Intelligent psychological tutoring system based on fused knowledge graph

Abstract

The invention discloses an intelligent psychological tutoring system based on a fused knowledge graph. The intelligent psychological tutoring system comprises a data acquisition module, a graph reasoning module, a strategy retrieval module and a content output module, the data acquisition module is used for acquiring multi-modal psychological data input by a user through an interactive interface; the map reasoning module is used for inputting the multi-modal psychological data into a pre-constructed fusion knowledge map for multi-level semantic mapping, performing multi-hop reasoning in the fusion knowledge map, and determining a current user state node; the strategy retrieval module is used for retrieving a tutoring strategy entity with a direct causal relationship in the fused knowledge graph according to the current user state node to obtain a target tutoring strategy entity; and the content output module is used for outputting the tutoring content corresponding to the target tutoring strategy entity to the interactive interface.

Description

TECHNICAL FIELD

[0001] The present application relates to an intelligent psychological counseling system based on a fusion knowledge graph. BACKGROUND

[0002] The field of psychological counseling is attempting to provide support with the aid of artificial intelligence technology. Existing technical solutions usually rely on pre-defined question and answer scripts, keyword matching-based dialogue engines, or simple queries using a single knowledge base to provide standardized psychological support content.

[0003] However, such traditional systems have obvious limitations such as rigid response patterns and difficulty in understanding the complex and dynamic psychological state of users. SUMMARY

[0004] The present application provides an intelligent psychological counseling system based on a fusion knowledge graph to solve the technical problems of rigid response patterns and difficulty in understanding the complex and dynamic psychological state of users in existing technologies.

[0005] To solve the above technical problems, the present application provides the following technical solutions:

[0006] The present application provides an intelligent psychological counseling system based on a fusion knowledge graph, comprising:

[0007] a data acquisition module, a graph reasoning module, a strategy retrieval module, and a content output module;

[0008] The data acquisition module is configured to acquire multi-modal psychological data input by a user through an interactive interface, the multi-modal psychological data including at least one of the following: text dialogue content, voice emotion features, and user-initiated emotion labels;

[0009] The graph reasoning module is configured to input the multi-modal psychological data into a pre-constructed fusion knowledge graph for multi-level semantic mapping and multi-hop reasoning in the fusion knowledge graph to determine a current user state node;

[0010] The strategy retrieval module is configured to retrieve a counseling strategy entity having a direct causal relationship in the fusion knowledge graph according to the current user state node to obtain a target counseling strategy entity;

[0011] The content output module is configured to output counseling content corresponding to the target counseling strategy entity to the interactive interface.

[0012] In an optional embodiment, the fusion knowledge graph comprises:

[0013] a static structure layer storing psychological concept entities, symptom entities, counseling strategy entities, and causal relationships therebetween;

[0014] The dynamic interaction layer stores user entities, dialogue turn entities, and temporal associations reflecting changes in user state; wherein the dynamic interaction layer is constructed based on historical session sequences.

[0015] In one optional embodiment, the strategy retrieval module is specifically used for:

[0016] Retrieve tutoring strategy entities with direct causal relationships in the static structure layer;

[0017] The retrieved tutoring strategy entities are sorted based on the feedback ratings of the same user for similar tutoring strategies recorded in the dynamic interaction layer.

[0018] In an optional embodiment, the content output module is further configured to:

[0019] The user status nodes generated in this session, the tutoring strategies adopted, and the real-time user feedback are updated to the dynamic interaction layer as new temporal associations.

[0020] In one optional embodiment, the graph inference module includes:

[0021] A text processing unit is configured to match at least one of the symptom entities in the static structure layer based on the text dialogue content and through entity links.

[0022] The speech processing unit is used to combine the speech emotion features to perform weight correction on the matched symptom entities;

[0023] The tag processing unit is used to calculate the state transition probability along the temporal association based on the user's historical state nodes in the dynamic interaction layer, and to determine the current user state node.

[0024] In one optional embodiment, the voice processing unit is specifically used for:

[0025] From the aforementioned voice emotion features, voice spectrum features and voice dynamic features are extracted, wherein the voice spectrum features are used to characterize the user's emotional valence, and the voice dynamic features are used to characterize the user's emotional arousal level.

