a state assessment method. Figure 1This is one of the flow charts of the state assessment method provided by the present invention. Figure 1 As shown, the method includes: Step 110: Receive the user's response to the evaluation question of the current round.

[0027] Specifically, in the state assessment process, a "round" refers to a specific question and answer, that is, a complete process of asking a state assessment question to the user and the user responding to it. The setting of multiple rounds helps to collect key information such as the user's emotional state and psychological stressors in a progressive manner. For example, the user's overall emotional state may be asked in the first round, and the specific sources of life or work stress may be further explored in the second round, and in the subsequent rounds, more detailed issues such as sleep quality and social interests may be explored.

[0028] The user's response to the current round of evaluation questions is the key basis for dynamically generating subsequent evaluation questions. Based on the user's answers in each round and the current scene of the conversation, the evaluation questions will be continuously adjusted to better adapt to the user's emotional state and conduct personalized in-depth evaluation. This design ensures the structure and flexibility of the conversation, helping to achieve accurate and user-friendly emotional evaluation.

[0029] Step 120, based on the scenario where the assessment question is located, the reply, and the historical dialogue of the user in the status assessment, determine the dialogue jump strategy and the emotional intention of the user in the scenario.

[0030] Specifically, each evaluation question is associated with a specific conversation scenario, such as "sleep scenario" or "emotional scenario". The scenario information provides a contextual basis for judging the user's response content, which helps to understand the user's emotional state and conversation needs. Based on the current conversation scenario, the jump strategy will also be different, such as whether to ask further questions in the same scenario, or jump to other related scenarios as needed (such as jumping from an emotional scenario to a sleep scenario).

[0031] The user's response in each round is a direct reflection of the emotional intention. By analyzing the emotional tendency of the current response (such as anxiety, fatigue, depression, etc.), the user's psychological state in the current scenario can be further clarified. Associating the user's response with the scenario can more accurately judge the user's emotional needs and provide support for the determination of emotional intentions and the formulation of dialogue jump strategies.

[0032] The user's historical conversation records during multiple rounds of evaluation are also an important basis for judging emotional states. For example, if a user mentioned "work pressure" many times in the previous rounds of conversation, then in the current round, if the user mentioned "feeling tired", the emotional intention can be determined as "stress-related fatigue" in combination with historical records. By referring to historical conversations, not only can the content of the conversation be kept consistent, but it can also help identify potential trends in emotional state changes, providing information support for generating continuous conversation content.

[0033] Based on the above information, the current conversation jump strategy can be generated. The jump strategy determines the direction of the subsequent conversation, that is, whether to continue to explore the current scene, or switch to a more suitable conversation scene according to the user's emotional state, or enter an open domain conversation, etc. For example, when the user expresses low emotions, the jump strategy may enter an open domain conversation to provide empathy support and provide more soothing conversations; if the user shows a clearer problem demand, the jump strategy may point to a specific task-based conversation node.

[0034] In addition, by analyzing the user's current response, conversation scenario, and the user's historical conversations, the user's emotional intent can be determined to ensure that the next round of conversation generation is closer to the user's emotional state. Emotional intent specifically refers to the corresponding emotions or states of users in different scenarios, which includes the user's emotional state in a specific scenario. For example, in a sleeping scenario, the user has difficulty falling asleep, and it lasts for 1-5 days. For another example, in an emotional scenario, the user becomes emotionally unstable. Emotional intent ensures that the next round of evaluation questions is closer to the user's emotional state, further improving the conversation experience.

[0035] Step 130, based on the dialogue jump strategy, determine the target dialogue model from the candidate dialogue models, and based on the target dialogue model, generate the next round of evaluation questions, and return to receive the user's reply to the next round of evaluation questions until the interaction ends.

[0036] Specifically, based on the dialogue jump strategy, the target dialogue model that best suits the current scenario is selected from multiple candidate dialogue models, and the model is used to generate the next round of evaluation questions. The dialogue jump strategy determines the direction of the dialogue, thereby ensuring that the selected dialogue model can meet the user's current needs and emotional state. For example, different dialogue models such as task-based dialogue models, open domain dialogue models, retrieval models, or generative models may be selected to adapt to the specific needs of users (such as guiding users to complete a task or providing emotional support).

