A method and system for assessing a psychological state

Abstract

The application discloses a psychological state evaluation method and system, and the method comprises the following steps: S1, inputting dialogue text of a dialogue window into a pre-trained language model, the language model extracting cognitive entities and attribution relationships between the cognitive entities from the dialogue text, and mapping the cognitive entities to a high-dimensional vector space to obtain feature vectors of the cognitive entities; S2, taking the cognitive entities as nodes, extracting core nodes from all nodes through a density peak clustering algorithm; S3, taking the core nodes as vertices and the attribution relationships between the core nodes as edges, constructing a weighted cognitive topology graph, calculating global structure entropy of the graph and degree centrality of the core nodes; S4, updating the weighted cognitive topology graph, the global structure entropy and the degree centrality in real time along with input of the dialogue text; based on the global structure entropy and the degree centrality, determining whether narrative topology collapse occurs; if the narrative topology collapse occurs, determining that the psychological state is abnormal. The application improves the scientificity and objectivity of psychological state evaluation.

Description

Technical Field

[0001] This invention relates to the fields of psychology and artificial intelligence, and in particular to a method and system for assessing psychological states. Background Technology

[0002] Current psychological health assessment methods based on natural language processing mostly focus on shallow sentiment computing, that is, classifying user input into positive, negative, neutral, and other emotional categories through text classification models, which can only achieve a general overview analysis of the text content.

[0003] Such methods can only perform surface-level emotion recognition and content statistics, ignoring the underlying cognitive network structure of the human mind and failing to reveal the essential changes in psychological states. Clinical and cognitive science research shows that psychological problems such as depression and anxiety are not simply fluctuations in emotional indicators, but rather external manifestations of rigid cognitive structures and limited cognitive pathways.

[0004] Existing technologies have significant limitations: related algorithms cannot quantify the connectivity, flexibility, and structural complexity of an individual's mind, making it difficult to explain the formation mechanism of psychological rigidity at the cognitive network level. Currently, the industry still lacks an effective computational method that can transform unstructured dialogue data into a mathematically meaningful cognitive topology map, objectively measure the flexibility of cognitive networks using structural entropy, and thus achieve automatic identification of narrative topology collapse.

[0005] Therefore, traditional psychological assessments rely on subjective descriptions and experience-based judgments, lacking unified, rigorous, and reproducible mathematical quantitative evidence, making it difficult to meet the needs for precise, digital, and interpretable mental health assessments.

[0006] The statements herein provide only background information in relation to this invention and do not necessarily constitute prior art. Summary of the Invention

[0007] The purpose of this invention is to provide a method and system for assessing psychological states, effectively solving the technical problems of traditional assessments, such as over-reliance on the counselor's subjective experience, simplistic assessment standards, and insufficient accuracy of assessment results. This invention transforms unstructured dialogues into dynamically changing weighted cognitive topology maps in real time. Based on the global structural entropy and degree centrality of the weighted cognitive topology map, it achieves an objective determination of narrative topological collapse phenomena. This invention accurately characterizes the connectivity, flexibility, and structural complexity of an individual's mind through cognitive topology maps, establishing a scientifically rigorous and unified quantitative standard for psychological state assessment, greatly improving the accuracy of assessment results.

[0008] To achieve the above objectives, the present invention provides a method for assessing psychological state, comprising the following steps: S1. Input the dialogue text of the dialogue window into the pre-trained language model. The language model extracts cognitive entities and attribution relationships between the cognitive entities from the dialogue text, and maps the cognitive entities to a high-dimensional vector space to obtain the feature vectors of the cognitive entities. S2. Using the cognitive entity as a node, extract the core node from all nodes using the density peak clustering algorithm; S3. Construct a weighted cognitive topology graph with the core nodes as vertices and the attribution relationships between the core nodes as edges; calculate the global structural entropy of the weighted cognitive topology graph and the degree centrality of the core nodes; S4. As dialogue text is input in each dialogue window, the weighted cognitive topology graph, the global structural entropy, and the degree centrality are updated in real time; based on the global structural entropy and the degree centrality, it is determined whether narrative topology collapse has occurred; if it has occurred, the psychological state is determined to be abnormal.