[0026] Using a pre-trained emotion state mapping model, the confidence impact factors of the speech spectral features and speech dynamic features on the severity of different symptom entities are calculated.

[0027] Based on the confidence level influence factor, the initial weights associated with the matched symptom entities are corrected to obtain the weight-corrected symptom entities.

[0028] In an optional embodiment, the causal relationships stored in the static structure layer are constructed through the following steps:

[0029] Structured knowledge triplets were extracted from authoritative psychology textbooks, diagnostic manuals, and published research papers on psychological counseling to form an initial set of causal relationships;

[0030] The initial set of causal relationships is imported into a pre-trained graph neural network for representation learning to obtain the embedding vectors of each entity;

[0031] Based on the embedded vector, the probability of a causal relationship between any two entities is calculated. Associations with a probability higher than a first preset threshold and not belonging to the initial causal relationship set are added as new implicit causal relationships to the static structure layer.

[0032] In an optional embodiment, the graph inference module includes a construction and update unit for constructing and updating the temporal associations through the following steps:

[0033] Define the current user status node as a status event;

[0034] Establish a directed temporal association between two consecutive state events;

[0035] The weights of the established temporal associations are adjusted based on the time interval between the two consecutive state events and the user's feedback rating of the tutoring strategy.

[0036] In an optional embodiment, the content output module is further configured to push a standardized assessment scale to the interactive interface within a preset time interval after the tutoring content is output.

[0037] Receive the user's completion results on the assessment scale, and quantify the completion results as an effectiveness score for this coaching session;

[0038] The effect score, the target tutoring strategy entity, and the current user status node are stored as a historical session sequence in the dynamic interaction layer.

[0039] This invention provides an intelligent psychological counseling system based on a fused knowledge graph. By constructing a two-layer knowledge graph that integrates static general knowledge and dynamic interaction history, and utilizing multimodal data for multi-hop reasoning, it achieves a deep, dynamic, and personalized understanding of the user's psychological state. This system overcomes the shortcomings of traditional solutions, such as rigid responses and isolated knowledge, intelligently generating counseling strategies highly adapted to the user's current state and historical feedback, thereby significantly improving the accuracy, personalization, and long-term effectiveness of psychological support. Simultaneously, the system possesses self-optimization capabilities, continuously learning and enhancing its adaptability through interaction. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a schematic diagram of the system module architecture of the intelligent psychological counseling system based on fused knowledge graph provided in an embodiment of the present invention. Detailed Implementation

[0042] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The following embodiments are implemented based on the technical solution of the present invention, and provide detailed implementation methods and specific operation processes, but the scope of protection of the present invention is not limited to the following embodiments.

[0043] In this invention, the terms "in one possible embodiment," "exemplary," or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "in one possible embodiment," "exemplary," or "for example" in this invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "in one possible embodiment," "exemplary," or "for example" is intended to present the relevant concepts in a specific manner.

[0044] Figure 1 This is a schematic diagram of the system module architecture of the intelligent psychological counseling system based on fused knowledge graphs provided in an embodiment of the present invention. Figure 1 As shown, a psychological counseling intelligent system 10 based on a fusion knowledge graph includes:

[0045] Data acquisition module 101, graph reasoning module 102, strategy retrieval module 103, and content output module 104;

[0046] The data acquisition module is used to acquire multimodal psychological data input by the user through the interactive interface. The multimodal psychological data includes at least one of the following: text dialogue content, voice emotional features, and emotion tags actively marked by the user.

[0047] The graph reasoning module is used to input the multimodal psychological data into a pre-constructed fusion knowledge graph for multi-level semantic mapping, and to perform multi-hop reasoning in the fusion knowledge graph to determine the current user state node;

[0048] The strategy retrieval module is used to retrieve tutoring strategy entities with direct causal relationships in the fused knowledge graph based on the current user state node, and obtain the target tutoring strategy entity.

[0049] The content output module is used to output the tutoring content corresponding to the target tutoring strategy entity to the interactive interface.

