A cognitive map-based psychological support dialogue interaction method

Abstract

The application discloses a psychological support dialogue interaction method based on a cognitive map, and comprises the following steps: receiving current round user text input, user explicit selection mode, historical session state, historical mode track, user basic portrait and emotion parameters; constructing a psychological task confidence vector to identify the current user psychological task; loading the corresponding mode strategy template according to matching; generating a current round reply according to a generated control parameter vector and round promotion rules; and monitoring and feeding back signals to the user, and writing back the current round mode, parameters, user acceptance and switching records to a session state library.

Description

Technical Field

[0001] This invention relates to the field of human-computer interaction technology, and in particular to a psychological support dialogue interaction method based on cognitive graphs. Background Technology

[0002] Existing large-scale chat systems often employ standardized prompts, uniform personalities, or a few fixed patterns for responses. While they can output fluent and coherent text, they often struggle to address the multiple core needs in psychological support scenarios, such as low stimulation, safe boundaries, empathic connection, and task progression. Users' actual needs in these scenarios vary significantly: some simply require companionship, some need to prioritize releasing negative emotions, some are in a high-arousal state and require reduced stimulation, and others possess self-management abilities and require guidance on cognitive breakdown or action planning. A standardized approach is prone to misaligned responses; for example, excessive rational analysis when users simply need to vent, overly didactic when users are under pressure, or overly superficial responses when users are in a self-exploration phase, failing to accurately match the user's real-time psychological state and core needs.

[0003] Another drawback of the existing approach is that even if the system has multiple "modes" set up internally, they are mostly simple prompt word switching at the operational level. They lack a calculable psychological task recognition mechanism, standardized mode template field definitions, scientific round-progression rules, and flexible switching mechanism. As a result, the system mode switching lacks reasonable transition and natural connection, giving users a sudden sense of "style change". It is difficult to form a stable and coherent psychological companionship experience and cannot meet the core service requirements of psychological support scenarios. Summary of the Invention

[0004] This invention aims to at least solve one of the technical problems existing in the prior art. Therefore, one objective of this invention is to propose a cognitive graph-based psychological support dialogue interaction method that improves the matching degree of user interaction dialogue.

[0005] According to an embodiment of the present invention, a psychological support dialogue interaction method based on cognitive graphs includes: receiving text input from the user in the current round. User explicit selection mode Historical session status Historical pattern trajectory User basic profile and emotional parameters Based on the above information, construct a confidence vector for the psychological task. Identify the current user's mental task; based on Match and load the corresponding pattern strategy template ;in accordance with Generate control parameter vector Generate the current round's response according to the round progression rules; monitor... User feedback signals When the switching conditions are met, a flexible mode switch is executed, and a unified personality anchor point is used. Maintain consistency of personality; write the current round's mode, parameters, user acceptance, and switching records back to the session state database.

[0006] According to the cognitive graph-based psychological support dialogue interaction method of the present invention, the present invention upgrades psychological AI from the traditional single general dialogue mode to a computable, designable, and replicable multi-mode control system, effectively breaking through the limitations of the prior art. It can significantly improve the accuracy of response and the degree of demand matching in different psychological task scenarios, and ensure the pertinence and effectiveness of psychological support services.

[0007] According to the cognitive graph-based psychological support dialogue interaction method of the present invention, the present invention, through standardized template field definitions and scientific round-progression rules, performs fine-grained control over the response generation process, effectively reducing the didactic feel, template-based rigidity, and style abruptness during mode switching that are common in existing psychological support dialogues, and is more in line with the core needs of long-term companionship psychological support, thereby improving the consistency and comfort of user experience.

[0008] According to the cognitive graph-based psychological support dialogue interaction method of the present invention, the present invention, with the help of structured result cards and conversation state write-back mechanism, enables the dialogue interaction results to go beyond the text communication level and be deposited into traceable, reusable and optimizable psychological support results, providing data support for subsequent rounds of interaction and long-term psychological state management of users, and realizing a closed-loop upgrade of psychological support services.