[0037] Once the target dialogue model is determined, the next round of evaluation questions suitable for the user's situation will be generated based on the model. The generated evaluation questions are then sent to the user and wait for the user's response. Each round of user responses will be received and processed, triggering a new dialogue jump strategy and dialogue model selection, gradually achieving a dynamic evaluation of the user's emotional state.

[0038] Step 140 , performing a status assessment based on the user's emotional intention in the scenario of the assessment questions in each round.

[0039] The analysis of emotional intent enables the entire evaluation process to accumulate the user's emotional expression and state changes in each round, so as to obtain more comprehensive emotional state information in the final evaluation. By recording and analyzing the emotional intent in each round, it is possible to identify the user's persistent emotional state in different scenarios, such as work pressure, sleep problems or life events.

[0040] Specifically, the state assessment judges the user's overall emotional health by summarizing the user's emotional reactions (such as anxiety, depression, emotional instability, etc.) in each round of conversation and combining the changing trends of these emotions in different scenarios. For example, if the user shows persistent difficulty falling asleep in multiple rounds of sleeping scenarios, accompanied by emotional intentions of depression, the assessment results may tend to be identified as excessive stress or potential depression symptoms. For another example, if the user frequently shows emotional instability in emotional scenarios, further attention may be needed to the user's emotional management and psychological state.

[0041] The state evaluation result here may be an evaluation result of a depressive emotional state or an evaluation result of an anxious emotional state. After the state evaluation result is obtained, it may be synchronized to a professional to assist the professional in diagnosing or treating the user's state.

[0042] The state assessment method provided by the present invention realizes flexible switching of dialogue modes through dialogue jump strategies, which can achieve accurate assessment and effectively provide emotional support. This flexible switching improves the natural fluency of the dialogue and enables users to have a more intimate experience in the interaction. At the same time, by dynamically generating evaluation questions in each round of dialogue, the user's emotional expression and state change trend are gradually accumulated, thereby realizing accurate analysis and real-time adjustment of the user's emotional state, effectively improving the accuracy and personalization level of state assessment.

[0043] Based on the above embodiment, the method of determining the dialogue jump strategy and the emotional intention of the user in the scenario based on the scenario where the evaluation question is located, the reply, and the historical dialogue of the user in the status evaluation includes: Generate input text based on the scenario where the assessment question is located, the reply, the historical dialogue of the user in the state assessment, and an input text template, wherein the input text template includes task description information of dialogue jump and emotion intention recognition; The input text is input into a large language model to obtain a dialogue jump strategy output by the large language model and the emotional intention of the user in the scenario.

[0044] Specifically, the input text template is used to provide a structured framework to help the model judge the direction of the conversation and the user's emotional intention. The beginning of the input text template provides a task summary, which is a brief description of the task that the large language model needs to complete. For example, "Task summary: Identify the user's emotional state and generate a jump strategy that meets the user's needs." This task summary helps the large language model clarify the overall goal of this next round of conversation and ensure that the generated content revolves around the evaluation or support of the user's emotions. The task summary directly guides the model to focus on the main task, making its output more targeted. The second module of the input text template is the current scene information, which contains the scene information of the evaluation question, such as labels such as "emotion", "sleep quality" or "stress". The scene information allows the large language model to quickly understand the context of the conversation, making it easier to focus on this scene when analyzing the user's response. For example, in the emotional scene, the large language model will focus on emotion-related vocabulary and expressions. The third module is the historical conversation part, which contains the user's conversation records in previous rounds. The historical conversation provides the changing trend and context information of the user's emotions. For example, if the historical record mentions "anxiety" many times, the model can identify the accumulation of emotions based on this trend, and then support the decision-making of the jump strategy and the generation of emotional intent. The fourth module of the input text template is the user's current round reply, which directly affects the large voice model's recognition of the user's emotional intent and the generation of the jump strategy. For example, if the user mentions "feeling very tired" in the current reply, the model will generate the corresponding emotional intent and jump strategy based on the scene and historical dialogue. In addition to the task summary, the input text template also clearly lists the specific requirements and precautions for the current task. For example, "Requirement: Provide emotional support" and "Note: Avoid asking sensitive topics directly." This module ensures that the model meets the expected interaction effect when generating the next round of evaluation questions. The next module is the strategy selection condition. The strategy selection condition in the template provides the model with specific rules for executing the jump strategy. For example, "If the user mentions emotional instability, enter the empathy dialogue mode" or "If stress is mentioned many times, switch to the stress scene." These conditions enable the model to flexibly select the dialogue mode under different emotional states, thereby achieving a balance between task orientation and emotional support. The last module is the format of the output of the large language model, such as "emotional intent: scene-node-state" and other formats. The following table shows the format of the emotion intent in the sleep scenario and emotional scenario:

[0045] The format of the jump strategy may be "current scene / switch to other scenes / enter empathy dialogue mode", etc. This standardized output format is convenient for subsequent modules to process and ensure the efficient generation of subsequent questions.

[0046] By generating input text according to the input text template and inputting it into a large language model, we can obtain the dialogue jump strategy and emotional intention. The main function of the dialogue jump strategy is to determine the next direction of the dialogue based on the scene and emotional state information in the dialogue. It will choose the appropriate direction of the dialogue according to the user's emotional state and reply content, such as continuing the current topic, switching to other scenes, or entering empathy dialogue mode when the mood is low. Through flexible jump strategies, evaluation questions can naturally connect different scenes and find a balance between task orientation and emotional support. The advantage of the dialogue jump strategy is that it ensures that the dialogue content dynamically adapts to user needs, making the evaluation interaction more smooth and coherent.

[0047] The role of emotional intent recognition is to identify the user's emotional state by analyzing the user's language and emotional expressions, and generate specific emotional labels, such as "anxious", "depressed" or "happy". These labels provide reference for subsequent conversation content, so that the generated responses can be closer to the user's actual emotions. The advantage of emotional intent recognition is that it makes the conversation more personalized and ensures that adaptive adjustments are made in a timely manner after the emotion is recognized, thereby providing more empathetic support.

[0048] Based on the above embodiment, the input text template also includes strategy selection description information of the dialogue jump strategy.

[0049] The strategy selection description information in the input text template provides the large language model with rule guidance on when and why to choose a certain dialogue jump strategy. The strategy selection description information details the applicable conditions and triggering scenarios of each jump strategy, so as to help the large language model understand and select different dialogue jump strategies. Specifically, when the user's emotional state is stable and there is still room for in-depth discussion on the current topic, the large language model will choose to continue the conversation in that scenario, that is, continue the current conversation. This strategy is applicable to situations where the user has not yet fully expressed or is interested in the current scenario. When the user's response is unclear, the expression is vague, or the core question is not answered, the large voice model can choose to repeat the current question to guide the user to provide a clearer response. This strategy ensures the integrity of information collection and reduces misunderstandings or missing information. When the user mentions a new scenario (such as switching from a stress topic to a sleep problem), the large voice model will choose to jump to the relevant dialogue scenario. This strategy supports flexible switching of dialogue content to adapt to the user's actual needs or emotional changes. When the user's current emotional state is highly related to the previous dialogue content, the model can choose to jump back to the historical dialogue. This strategy keeps the dialogue coherent, especially when it is necessary to revisit the content that has been discussed. When a user expresses negative emotions or seeks support, the large voice model can choose to enter the empathy response mode and provide soothing and supportive responses rather than further task-oriented questions. This strategy is suitable for scenarios that help users relieve emotions and provide psychological support. When the current conversation has fully completed the task objectives and the user's emotions are stable, the large voice model will choose to end this round of conversation. This strategy ends the interaction at the right time, avoiding excessive intervention and providing a moderate sense of conversation closure.

[0050] Based on the above embodiment, the candidate dialogue models include a chat dialogue model and a task-based dialogue model for state assessment tasks; The step of determining a target dialogue model from each candidate dialogue model based on the dialogue jump strategy includes: When the dialogue jump strategy is empathy response, determining the chat dialogue model as the target dialogue model; When the dialogue jump strategy is not the empathic response, the task-based dialogue model is determined to be the target dialogue model.