[0009] Optionally, the method further includes the step of: S5. The counselor modifies the weighted cognitive graph based on the intervention strategy; calculates the new global structural entropy and node degree centrality for the modified weighted cognitive topology graph, and verifies the effectiveness of the intervention strategy based on the calculation results.

[0010] Optionally, the nodes include a first node to an Nth node, where N is the total number of cognitive entities extracted by the language model. Step S2 includes: S21, for the first m The local density is calculated from the eigenvectors of the nodes. and relative distance ; ; ; in, For the first m The eigenvector of the node and the first n Cosine distance between the feature vectors of nodes This is the cutoff distance threshold; For the first n Local density of the eigenvectors of a node; ; ; S22, if and If the product of the first and second digits is greater than the set first threshold, then the first digit... m The node is the core node.

[0011] Optionally, the weighted cognitive topology graph is a directed graph, which includes M core nodes, namely the first core node to the Mth core node; make Indicates from the first i The core node points to the first j The weights of the edges in the core nodes. ; in, For the first i Core nodes and the first j The core nodes exchange information within the dialog window. ; For the first i Core nodes and the first j The co-occurrence probability of core nodes in the dialog window; , For the first i Core node, the first j The probability that a core node appears alone in the dialog window.

[0012] Optional, let This represents the weighted cognitive topology map. , , Let represent the set of core nodes and the set of edges in the weighted cognitive topology graph, respectively; This represents the global structural entropy. ; in, core node v The weighting degree, The total edge weights of the weighted cognitive topology graph are given.

[0013] Optional, let , This represents the global structural entropy before and after the weighted cognitive topology map is updated. ; in, After the weighted cognitive topology graph is updated, the core node set consists of newly added core nodes and core nodes whose weighting degree has changed.

[0014] Optionally, in step S4, narrative topological collapse is determined to have occurred when both of the following conditions are met simultaneously: Condition 1: The global structural entropy continuously decreases over T consecutive time windows, and the global structural entropy of each of these T time windows is lower than a preset healthy baseline threshold. ; Condition 2: In the cognitive topology graph, the proportion of the number of connected edges of the maximum degree node to the total number of edges in the graph is greater than a preset proportion threshold.

[0015] Optionally, the core nodes include a first core node, whose corresponding cognitive entity is "self"; in step S5, the counselor modifies the weighted cognitive map including: Modify the edge weights of the weighted cognitive graph, and / or add intervention nodes to the weighted cognitive topology graph, connect the intervention nodes to the first core node and set corresponding edge weights; the intervention nodes are positive nodes.

[0016] The present invention also provides a psychological state assessment system for implementing the psychological state assessment method as described in the present invention, comprising: The entity extraction module is used to extract cognitive entities and attribution relationships between cognitive entities from the dialogue text, and generate corresponding feature vectors for the cognitive entities. The core node mining module is used to convert the cognitive entities into corresponding nodes and cluster all nodes to obtain core nodes. The graph construction module is used to calculate edge weights for the core nodes and construct a weighted cognitive topology graph. The computation module is used to calculate the global structural entropy of the weighted cognitive topology graph and the degree centrality of the core nodes; The diagnostic module determines whether narrative topological collapse has occurred based on the global structural entropy and the degree centrality of the core nodes.

[0017] Optionally, the psychological state assessment system further includes: The graph rendering and display module is used to render the weighted cognitive topology graph into a three-dimensional force-guided graph and display it in a visualization interface. The interface interaction editing module is used to receive the editing operation performed by the user on the three-dimensional force guidance map in the visualization interface, and provide the editing operation to the map construction module to update the cognitive topology map; the map rendering and display module refreshes the three-dimensional force guidance map in the visualization interface based on the updated cognitive topology map.