[0009] According to the cognitive graph-based psychological support dialogue interaction method of the present invention, the present invention forms a close upstream and downstream technical connection with patents related to emotion parameters and emotion regulation, and builds a logically coherent and functionally complementary technical system, which is convenient for further integration to form a complete product patent chain, laying a solid foundation for subsequent productization and technology expansion.

[0010] According to some embodiments of the present invention, the psychological task confidence vector It is determined by the fusion of three types of signals: explicit signals, semantic signals, and state signals. The psychological tasks mentioned include one or more of the following: providing reassuring companionship, releasing stress, calming and relaxing, sorting out problems, deconstructing cognition, self-exploration, discovering strengths, and dream interpretation.

[0011] According to some embodiments of the present invention, psychological task recognition adopts a dual-channel structure of rule engine and classification model: first, the rule engine filters strongly constrained tasks, and then the classification model outputs probability ranking for the remaining candidate tasks.

[0012] According to some embodiments of the present invention, pattern strategy template It includes tone constraints, empathy intensity, question frequency, suggestion intensity, explanation depth, prohibited behaviors, stage goals, output format, maximum progress per round, and mode exit conditions.

[0013] According to some embodiments of the present invention, pattern strategy template It also includes the risk boundary field. With personality anchor field It is used to limit highly stimulating content and maintain personality continuity.

[0014] According to some embodiments of the present invention, the round-progression rules include: advancing only one point at a time; asking a maximum of one question at a time; prioritizing stabilization in high-awakening / high-load states; not allocating high-cost action tasks in low-resource states; and performing sorting or cognitive reconstruction only when the user has a clear sense of purpose.

[0015] According to some embodiments of the present invention, flexible mode switching employs a transition segment control mechanism; assuming the previous mode is... Candidate new modes are ,when And switch benefits Greater than the cost of switching At that time, the execution mode is switched.

[0016] According to some embodiments of the present invention, the switching cost ,in For pattern differentiation, The time interval between the last switchover. This represents the user's stable state coefficient.

[0017] According to some embodiments of the present invention, a unified personality anchor point This includes pronoun usage, response temperature, ethical boundaries, and taboo expressions, ensuring that the linguistic personality and assistant identity remain consistent before and after the switch.

[0018] According to some embodiments of the present invention, closed-loop updates include evaluating the effectiveness of the mode based on subsequent 1-3 rounds of user feedback, reducing the mode priority with negative feedback, and increasing the mode retention time with positive feedback.

[0019] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0020] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0021] Figure 1 This is a flowchart of the overall modules of the psychological support dialogue interaction method in this embodiment of the invention;

[0022] Figure 2 This is a logic diagram of psychological task recognition in an embodiment of the present invention;

[0023] Figure 3 This is a control flowchart for flexible mode switching in an embodiment of the present invention;

[0024] Figure 4 This is a diagram of the overall system architecture upon which the embodiments of the present invention are based;

[0025] Figure 5 This is a schematic diagram of the timing of a single-turn dialogue interaction in an embodiment of the present invention. Detailed Implementation

[0026] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0027] This embodiment provides a psychological support dialogue interaction method based on cognitive graphs, such as... Figure 1-5 As shown, it includes six core steps: information reception, psychological task recognition, pattern template matching, response generation, pattern switching, and closed-loop update. Each step is interconnected, forming a complete interactive closed loop.

[0028] Step S1: Receiving Multi-Dimensional Information

[0029] Receive text input from the current user in the current round User explicit selection mode Historical session status Historical pattern trajectory User basic profile and emotional parameters The above information provides basic data support for subsequent psychological task identification and response generation.

[0030] Among them, user text input The current round of text messages sent to the user, including their emotional concerns and needs, such as "I've been under a lot of work pressure lately and can't sleep at night"; the user can also explicitly select a mode. For user-selected psychological support modes, such as the "stress relief" mode manually chosen by the user; historical conversation status. This includes all session records prior to the current round, including user input history, system responses history, and interaction duration; historical pattern trajectories. This includes sequences of psychological support patterns previously used by the user, such as "reassuring companionship → stress release → cognitive deconstruction"; and a basic user profile. This includes basic information such as user age, gender, occupation, baseline psychological state, and past psychological support needs; emotional parameters. This refers to the current emotional characteristics of users extracted through a text sentiment analysis model, including emotion type (anxiety, depression, calmness, etc.) and emotion intensity (high, medium, low), such as "anxiety, high emotion intensity".