[0050] For example, the various modules of the system can exchange data via API interfaces. Furthermore, the data acquisition module can integrate speech recognition and emotion annotation components to convert the raw dialogue into structured multimodal data packets. The graph reasoning module can call the embedded graph computing engine to execute reasoning algorithms on the fused knowledge graph loaded into memory.

[0051] In one optional embodiment, the fused knowledge graph includes:

[0052] The static structure layer stores psychological concept entities, symptom entities, counseling strategy entities, and the causal relationships between them;

[0053] The dynamic interaction layer stores user entities, dialogue turn entities, and temporal associations reflecting changes in user state; wherein the dynamic interaction layer is constructed based on historical session sequences.

[0054] For example, the static structure layer can be stored as an RDF graph in the Neo4j graph database, containing entities such as "anxiety symptoms" and "cognitive behavioral therapy," as well as relationships such as "cause" and "relief." The dynamic interaction layer can be stored as a sequence graph, with each user having an independent subgraph. Nodes record the state inferred after each session, and the attributes on the edges record the context of state transitions (such as triggering strategies and feedback scores).

[0055] In one optional embodiment, the strategy retrieval module is specifically used for:

[0056] Retrieve tutoring strategy entities with direct causal relationships in the static structure layer;

[0057] The retrieved tutoring strategy entities are sorted based on the feedback ratings of the same user for similar tutoring strategies recorded in the dynamic interaction layer.

[0058] For example, the strategy retrieval module first queries the static structure layer for all strategy entities that have a "recommendation guidance" relationship with the "current user state node". Then, it accesses the dynamic layer to query the user's historical feedback ratings (e.g., 1-5 points) for different strategies under a certain state, and calculates the average or weighted score to sort the candidate strategies in descending order.

[0059] In an optional embodiment, the content output module is further configured to:

[0060] The user status nodes generated in this session, the tutoring strategies adopted, and the real-time user feedback are updated to the dynamic interaction layer as new temporal associations.

[0061] For example, after the tutoring content is output, the system collects real-time user feedback (such as "helpful" button clicks and satisfaction slider ratings) through an interactive interface. A background thread of the content output module packages this feedback data, the user state node "User State Node_A" determined in this instance, and the adopted strategy "Strategy_X" into a JSON object. This object is then used as a new temporal edge from the user's previous state node to "User State Node_A" via the graph database's update API.

[0062] In one optional embodiment, the graph inference module includes:

[0063] A text processing unit is configured to match at least one of the symptom entities in the static structure layer based on the text dialogue content and through entity links.

[0064] The speech processing unit is used to combine the speech emotion features to perform weight correction on the matched symptom entities;

[0065] The tag processing unit is used to calculate the state transition probability along the temporal association based on the user's historical state nodes in the dynamic interaction layer, and to determine the current user state node.

[0066] For example, the text processing unit uses the BERT model for named entity recognition and linking. The speech processing unit uses an open-source toolkit to extract Mel spectrogram features and fundamental frequency envelopes from the speech. The tag processing unit implements a Hidden Markov Model whose state transition matrix is ​​dynamically adjusted based on the user's historical state sequence in the dynamic interaction layer to calculate the most likely new state.

[0067] In one optional embodiment, the voice processing unit is specifically used for:

[0068] From the aforementioned voice emotion features, voice spectrum features and voice dynamic features are extracted, wherein the voice spectrum features are used to characterize the user's emotional valence, and the voice dynamic features are used to characterize the user's emotional arousal level.

[0069] Using a pre-trained emotion state mapping model, the confidence impact factors of the speech spectral features and speech dynamic features on the severity of different symptom entities are calculated.

[0070] Based on the confidence level influence factor, the initial weights associated with the matched symptom entities are corrected to obtain the weight-corrected symptom entities.

[0071] For example, speech spectral features can be specifically statistical measures (such as mean and variance) of Mel-frequency cepstral coefficients, used to determine whether a sound is bright (positive) or dull (negative). Speech dynamic features can be specifically the standard deviation and range of the fundamental frequency over a period of time. The emotion state mapping model can be a simple neural network whose input is a concatenated vector of the above features, and whose output is an adjustment coefficient. For example, when speech features indicate "high arousal, negative valence," the model outputs a coefficient greater than 1.0, amplifying the weights associated with the relevant symptom entities.