[0018] Compared with the prior art, the beneficial effects of the present invention include: This invention automatically transforms unstructured dialogue text into a weighted cognitive topology graph, dynamically adjusting the graph's topology as the dialogue window updates. By accurately characterizing the connectivity, flexibility, and structural complexity of an individual's mind through this cognitive topology graph, and further using the global structural entropy and degree centrality of this weighted cognitive topology graph, it achieves an objective determination of narrative topological collapse phenomena. This invention establishes a scientifically rigorous and unified quantitative standard for psychological state assessment, significantly improving the accuracy of assessment results. This invention quantifies changes in psychological state at the cognitive network level, breaking through the limitations of traditional methods that rely solely on superficial emotional polarity analysis, and enhancing the objectivity and quantifiability of psychological assessment.

[0019] After assessing the psychological state, the support counselor modifies the cognitive topology graph (including modifying edge weights and adding intervention nodes) and recalculates the global structural entropy and degree centrality based on the intervention strategy. This allows for a quantitative comparison of the network structure changes before and after the intervention, intuitively verifying the effectiveness of the intervention strategy. It provides quantifiable and traceable feedback for psychological intervention, enhancing the scientific rigor and repeatability of the intervention plan.

[0020] This invention extracts cognitive entities as nodes from unstructured dialogues and uses a density peak clustering algorithm to extract core cognitive nodes from all nodes, reducing node redundancy, highlighting key cognitive entities, and reducing graph size and computational overhead. The selection criteria for core nodes are clear and reproducible, avoiding subjective selection bias and ensuring the stability of subsequent graph construction and global structural entropy calculation.

[0021] In this invention, edge weights are integrated with co-occurrence frequency and point mutual information. Co-occurrence frequency characterizes the association strength, while point mutual information delineates the association reliability, making the weighted cognitive topology graph more closely resemble the characteristics of real cognitive associations. The edge weight calculation is entirely data-driven, without any subjective human parameters, improving the objectivity and consistency of the graph's topology.

[0022] This invention uses an incremental update method to calculate the global structural entropy, making local corrections only for newly added nodes and nodes with changed weights, without requiring a full graph recalculation. This significantly reduces computation time in real-time dialogue scenarios, improves the real-time performance of graph updates and global structural entropy calculations, and adapts to the needs of online continuous dialogue evaluation.

[0023] This invention employs a dual-condition approach—"continuous decrease in global structural entropy + over-centralization of negative core nodes"—to jointly determine narrative topology collapse. This approach fully considers the overall flexibility and connectivity of cognitive networks, reducing the risk of misjudgment based on a single indicator. The judgment criteria are unified, quantifiable, and reproducible, improving the stability and accuracy of abnormal psychological state identification.

[0024] This invention renders a weighted cognitive topology map as a three-dimensional force-directed graph, intuitively presenting cognitive entities, attribution relationships, and weight differences. It also supports interactive editing of the map within a visual interface, allowing counselors to intuitively observe and adjust cognitive structures and receive immediate quantitative feedback, thus improving the system's interactivity, interpretability, and clinical usability. Attached Figure Description

[0025] Figure 1 This is a flowchart of a psychological state assessment method in an embodiment of the present invention.

[0026] Figure 2 This is a flowchart of a psychological state assessment method in one embodiment of the present invention.

[0027] Figure 3 This is a schematic diagram illustrating how unstructured dialogue text is converted into feature vectors and semantic clustering is performed using the LLM model and DPC algorithm in an embodiment of the present invention.

[0028] Figure 4 A schematic diagram of the structure of a weighted cognitive topology map of a healthy mind.

[0029] Figure 5 A schematic diagram of the structure of a weighted cognitive topology map in which the mind is in a state of narrative topological collapse.

[0030] Figure 6 This is a schematic diagram of the structure of the psychological state assessment system in an embodiment of the present invention.

[0031] Figure 7 This is a schematic diagram illustrating the interaction between a counselor and a psychological state assessment system through a visual interface, as described in an embodiment of the present invention. Detailed Implementation

[0032] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. The advantages and features of the present invention will become clearer from the following description. It should be noted that the drawings are in a very simplified form and use non-precise proportions, used only to facilitate and clearly illustrate the embodiments of the present invention. Please refer to the drawings to make the objectives, features, and advantages of the present invention more apparent and understandable. It should be understood that the structures, proportions, sizes, etc., depicted in the accompanying drawings are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed in the specification, and are not intended to limit the implementation conditions of the present invention. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in proportions, or adjustments to the size, without affecting the effects and objectives achieved by the present invention, should still fall within the scope of the technical content disclosed in the present invention.