[0031] Step S2: Construction of Confidence Vectors for Mental Tasks and Recognition of Mental Tasks

[0032] Based on the multi-dimensional information received in step S1, a confidence vector for the psychological task is constructed. It also identifies the core psychological tasks of the current user through a dual-channel structure of rule engine and classification model.

[0033] Specifically, psychological task confidence vector From explicit signals semantic signals Status signals The three types of signals are fused together, and the fusion formula is as follows: ,in The fusion function employs a weighted summation method to fuse the three types of signals. The weight coefficients are determined through training with a large amount of psychological support interaction data to ensure the accuracy of the fusion results.

[0034] explicit signal Explicit selection mode for corresponding users The signal strength, for example, when the user selects the "pressure relief" mode, The signal value corresponding to "pressure relief" is the highest; semantic signal For user text input The semantic feature extraction results are used to extract psychological need keywords (such as "stress," "insomnia," and "confide") from the text using the BERT model, and mapped to signal values ​​for the corresponding psychological tasks; state signals From historical session state Historical pattern trajectory User basic profile and emotional parameters The data is generated through fusion, reflecting the user's current psychological state and historical demand trends.

[0035] The psychological task recognition adopts a dual-channel structure, such as Figure 2 As shown, the rule engine first filters out strongly constrained tasks. That is, when the signal value corresponding to a certain psychological task exceeds a preset threshold (such as 0.8), the task is directly determined as the current core psychological task. If there is no psychological task with a signal value exceeding the threshold, the classification model (using a CNN-LSTM hybrid model) outputs the probability ranking of the remaining candidate psychological tasks and selects the 1-2 tasks with the highest probability as the current core psychological tasks.

[0036] The psychological tasks in this invention include one or more of the following: reassuring companionship, stress release, calm relaxation, problem sorting, cognitive decomposition, self-exploration, strengths discovery, and dream interpretation. For example, if a user inputs "I've been under a lot of work pressure lately and can't sleep at night," and combines this with the emotional parameters "anxiety, high emotional intensity," the core psychological tasks identified are "stress release" and "calm relaxation."

[0037] Step S3: Pattern Strategy Template Matching and Loading

[0038] Based on the psychological task confidence vector identified in step S2 Match and load the corresponding pattern strategy template from the system's pattern strategy template library. The pattern strategy template library pre-stores templates corresponding to various psychological tasks. Each template contains complete field definitions to ensure the standardization and relevance of responses.

[0039] Specifically, pattern strategy template It includes tone constraints, empathy intensity, question frequency, suggestion intensity, explanation depth, prohibited behaviors, stage goals, output format, maximum progress per round, and mode exit conditions, as well as risk boundary fields. With personality anchor field Among them, the risk boundary field The personality anchor field is used to restrict highly stimulating content (such as avoiding mentioning extremely negative words). Used to maintain personality continuity and serve as a unified personality anchor for subsequent actions. Provides the foundation.

[0040] For example, when the core psychological task is "stress release", the matching pattern strategy template The settings are as follows: tone of voice is gentle and empathetic, empathy intensity is high, questioning frequency is low (avoid excessive questioning), suggestion intensity is low (prioritize listening, do not rush to give advice), explanation depth is shallow, prohibited behaviors include avoiding preaching and mentioning mandatory words such as "you should", the stage goal is to help users release negative emotions, the output format is short sentences and conversational (close to the tone of everyday conversation), the maximum progress per round is 1 (only one emotional point is discussed at a time), and the mode exit condition is when the user's emotional intensity decreases to moderate or below. Step S4: Control parameter generation and current round response generation.