[0051] Specifically, the chat dialogue model is a type of open domain dialogue model. The open domain dialogue model focuses on natural communication without specific task restrictions, is applicable to a wide range of topics, and is highly flexible. The chat dialogue model focuses on open, non-task-oriented communication with users, and is usually used in scenarios that require emotional expression, such as emotional support and empathetic responses. The goal of this model is to provide a natural and warm communication experience, which is suitable for users to express emotions, get comfort, or relax. It can identify and respond to the user's emotional state, generate responses in a gentle and caring tone, guide users to express emotions, and provide psychological comfort and companionship. The task-based dialogue model focuses on structured and goal-oriented dialogues, and is mainly used to complete specific state assessment tasks. In this model, the dialogue is carried out around specific goals, such as the assessment of emotional state, the inquiry of specific questions, etc. The design of the task-based dialogue model is more process-oriented, and can advance the dialogue according to the predetermined logic, guide users to provide accurate answers, and ensure the completeness and accuracy of the assessment.

[0052] When the dialogue jump strategy is empathetic response, the chat dialogue model is selected as the target dialogue model. At this time, the goal of the chat dialogue model is to provide a more emotionally supportive response, focusing on the user's emotional expression and needs, and is suitable for use in situations where the user's emotions fluctuate greatly or need emotional support.

[0053] When the dialogue jump strategy is not empathetic response, the task-based dialogue model is selected as the target dialogue model. The task-based dialogue model focuses on guiding users to complete specific state assessments. The model generates structured questions to guide users to provide necessary information, such as current mental state, emotions, etc.

[0054] Based on the above embodiment, when the dialogue jump strategy is not the empathic response, generating the next round of evaluation questions based on the target dialogue model includes: Based on the emotional intention in the scenario, updating the scenario state of each scenario used for the status evaluation; Determine the scene to be answered based on the scene status of each scene, the scene question and answer process of the state evaluation, and the dialogue jump strategy; Based on the target dialogue model and the question-and-answer scenario, the evaluation questions for the next round are generated.

[0055] Specifically, during the state evaluation process, each scene state marks whether the current scene has been asked. For example, a scene may contain multiple nodes, and the nodes represent specific detailed questions or emotional states. When the model recognizes that the user shows relevant emotional intentions (such as anxiety, depression, etc.) in a specific scene, the scene state will be updated to "asked". This update ensures that subsequent conversations will not re-enter the scene unless necessary, but focus on the unfinished part, thereby improving the efficiency of the conversation and user experience. Emotional intent and scene state are directly associated with the state of the scene node. For example, in the "sleep problem" scene, when the emotional intent marks that the user has "difficulty falling asleep", the relevant nodes of the scene will be updated to "asked". Conversely, if the user does not show the emotional intention associated with the scene, the scene node will remain in the "unasked" state for further questioning.

[0056] When determining the scenarios to be asked and answered, the scenario status, scenario question-answering process, and conversation jump strategy work together. First, the scenario status records whether each scenario node has been asked. By checking the scenario status, it is possible to identify which scenarios have not been asked or have been partially completed, ensuring that the subsequent conversation focuses on the unfinished parts. Next, the scenario question-answering process provides the overall path of the conversation, defines the logical order and relevance between the scenarios, and helps to make choices in the established order in the unasked scenarios to maintain the fluency and systematicness of the conversation. On this basis, the conversation jump strategy decides whether to follow the original process or jump to other scenarios based on the user's response in this round and the user's historical conversation in the status evaluation to meet the user's emotional and information needs. In this way, the three work together: the scenario status ensures that the completed content will not be asked repeatedly, the scenario question-answering process provides a stable conversation structure, and the jump strategy flexibly switches between different scenarios based on the user's real-time emotions.

[0057] After the question-and-answer scenario is determined, the target dialogue model (such as a task-based dialogue model) will generate specific evaluation questions based on the scenario. Task-based dialogue models focus on collecting structured information and will generate detailed questions around the theme of the scenario. For example, when the question-and-answer scenario is "sleep", the model may generate the question "How was your sleep in the past week?" The target dialogue model will refer to the specific content of the question-and-answer scenario (such as "difficulty falling asleep", "waking up early", etc.) to generate questions to ensure that the questions are in line with the user's actual emotional state, while avoiding repeated or invalid questions. This design is particularly applicable under the task-based dialogue model, so that each round of dialogue can effectively obtain valuable information, and ultimately provide comprehensive data support for status evaluation.