[0033] Currently, most mental health assessments based on natural language processing rely on superficial sentiment analysis, which can only identify positive, negative, and neutral emotions, failing to reflect deeper changes in mental state. Clinical research indicates that problems such as depression and anxiety are essentially due to rigid cognitive structures and limited thought pathways. Existing technologies cannot quantify the connectivity and complexity of an individual's mind. Traditional mental state assessment methods primarily depend on the subjective experience of counselors, lacking unified, rigorous, and reproducible quantitative evidence, and thus cannot meet the needs for accurate, numerical, and interpretable assessments.

[0034] This invention provides a method and system for assessing psychological states, effectively solving the technical problems of traditional assessments, such as over-reliance on counselors' subjective experience, singular assessment standards, and insufficient accuracy of assessment results. This invention transforms unstructured dialogues into dynamically changing weighted cognitive topology maps in real time. Based on the global structural entropy and degree centrality of the weighted cognitive topology maps, it achieves objective determination of narrative topological collapse phenomena. This invention accurately characterizes the connectivity, flexibility, and structural complexity of an individual's mind through cognitive topology maps, establishing a scientifically rigorous and unified quantitative standard for psychological state assessment, greatly improving the accuracy of assessment results.

[0035] Figure 1 This is a flowchart of a psychological state assessment method according to an embodiment of the present invention. Figure 1 As shown, the psychological state assessment method includes the following steps: S1. Input the dialogue text of the dialogue window into the pre-trained language model. The language model extracts cognitive entities and attribution relationships between the cognitive entities from the dialogue text, and maps the cognitive entities to a high-dimensional vector space to obtain the feature vectors of the cognitive entities.

[0036] The dialogue window can be defined by the user, for example, one dialogue window for every 10 sentences or one dialogue window for every 500 characters; this invention does not impose any restrictions. The dialogue text of this invention is unstructured dialogue text. Unstructured dialogue refers to natural spoken dialogue without a fixed format, fixed questions, or fixed options, where words flow freely from the speaker's thoughts.

[0037] According to Weiner's attribution theory, the attribution relationships include: 1) Internal stable attribution: Attributing the outcome of an event to its own persistent and stable characteristics. For example, "I didn't do well on the exam because I'm not capable enough."

[0038] 2) Internal instability attribution: Attributing the outcome of an event to its own temporary and variable factors. For example, "I didn't do well on the exam because I didn't try hard enough this time." 3) External stable attribution: Attributing the outcome of an event to long-term, fixed external conditions. For example, "I didn't do well on the exam because the questions were too difficult." 4) External instability attribution: Attributing the outcome of an event to external, accidental, and temporary factors. For example, "I didn't do well on the exam because I was unlucky."

[0039] The language model used in this embodiment is an LLM (Large Language Model). First, the pre-trained LLM model is fed instructions containing psychologically specific prompts, such as: "Please identify the cognitive subject, object, and attribution relationship between them in the following text, and extract the causal link." Then, the LLM model generates cognitive entities based on the input dialogue text. Example: When the input text is "I messed up this exam, I'm destined to be a loser," the LLM model outputs one cognitive entity as "self" and another as "exam." The edge between these two cognitive entities (also called a relation edge) is an "internal stable attribution," and the causal link is "failed the exam -> self-denial (loser)." How to train the LLM model is existing technology and will not be elaborated upon in this invention.

[0040] By using the Embedding layer of the LLM model, cognitive entities are mapped to a high-dimensional vector space to obtain the corresponding feature vectors. This is existing technology and will not be elaborated upon in this invention.

[0041] S2. Using cognitive entities as nodes, the core nodes are extracted from all nodes through the Density Peak Clustering (DPC) algorithm.