[0072] In an optional embodiment, the causal relationships stored in the static structure layer are constructed through the following steps:

[0073] Structured knowledge triplets were extracted from authoritative psychology textbooks, diagnostic manuals, and published research papers on psychological counseling to form an initial set of causal relationships;

[0074] The initial set of causal relationships is imported into a pre-trained graph neural network for representation learning to obtain the embedding vectors of each entity;

[0075] Based on the embedded vector, the probability of a causal relationship between any two entities is calculated. Associations with a probability higher than a first preset threshold and not belonging to the initial causal relationship set are added as new implicit causal relationships to the static structure layer.

[0076] For example, in the construction phase, triples are first extracted from the literature using an information extraction tool. Then, a pre-trained graph neural network model on a general knowledge graph is loaded using a library such as DGL or PyTorch Geometric, and fine-tuned with an initial set of causal relationships. Finally, the cosine similarity of the embedding vectors between all entity pairs is calculated, and associations above a threshold (e.g., 0.85) that are not explicitly documented in the literature are added to the graph as valuable implicit causal relationships.

[0077] In an optional embodiment, the graph inference module includes a construction and update unit for constructing and updating the temporal associations through the following steps:

[0078] Define the current user status node as a status event;

[0079] Establish a directed temporal association between two consecutive state events;

[0080] The weights of the established temporal associations are adjusted based on the time interval between the two consecutive state events and the user's feedback rating of the tutoring strategy.

[0081] For example, the specific construction and update steps are as follows: When a user completes the Nth session, the system creates a state event node S_N. Then, in the Cypher query language of the graph database, a command is executed to create a relation from S_{N-1} to S_N: NEXT. The weight attribute of this relation is calculated according to the formula weight = base_score / ln(time_interval + 1), where base_score is the effectiveness score of this tutoring session, and time_interval is the number of days between two sessions.

[0082] In an optional embodiment, the content output module is further configured to push a standardized assessment scale to the interactive interface within a preset time interval after the tutoring content is output.

[0083] Receive the user's completion results on the assessment scale, and quantify the completion results as an effectiveness score for this coaching session;

[0084] The effect score, the target tutoring strategy entity, and the current user status node are stored as a historical session sequence in the dynamic interaction layer.

[0085] For example, 24 hours after the system outputs the coaching content, it automatically pushes a questionnaire to the APP message center. After the user completes the questionnaire, the system calculates the raw score according to the standard scoring rules and normalizes it to an effectiveness score of 0-10. Subsequently, the system appends the record (user ID, timestamp, status node, strategy entity, effectiveness score) to the user's dynamic interaction log table and updates the attributes of the corresponding node in the graph at the same time.

[0086] Based on the above structure, the intelligent psychological counseling system based on a fused knowledge graph provided by this invention constructs a two-layer knowledge graph that integrates static general knowledge and dynamic interaction history, and utilizes multimodal data for multi-hop reasoning to achieve a deep, dynamic, and personalized understanding of the user's psychological state. This system overcomes the shortcomings of traditional solutions, such as rigid responses and isolated knowledge, and intelligently generates counseling strategies highly adapted to the user's current state and historical feedback, thereby significantly improving the accuracy, personalization, and long-term effectiveness of psychological support. Simultaneously, the system possesses self-optimization capabilities, continuously learning and enhancing its adaptability through interaction.

[0087] Furthermore, it should be noted that the present invention can be provided as a method, apparatus, or computer program product. Therefore, embodiments of the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, embodiments of the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code.

[0088] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0089] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing terminal equipment to cause a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0090] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0091] Finally, it should be noted that the above description represents a preferred embodiment of the present invention. It should be pointed out that although preferred embodiments have been described, those skilled in the art, once they understand the basic inventive concept of the present invention, can make various improvements and modifications without departing from the principles described herein. These improvements and modifications should also be considered within the scope of protection of the present invention. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the embodiments of the present invention.