[0058] After the question-and-answer scenario is determined, the type of target dialogue model will determine how to generate the next round of evaluation questions based on the question-and-answer scenario. For the task-based dialogue model, it focuses on structured information collection and can generate detailed questions around specific scenario topics. For example, when the question-and-answer scenario is "sleep", the task-based dialogue model may generate specific questions such as "How was your sleep in the past week?" or "Are there frequent early awakenings?" The design of this model ensures the rigor of the dialogue and the accuracy of the information, so that each round of questions can effectively obtain valuable data, which helps to ensure the comprehensiveness and accuracy of the status assessment.

[0059] For the chat dialogue model, if the dialogue jump strategy indicates that the user needs more emotional support, the chat dialogue model will be enabled. This model pays more attention to the expression of user emotions, generates responses in a gentle and empathetic way, and guides users to naturally share more emotional details. For example, in the "sleep" scenario, the chat dialogue model may ask open-ended questions such as "Are you bothered by your recent sleep situation?" or "Is there anything special that makes it difficult for you to fall asleep?" to encourage users to express their personal feelings, thereby enhancing understanding of their emotions. This type of empathetic dialogue provides users with emotional guidance and makes the evaluation process more humane.

[0060] Based on the above embodiment, generating the next round of evaluation questions based on the target dialogue model and the question-answering scenario includes: Based on the reply, updating the scenario dialogue stack of each scenario used for the state evaluation; Based on the target dialogue model, the scenario to be asked and answered, and the scenario dialogue stack of the scenario to be asked and answered, the evaluation questions for the next round are generated.

[0061] Specifically, the scene conversation stack is a structure used to store the conversation history in a specific scene. Each scene has an independent conversation stack, which is used to record all the questions in the scene, the user's replies, and the context information during the conversation. Through the scene conversation stack, the user's communication content in the scene can be traced back at any time, so that the continuity and consistency of the conversation can be maintained. When the user provides a new reply in the current scene, this information will be added to the conversation stack of the corresponding scene. For example, in the "sleep scene", if the user is asked "Do you often suffer from insomnia" and answers "Occasionally", the question and the user's reply will be recorded in the conversation stack of the "sleep scene". In this way, the stack continuously accumulates the user's answers and context information to form a complete conversation track. The stack is updated in real time, ensuring that the content of each round of interaction can be referenced by subsequent conversations to avoid information loss and duplication.

[0062] In each round of dialogue, generating new evaluation questions requires comprehensive consideration of the target dialogue model, the current question-to-be-answered scenario, and the historical information in the scenario dialogue stack. The type of target dialogue model affects the style and content of the generated questions. If the target dialogue model is a task-based dialogue model, the generated questions will tend to be more structured and specific to ensure that accurate state evaluation data is collected; if the target model is a chat dialogue model, the generated questions may be more open-ended to allow users to express their emotions freely. The question-to-be-answered scenario defines the specific topic and focus of the current dialogue, such as "sleep problems" or "anxiety problems." When generating questions, referring to the question-to-be-answered scenario can ensure that the questions revolve around specific topics and keep the conversation focused. For example, in the "sleep problems" scenario, the generated question may be "Do you usually need more than half an hour to fall asleep?" to further explore the user's sleep situation. The scenario dialogue stack records all historical dialogues under the scenario, providing background information for the generation of new questions. For example, in the "Sleep Scenario", the user has been asked "Have you had any sleep problems recently?" and the answer is "Often", then the next round of questions generated can continue to ask more specific details based on the answer, such as "What time do you usually go to bed?" or "Is there any special reason for not being able to sleep?" By referring to the scenario dialogue stack, repeated questions can be avoided, ensuring that each question further explores the previous information, thereby increasing the depth and coherence of the conversation.

[0063] Based on the above embodiment, the step of generating the next round of evaluation questions based on the target dialogue model, the scenario to be asked and answered, and the scenario dialogue stack of the scenario to be asked and answered includes: Generate a question and answer prompt based on the scenario to be asked and answered, the scenario dialogue stack of the scenario to be asked and answered, and the scenario question and answer process; Based on the target dialogue model, an evaluation question corresponding to the question and answer prompt is generated as the evaluation question for the next round.