[0041] Based on the mode strategy template loaded in step S3 Generate control parameter vector The definitions and values ​​of each parameter are as follows:

[0042] Empathy intensity coefficient, with a value range of 0-1. The higher the value, the stronger the empathy in the response. For example, the value is 0.8 in the "stress release" mode and 0.5 in the "cognitive decomposition" mode.

[0043] : Frequency of asking questions, ranging from 0 to 1 (0 means no questions asked, 1 means a maximum of 1 question asked). For example, in the "Stress Release" mode, the value is 0.2 (asking questions occasionally to guide the venting), and in the "Problem Sorting" mode, the value is 0.8 (asking questions appropriately to help sort out the problems).

[0044] The suggestion intensity coefficient ranges from 0 to 1. The higher the value, the higher the proportion of suggestions in the response. For example, the value is 0.2 in the "stress release" mode and 0.8 in the "action planning" mode.

[0045] : Explanation depth coefficient, with a value range of 0-1. The higher the value, the more detailed the explanatory content of the reply. For example, the value is 0.7 in the "self-exploration" mode and 0.3 in the "peace of mind companionship" mode.

[0046] User cognitive load coefficient, with a value range of 0-1. The lower the value, the simpler and easier the response, avoiding cognitive burden on the user. For example, the value is 0.2 when the user is in a high arousal state and 0.7 when the user is in a calm state.

[0047] : Response structure coefficient, with a value of 0 (no fixed structure) or 1 (fixed structure). For example, in the "cognitive decomposition" mode, the value is 1 (responding according to the "problem-reason-suggestion" structure), and in the "peace of mind companionship" mode, the value is 0 (responding in a free-talking style).

[0048] Generate control parameter vector Then, the current round's response is generated according to the round progression rules, which specifically include:

[0049] One step at a time: Each round of responses focuses on only one psychological need or emotional point, avoiding the simultaneous advancement of multiple tasks and preventing cognitive burden on the user;

[0050] Ask a maximum of one question at a time: Each round of responses should contain a maximum of one guiding question to avoid excessive questioning that could cause user resistance;

[0051] High arousal / high load states are prioritized for stabilization: when user emotion parameters When the symptoms are high arousal (e.g., anxiety, anger) or high cognitive load, stabilizing responses (e.g., empathic comfort) are generated first, without cognitive decomposition or action planning.

[0052] Do not assign high-cost action tasks when users are in a low-energy, low-motivation state: Do not give complex action tasks (such as "make a weekly plan"), but only give simple and easy-to-implement suggestions (such as "drink a glass of warm water and rest for 5 minutes").

[0053] Only perform analysis or cognitive restructuring when users have a clear sense of purpose: When users clearly express their needs (such as "I want to know how to relieve work pressure"), then carry out tasks such as problem analysis and cognitive restructuring to avoid proactive over-analysis.

[0054] For example, if a user inputs "I've been under a lot of work pressure lately and can't sleep at night," the core psychological task identified is "stress release," and the corresponding pattern strategy template is loaded. Then, the control parameter vector is generated. Following the round progression rules, the current round response is generated as follows: "I can sense that you've been under a lot of pressure lately. It must be very difficult for you to sleep at night. Would you be willing to tell me what exactly is causing you so much stress at work?" This response meets the requirements of high empathy intensity, low question frequency, and low cognitive load, and it only revolves around the task of "stress release".

[0055] Step S5: Flexible mode switching

[0056] Monitoring step S2 generates the psychological task confidence vector User feedback signals When the switching conditions are met, a flexible mode switch is executed, and a unified personality anchor point is used. Maintain consistency in personality to avoid the problem of "style abrupt change" when switching patterns.

[0057] User feedback signals This includes user text feedback (such as "I don't want to talk about this, I want to talk about something else") and interactive behavior feedback (such as quick replies, long periods without replies, and repeatedly sending the same content). By analyzing the feedback signals, we can determine the user's current acceptance of the current mode and changes in their needs.

[0058] Flexible mode switching employs a transition segment control mechanism, such as... Figure 3 (The flowchart for flexible mode switching is shown.) Let the previous mode be... Candidate new modes are The switching condition is: (in (The threshold value is 0.3), and the switching benefit... Greater than the cost of switching .