Claims

1. A psychological counseling intelligent system based on fused knowledge graphs, characterized in that, include: Data acquisition module, graph reasoning module, strategy retrieval module, and content output module; The data acquisition module is used to acquire multimodal psychological data input by the user through the interactive interface. The multimodal psychological data includes at least one of the following: text dialogue content, voice emotional features, and emotion tags actively marked by the user. The graph reasoning module is used to input the multimodal psychological data into a pre-constructed fusion knowledge graph for multi-level semantic mapping, and to perform multi-hop reasoning in the fusion knowledge graph to determine the current user state node; The strategy retrieval module is used to retrieve tutoring strategy entities with direct causal relationships in the fused knowledge graph based on the current user state node, and obtain the target tutoring strategy entity. The content output module is used to output the tutoring content corresponding to the target tutoring strategy entity to the interactive interface.

2. The intelligent psychological counseling system based on fused knowledge graph as described in claim 1, characterized in that, The fused knowledge graph includes: The static structure layer stores psychological concept entities, symptom entities, counseling strategy entities, and the causal relationships between them; The dynamic interaction layer stores user entities, dialogue turn entities, and temporal associations reflecting changes in user state; wherein the dynamic interaction layer is constructed based on historical session sequences.

3. The intelligent psychological counseling system based on fused knowledge graph as described in claim 2, characterized in that, The strategy retrieval module is specifically used for: Retrieve tutoring strategy entities with direct causal relationships in the static structure layer; The retrieved tutoring strategy entities are sorted based on the feedback ratings of the same user for similar tutoring strategies recorded in the dynamic interaction layer.

4. The intelligent psychological counseling system based on fused knowledge graph as described in claim 3, characterized in that, The content output module is also used for: The user status nodes generated in this session, the tutoring strategies adopted, and the real-time user feedback are updated to the dynamic interaction layer as new temporal associations.

5. The intelligent psychological counseling system based on fused knowledge graph as described in claim 2, characterized in that, The graph reasoning module includes: A text processing unit is configured to match at least one of the symptom entities in the static structure layer based on the text dialogue content and through entity links. The speech processing unit is used to combine the speech emotion features to perform weight correction on the matched symptom entities; The tag processing unit is used to calculate the state transition probability along the temporal association based on the user's historical state nodes in the dynamic interaction layer, and to determine the current user state node.

6. The intelligent psychological counseling system based on fused knowledge graph as described in claim 5, characterized in that, The speech processing unit is specifically used for: From the aforementioned voice emotion features, voice spectrum features and voice dynamic features are extracted, wherein the voice spectrum features are used to characterize the user's emotional valence, and the voice dynamic features are used to characterize the user's emotional arousal level. Using a pre-trained emotion state mapping model, the confidence impact factors of the speech spectral features and speech dynamic features on the severity of different symptom entities are calculated. Based on the confidence level influence factor, the initial weights associated with the matched symptom entities are corrected to obtain the weight-corrected symptom entities.

7. The intelligent psychological counseling system based on fused knowledge graph as described in claim 2, characterized in that, The causal relationships stored in the static structure layer are constructed through the following steps: Structured knowledge triplets were extracted from authoritative psychology textbooks, diagnostic manuals, and published research papers on psychological counseling to form an initial set of causal relationships; The initial set of causal relationships is imported into a pre-trained graph neural network for representation learning to obtain the embedding vectors of each entity; Based on the embedded vector, the probability of a causal relationship between any two entities is calculated. Associations with a probability higher than a first preset threshold and not belonging to the initial causal relationship set are added as new implicit causal relationships and added to the static structure layer.

8. The intelligent psychological counseling system based on fused knowledge graph as described in claim 2, characterized in that, The graph inference module includes a construction and update unit, used to construct and update the temporal associations through the following steps: Define the current user status node as a status event; Establish a directed temporal association between two consecutive state events; The weights of the established temporal associations are adjusted based on the time interval between the two consecutive state events and the user's feedback rating of the tutoring strategy.

9. The intelligent psychological counseling system based on fused knowledge graph as described in claim 1, characterized in that, The content output module is also used to push the standardized assessment scale to the interactive interface within a preset time interval after the tutoring content is output. Receive the user's completion results on the assessment scale, and quantify the completion results as an effectiveness score for this coaching session; The effect score, the target tutoring strategy entity, and the current user status node are stored as a historical session sequence in the dynamic interaction layer.

Images