[0064] Figure 2 This is the process of generating the next round of evaluation questions based on the question and answer prompts provided by the present invention. Figure 2As shown in the figure, first, the scenario to be asked and answered is the basis for question generation. Secondly, through the scenario dialogue stack, historical dialogues related to the current scenario are retrieved to ensure that they are carried out on the basis of previous questions and answers to maintain the continuity of the dialogue. For example, the user may have been asked whether he has mental distress before and relevant feedback was obtained. The scenario dialogue stack provides background information for subsequent questions, which helps to avoid repeated questions and make the dialogue more contextual. Then, the content to be asked in the next round is selected by referring to the pre-designed scenario question and answer process. The scenario question and answer process defines the recommended inquiry sequence and logical relationship in a specific scenario. For example, in the "sleeping scenario", the process may suggest further confirmation of "whether the user's sleeping time exceeds half an hour". In this way, each question is raised under the guidance of the established process to ensure the comprehensiveness and systematicness of information collection. Finally, these three parts of information are spliced ​​into prompts and input into the target dialogue model to generate specific evaluation questions. The task-based dialogue model will generate specific questions based on the current scenario and the content to be asked in the process. For example, in the "falling asleep scenario", the question generated by the model might be "Does it usually take you more than half an hour to fall asleep?" This question not only accurately captures specific data related to falling asleep, but also maintains the appropriateness of the question, making the conversation meet the evaluation objectives.

[0065] Based on the above embodiments, Figure 3 This is the second flow chart of the state assessment method provided by the present invention. Figure 3 As shown in the figure, the user first provides the basic information required for status assessment through voice input. This voice input is converted into text data by the voice recognition module for subsequent processing. After conversion into text, the natural language understanding module understands the text input, mainly including scene recognition, intent recognition and task decision. Scene recognition is used to determine which scene the input content belongs to, such as "sleep problems" or "emotional state"; intent recognition analyzes the emotions expressed by the user in the scene; task decision determines the next dialogue jump strategy based on the scene and intent, such as choosing an empathetic response when the user shows emotional distress. Finally, the output generated by the natural language understanding module includes scenes, emotional intentions and dialogue jump strategies, and passes this information to the dialogue management module.

[0066] The dialogue management module performs overall control and state management of the dialogue based on the information transmitted by the natural language understanding module. Its functions include dynamic storage of historical dialogues and dialogue flow guidance. The dynamic storage record of historical dialogues is the scene stack, which stores the current dialogue content and previous interaction information, maintaining coherence and context relevance. The dialogue flow guidance determines the next evaluation dialogue process based on the user's emotional intention and dialogue jump strategy. For example, if the user is expressing an emotional problem, a new scene may be selected or an appropriate guiding question may be generated. The dialogue management module outputs dialogue strategy guidance and historical dialogue information to provide guidance for the natural language generation module.

[0067] The natural language generation module generates appropriate text responses based on the historical dialogue information and dialogue strategy guidance provided by the dialogue management module. For example, for an input such as "feeling emotionally unstable", the natural language generation module may generate caring responses such as "Do you have any special stressors recently?" or "Want to talk to me about your specific problems?" The generated text responses are finally converted into speech outputs through the text-to-speech module, completing the entire interaction process and feeding back the responses to the user in the form of speech.

[0068] The state assessment device provided by the present invention is described below. The state assessment device described below and the state assessment method described above can be referred to each other.

[0069] Figure 4 is a schematic diagram of the structure of the state assessment device provided by the present invention, such as Figure 5 As shown, the device comprises: A reply receiving unit 410 is used to receive a user's reply to the evaluation question of the current round; A jump strategy and emotional intention determination unit 420, configured to determine a dialogue jump strategy and the emotional intention of the user in the scenario based on the scenario in which the evaluation question is located, the reply, and the historical dialogue of the user in the state evaluation; A model selection and evaluation question generation unit 430 is used to determine a target dialogue model from the candidate dialogue models based on the dialogue jump strategy, generate the next round of evaluation questions based on the target dialogue model, and return to receive the user's reply to the next round of evaluation questions until the interaction ends; The state evaluation unit 440 is used to perform state evaluation based on the emotional intention of the user in the scenario of the evaluation questions in each round.