[0059] Among them, switching benefits For the new model The degree of matching with the user's current mental task, ranging from 0 to 1; the higher the degree of matching, the greater the benefit; switching cost. , The cost calculation function, For pattern difference ( and (the degree of difference in the fields) The time interval between the last switch (the shorter the time interval, the higher the cost). This is the user's stable state coefficient (the more unstable the user's state, the higher the cost).

[0060] Unified personality anchor This includes pronoun habits (e.g., always using "I" as the first person, avoiding switching to "this assistant"), responsiveness (e.g., always remaining gentle and patient, not suddenly becoming cold or harsh), ethical boundaries (e.g., not involving privacy breaches, not inciting negative behavior), and taboo expressions (e.g., avoiding the use of offensive or discriminatory words). Regardless of the mode switching, a unified personality anchor must be followed. This ensures that the language persona aligns with the assistant's identity.

[0061] For example, the previous round pattern was "stress relief" ( The system detected user feedback that "I feel much better now and want to know how to adjust my sleep schedule," and identified "calm and relaxed" as the candidate new mode. ), calculated Switching benefits Switching costs If the switching conditions are met, a flexible mode switch is executed, and the transitional reply is: "I'm so happy to see that you're in a better mood~ Since you want to adjust your schedule, let's talk about how to help you sleep more soundly at night." This transitional reply not only achieves mode switching but also maintains the consistency of tone through a unified personality anchor.

[0062] Step S6: Closed-loop update

[0063] This round of mode ( ), control parameter vector ( User acceptance (through user feedback signals) The evaluation and mode switching records (if any) are written back to the session state database to achieve closed-loop updates and provide data support for subsequent rounds of interaction.

[0064] Specifically, the closed-loop update includes evaluating the effectiveness of the current mode based on subsequent 1-3 rounds of user feedback: if users give positive feedback (such as "What you said is very helpful" or "I feel much better"), the priority of the mode is increased to extend its retention time; if users give negative feedback (such as "I don't want to talk about this" or "This suggestion is useless"), the priority of the mode is reduced to shorten its retention time, and the field parameters of the mode are adjusted (such as reducing empathy intensity and reducing the frequency of questions) to optimize the subsequent response effect.

[0065] For example, if the "stress release" mode is used in this round and the user gives positive feedback in the following 1-3 rounds, then the priority of the "stress release" mode will be increased, and this mode will be prioritized when users have similar psychological needs in the future. If the user gives negative feedback, then the priority of this mode will be decreased, and its empathy intensity coefficient will be adjusted. Question frequency Optimize the pattern template by adjusting parameters such as these.

[0066] System Architecture and Core Module Description

[0067] The cognitive graph-based psychological support dialogue interaction method of the present invention can be implemented through a corresponding system, the architecture of which is as follows: Figure 4 As shown, it includes an information receiving module, a psychological task recognition module, a pattern template matching module, a response generation module, a pattern switching module, a closed-loop update module, and a cognitive graph database. The functions of each module are as follows:

[0068] Information receiving module: Used to receive text input from the user in the current round. User explicit selection mode Historical session status Historical pattern trajectory User basic profile and emotional parameters The above information is then transmitted to the psychological task recognition module.

[0069] Psychological Task Recognition Module: Connects to the information receiving module, used to construct a psychological task confidence vector based on the received multi-dimensional information. Through a dual-channel structure of rule engine and classification model, the system identifies the core psychological task of the current user and transmits the identification results to the pattern template matching module.

[0070] Pattern template matching module: Connects to the psychological task recognition module and is used to match the psychological task confidence vector. Match and load the corresponding pattern strategy template from the cognitive graph database. , pattern strategy template Transmitted to the response generation module;

[0071] Response generation module: Connects to the pattern template matching module, used to match responses based on pattern strategy templates. Generate control parameter vector The system generates the current round's response according to the round progression rules, sends the response to the user, and simultaneously transmits the relevant data to the mode switching module.