[0070] The state assessment device provided by the present invention realizes flexible switching of dialogue modes through dialogue jump strategies, which can achieve accurate assessment and effectively provide emotional support. This flexible switching improves the natural fluency of the dialogue and enables users to have a more intimate experience in the interaction. At the same time, by dynamically generating evaluation questions in each round of dialogue, the user's emotional expression and state change trend are gradually accumulated, thereby realizing accurate analysis and real-time adjustment of the user's emotional state, effectively improving the accuracy and personalization level of emotional assessment.

[0071] Based on any of the above embodiments, the jump strategy and emotional intention determination unit is specifically used to: Generate input text based on the scenario where the assessment question is located, the reply, the historical dialogue of the user in the state assessment, and an input text template, wherein the input text template includes task description information of dialogue jump and emotion intention recognition; The input text is input into a large language model to obtain a dialogue jump strategy output by the large language model and the emotional intention of the user in the scenario.

[0072] The input text template also includes strategy selection description information of the dialogue jump strategy.

[0073] Based on any of the above embodiments, the model selection and evaluation question generation unit is specifically used for: The candidate dialogue models include a chat dialogue model and a task-based dialogue model oriented to a state assessment task; When the dialogue jump strategy is empathy response, determining the chat dialogue model as the target dialogue model; When the dialogue jump strategy is not the empathic response, the task-based dialogue model is determined to be the target dialogue model.

[0074] Based on the emotional intention in the scenario, updating the scenario state of each scenario used for the status evaluation; Determine the scene to be answered based on the scene status of each scene, the scene question and answer process of the state evaluation, and the dialogue jump strategy; Based on the target dialogue model and the question-and-answer scenario, the evaluation questions for the next round are generated.

[0075] Based on the reply, updating the scenario dialogue stack of each scenario used for the state evaluation; Based on the target dialogue model, the scenario to be asked and answered, and the scenario dialogue stack of the scenario to be asked and answered, the evaluation questions for the next round are generated.

[0076] Generate a question and answer prompt based on the scenario to be asked and answered, the scenario dialogue stack of the scenario to be asked and answered, and the scenario question and answer process; Based on the target dialogue model, an evaluation question corresponding to the question and answer prompt is generated as the evaluation question for the next round.

[0077] Figure 5 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 5 As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530 and a communication bus 540, wherein the processor 510, the communication interface 520 and the memory 530 communicate with each other through the communication bus 540. The processor 510 may call the logic instructions in the memory 530 to execute the state evaluation method, which includes: Receive user responses to the current round of evaluation questions; Based on the scenario where the evaluation question is located, the reply, and the historical dialogue of the user in the status evaluation, determine the dialogue jump strategy and the emotional intention of the user in the scenario; Based on the dialogue jump strategy, determine the target dialogue model from the candidate dialogue models, generate the next round of evaluation questions based on the target dialogue model, and return to receive the user's reply to the next round of evaluation questions until the interaction ends; Based on the emotional intention of the user in the scenario of the evaluation questions in each round, a state evaluation is performed.

[0078] In addition, the logic instructions in the above-mentioned memory 530 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the relevant technology or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0079] On the other hand, the present invention further provides a computer program product, the computer program product comprising a computer program, the computer program can be stored on a non-transitory computer-readable storage medium, when the computer program is executed by a processor, the computer can execute the state assessment method provided by the above methods, the method comprising: Receive user responses to the current round of evaluation questions; Based on the scenario where the evaluation question is located, the reply, and the historical dialogue of the user in the status evaluation, determine the dialogue jump strategy and the emotional intention of the user in the scenario; Based on the dialogue jump strategy, determine the target dialogue model from the candidate dialogue models, generate the next round of evaluation questions based on the target dialogue model, and return to receive the user's reply to the next round of evaluation questions until the interaction ends; Based on the emotional intention of the user in the scenario of the evaluation questions in each round, a state evaluation is performed.

[0080] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to perform the state assessment method provided by the above methods, the method comprising: Receive user responses to the current round of evaluation questions; Based on the scenario where the evaluation question is located, the reply, and the historical dialogue of the user in the status evaluation, determine the dialogue jump strategy and the emotional intention of the user in the scenario; Based on the dialogue jump strategy, determine the target dialogue model from the candidate dialogue models, generate the next round of evaluation questions based on the target dialogue model, and return to receive the user's reply to the next round of evaluation questions until the interaction ends; Based on the emotional intention of the user in the scenario of the evaluation questions in each round, a state evaluation is performed.

[0081] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0082] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiment.

[0083] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A state assessment method, characterized in that: include: Receive user responses to the current round of evaluation questions; Determine a dialogue jump strategy and the user's emotional intention in the scenario based on the scenario in which the assessment question is located, the reply, and the user's historical dialogue in the status assessment; Based on the dialogue jump strategy, determine the target dialogue model from the candidate dialogue models, generate the next round of evaluation questions based on the target dialogue model, and return to receive the user's reply to the next round of evaluation questions until the interaction ends; Based on the emotional intention of the user in the scenario of the evaluation questions in each round, a state evaluation is performed.

2. The state assessment method according to claim 1, characterized in that: The determining of the dialogue jump strategy and the emotional intention of the user in the scenario based on the scenario where the evaluation question is located, the reply, and the historical dialogue of the user in the status evaluation includes: Generate input text based on the scenario where the assessment question is located, the reply, the historical dialogue of the user in the state assessment, and an input text template, wherein the input text template includes task description information of dialogue jump and emotion intention recognition; The input text is input into a large language model to obtain a dialogue jump strategy output by the large language model and the emotional intention of the user in the scenario.

3. The state assessment method according to claim 2, characterized in that: The input text template also includes strategy selection description information of the dialogue jump strategy.

4. The state assessment method according to claim 1, characterized in that: The candidate dialogue models include a chat dialogue model and a task-based dialogue model oriented to a state assessment task; The step of determining a target dialogue model from each candidate dialogue model based on the dialogue jump strategy includes: When the dialogue jump strategy is empathy response, determining the chat dialogue model as the target dialogue model; When the dialogue jump strategy is not the empathic response, the task-based dialogue model is determined to be the target dialogue model.

5. The state assessment method according to claim 4, characterized in that: In the case where the dialogue jump strategy is not the empathic response, generating the next round of evaluation questions based on the target dialogue model includes: Based on the emotional intention in the scenario, updating the scenario state of each scenario used for the status evaluation; Determine the scene to be answered based on the scene status of each scene, the scene question and answer process of the state evaluation, and the dialogue jump strategy; Based on the target dialogue model and the question-and-answer scenario, the evaluation questions for the next round are generated.

6. The state assessment method according to claim 5, characterized in that: The generating the next round of evaluation questions based on the target dialogue model and the question-answering scenario includes: Based on the reply, updating the scenario dialogue stack of each scenario used for the state evaluation; Based on the target dialogue model, the scenario to be asked and answered, and the scenario dialogue stack of the scenario to be asked and answered, the evaluation questions for the next round are generated.

7. The state assessment method according to claim 6, characterized in that: The generating the next round of evaluation questions based on the target dialogue model, the scenario to be asked and answered, and the scenario dialogue stack of the scenario to be asked and answered includes: Generate a question and answer prompt based on the scenario to be asked and answered, the scenario dialogue stack of the scenario to be asked and answered, and the scenario question and answer process; Based on the target dialogue model, an evaluation question corresponding to the question and answer prompt is generated as the evaluation question for the next round.

8. A state assessment device, characterized in that: include: A response receiving unit, used to receive the user's response to the evaluation question of the current round; A jump strategy and emotional intention determination unit, used to determine a dialogue jump strategy and the emotional intention of the user in the scenario based on the scenario where the evaluation question is located, the reply, and the historical dialogue of the user in the state evaluation; A model selection and evaluation question generation unit, configured to determine a target dialogue model from the candidate dialogue models based on the dialogue jump strategy, generate evaluation questions for the next round based on the target dialogue model, and return to receive the user's reply to the evaluation questions for the next round until the interaction ends; The state evaluation unit is used to perform state evaluation based on the emotional intention of the user in the scenario of the evaluation questions in each round.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the state assessment method according to any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the state assessment method according to any one of claims 1 to 7 is implemented.

11. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the state assessment method according to any one of claims 1 to 7 is implemented.

Images