[0072] Mode switching module: Connects the response generation module and the information receiving module, used to monitor the confidence vector of the psychological task. User feedback signals Determine if the switching conditions are met; if so, execute the flexible mode switch and use a unified personality anchor point. Maintaining consistency of personality;

[0073] Closed-loop update module: Connects the mode switching module and the cognitive graph database, and is used to write back the current mode, control parameters, user acceptance, and switching records to the session state database to achieve closed-loop update, while optimizing the mode strategy template library;

[0074] Cognitive Graph Database: Used to store basic user profiles, historical conversation data, pattern strategy template library, psychological task tag library, and other data, providing data support for the operation of each module.

[0075] Among them, the psychological task tag library in the cognitive graph database is built based on a large number of psychological support cases and includes eight types of psychological tasks such as reassuring companionship and stress release. Each task corresponds to multiple keyword tags (such as "stress release" corresponding to tags such as "stress", "anxiety", and "insomnia"), which are used to quickly match user needs. The templates in the pattern strategy template library can be continuously optimized based on user feedback and interaction data to ensure the adaptability and effectiveness of the templates.

[0076] According to the cognitive graph-based psychological support dialogue interaction method of the present invention, the present invention upgrades psychological AI from the traditional single general dialogue mode to a computable, designable, and replicable multi-mode control system, effectively breaking through the limitations of the prior art. It can significantly improve the accuracy of response and the degree of demand matching in different psychological task scenarios, and ensure the pertinence and effectiveness of psychological support services.

[0077] According to the cognitive graph-based psychological support dialogue interaction method of the present invention, the present invention, through standardized template field definitions and scientific round-progression rules, performs fine-grained control over the response generation process, effectively reducing the didactic feel, template-based rigidity, and style abruptness during mode switching that are common in existing psychological support dialogues, and is more in line with the core needs of long-term companionship psychological support, thereby improving the consistency and comfort of user experience.

[0078] According to the cognitive graph-based psychological support dialogue interaction method of the present invention, the present invention, with the help of structured result cards and conversation state write-back mechanism, enables the dialogue interaction results to go beyond the text communication level and be deposited into traceable, reusable and optimizable psychological support results, providing data support for subsequent rounds of interaction and long-term psychological state management of users, and realizing a closed-loop upgrade of psychological support services.

[0079] According to the cognitive graph-based psychological support dialogue interaction method of the present invention, the present invention forms a close upstream and downstream technical connection with patents related to emotion parameters and emotion regulation, and builds a logically coherent and functionally complementary technical system, which is convenient for further integration to form a complete product patent chain, laying a solid foundation for subsequent productization and technology expansion.

[0080] Specific application scenario examples

[0081] To further illustrate the implementation effect of the present invention, a detailed description is provided in conjunction with a specific application scenario, as follows: User A (28 years old, white-collar worker), basic user profile For individuals who "face long-term work pressure, have a history of needing stress relief and relaxation, and have a baseline emotional level of moderate anxiety," the specific interaction process is as follows:

[0082] Step S1: The information receiving module receives the text input from user A in the current round. "The project deadline is fast approaching, and I'm working overtime until midnight every day. I feel like I can't hold on much longer," the user explicitly selected the mode. "Stress Relief", Historical Session Status The system responded with empathy in response to the user's previous complaint about work stress, reflecting historical patterns. "Peace of mind companionship → Stress release", emotional parameters "Anxiety, high intensity of emotion";

[0083] Step S2: Based on the above information, the psychological task recognition module constructs a psychological task confidence vector. The signal value for "stress release" is 0.85, and the signal value for "calm relaxation" is 0.7. The rule engine selects "stress release" as a strongly constrained task and determines the current core psychological task as "stress release".

[0084] Step S3: The pattern template matching module matches and loads the pattern strategy template corresponding to "Pressure Relief". The template fields are set as follows: tone constraints: gentle empathy, high empathy intensity, low question frequency, low suggestion intensity, shallow explanation depth, prohibition of preaching, and release of negative emotions as a stage goal;

[0085] Step S4: The response generation module generates the control parameter vector. Following the round-based progression rules, the following reply is generated: "Working overtime until midnight is really tough. The pressure of the project deadline must be overwhelming, right? Don't force yourself. Tell me what's bothering you the most. I'll always be here for you."

[0086] Step S5: User A provides feedback "Thank you, it feels much better to talk about it. I was just worried about not being able to complete the project." (Psychological task confidence vector detected) The signal value corresponding to "problem analysis" in the middle rises to 0.65, calculated as follows: If the switching conditions are not met, maintain the "pressure relief" mode;

[0087] Step S6: The closed-loop update module writes the current "pressure relief" mode, control parameter vector, and user acceptance (positive feedback) back to the session state library, increasing the priority of the "pressure relief" mode.

[0088] Through the above-described interaction process, the present invention can accurately match the user's psychological needs, provide responses that fit the user's state, avoid misaligned responses, maintain personality consistency, and provide users with stable and consistent psychological support services, thus overcoming the shortcomings of the prior art.

[0089] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0090] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A psychological support dialogue interaction method based on cognitive mapping, characterized in that, include: Receive text input from the current user in the current round User explicit selection mode Historical session status Historical pattern trajectory User basic profile and emotional parameters ; Based on the above information, a confidence vector for the psychological task is constructed. Identify the current user's mental task; based on Match and load the corresponding pattern strategy template ;in accordance with Generate control parameter vector Generate the current round's response according to the round progression rules; monitor... User feedback signals When the switching conditions are met, a flexible mode switch is executed, and a unified personality anchor point is used. Maintain consistency of personality; write the current round's mode, parameters, user acceptance, and switching records back to the session state database.

2. The cognitive graph-based psychological support dialogue interaction method according to claim 1, characterized in that, The psychological task confidence vector It is determined by the fusion of three types of signals: explicit signals, semantic signals, and state signals. The psychological tasks mentioned include one or more of the following: providing reassuring companionship, releasing stress, calming and relaxing, sorting out problems, deconstructing cognition, self-exploration, discovering strengths, and dream interpretation.

3. The cognitive graph-based psychological support dialogue interaction method according to claim 1, characterized in that, The psychological task recognition adopts a dual-channel structure of rule engine and classification model: first, the rule engine filters strongly constrained tasks, and then the classification model outputs the probability ranking of the remaining candidate tasks.

4. The cognitive graph-based psychological support dialogue interaction method according to claim 1, characterized in that, Pattern Strategy Template It includes tone constraints, empathy intensity, question frequency, suggestion intensity, explanation depth, prohibited behaviors, stage goals, output format, maximum progress per round, and mode exit conditions.

5. The cognitive graph-based psychological support dialogue interaction method according to claim 4, characterized in that, Pattern Strategy Template It also includes the risk boundary field. With personality anchor field It is used to limit highly stimulating content and maintain personality continuity.

6. The cognitive graph-based psychological support dialogue interaction method according to claim 1, characterized in that, The round-based progression rules include: advancing only one point at a time; asking a maximum of one question at a time; prioritizing stabilization in high-awakening / high-load states; not allocating high-cost action tasks in low-resource states; and only performing sorting or cognitive restructuring when the user has a clear sense of purpose.

7. The cognitive graph-based psychological support dialogue interaction method according to claim 1, characterized in that, The flexible mode switching adopts a transition control mechanism; assuming the previous mode is... Candidate new modes are ,when And switch benefits Greater than the cost of switching At that time, the execution mode is switched.

8. The cognitive graph-based psychological support dialogue interaction method according to claim 7, characterized in that, Switching Cost ,in For pattern differentiation, The time interval between the last switchover. This represents the user's stable state coefficient.

9. The cognitive graph-based psychological support dialogue interaction method according to claim 1, characterized in that, Unified personality anchor This includes pronoun usage, response temperature, ethical boundaries, and taboo expressions, ensuring that the linguistic personality and assistant identity remain consistent before and after the switch.

10. The cognitive graph-based psychological support dialogue interaction method according to claim 1, characterized in that, Closed-loop updates include evaluating the effectiveness of the model based on subsequent 1-3 rounds of user feedback, reducing the model's priority based on negative feedback, and increasing the model's retention time based on positive feedback.

Images