[0042] In this embodiment, within a certain dialog window, the LLM model extracts N cognitive entities. These N cognitive entities are respectively designated as the first node to the Nth node. Step S2 includes: S21, for the first m Calculate the local density of the eigenvectors of a node and relative distance ; .

[0043] .

[0044] in, For the first m The eigenvector of the node and the first n Cosine distance between the feature vectors of nodes This is the cutoff distance threshold. For the first n Local density of the eigenvectors of a node. .

[0045] .

[0046] S22, if and If the product of the first and second digits is greater than the set first threshold, then the second digit... m The node is the core node.

[0047] Figure 3 This is a schematic diagram illustrating how, in an embodiment of the present invention, unstructured dialogue text is converted into feature vectors and semantic clustering is performed using an LLM model and a DPC algorithm. For example... Figure 3 As shown, the left side represents the visitor's original unstructured expression (such as "I'm useless," "I'm a failure," "Nobody loves me"), which is transformed into a dense feature vector in a high-dimensional space through the LLM embedding layer. Subsequently, a density peak clustering algorithm merges high-dimensional feature vectors that are close in distance and have consistent semantic orientation into a core node, thereby eliminating redundant expressions in natural language and completing the graph nodeization. This invention reduces node redundancy, highlights key cognitive entities, and reduces the graph size and computational overhead (detailed later). The core node selection criteria are clear and reproducible, avoiding subjective selection bias and ensuring the stability of subsequent graph construction and global structural entropy calculation.

[0048] S3. Construct a weighted cognitive topology graph with core nodes as vertices and attribution relationships between core nodes as edges; calculate the global structural entropy and degree centrality of the core nodes in the weighted cognitive topology graph.

[0049] The weighted cognitive topology graph in this embodiment is a directed graph, which includes M core nodes, namely the first core node to the Mth core node.

[0050] make Indicates from the first i The core node points to the first j The weights of the edges in the core nodes. ; ; For the first i Core nodes and the first j The co-occurrence probability of core nodes in the dialog window; , For the first i Core node, the first j The probability that a core node appears alone in the dialog window.

[0051] For the first i Core nodes and the first jPointwise Mutual Information (PMI) of core nodes in the dialog window is used to determine whether two core nodes are truly related. The greater the PMI between two core nodes, the higher the correlation between them, such as "exam" and "failure"; the smaller the PMI, the lower the correlation between them, such as "exam" and "eat".

[0052] For example: Scenario A: The core nodes "exam" and "failure" appear frequently and simultaneously, indicating high mutual information between them. → The edge weight between these two core nodes is very high, representing a strong negative perception.

[0053] Scenario B: The core nodes "exam" and "eat" occasionally appear together. The co-occurrence probability is not low, but the mutual information between nodes is very low. The edge weight is lowered by the mutual information between nodes to avoid misjudging it as a strong association.

[0054] In this invention, edge weights are integrated with co-occurrence frequency and point mutual information. Co-occurrence frequency characterizes the association strength, while point mutual information delineates the association reliability, making the weighted cognitive topology graph more closely resemble the characteristics of real cognitive associations. The edge weight calculation is entirely data-driven, without any subjective human parameters, improving the objectivity and consistency of the graph's topology.

[0055] make This represents a weighted cognitive topology graph. , , Let represent the set of core nodes and the set of edges in the weighted cognitive topology graph, respectively; let Represents global structural entropy. ; in, core node v The weighting degree, The total edge weights of the weighted cognitive topology graph are given.

[0056] core node v Degree centrality, for core nodes v The proportion of connected edges to the total number of edges in the graph is calculated using the following formula: .

[0057] Narrative topology collapse is commonly used to diagnose anxiety and depressive tendencies in individuals and to measure mental flexibility. It can be simply understood as the content of a person's speech / thoughts gradually narrowing, solidifying, and locking down from a diverse, open, and flexible state. All thoughts revolve around a few negative core nodes, and the cognitive network becomes rigid, singular, and inflexible. Higher global structural entropy indicates more diverse thinking content and greater mental flexibility; higher global structural entropy indicates more rigid thinking. Figure 3 As shown, the dialogue gradually revolves around the core node "self-blame." At this point, the global structural entropy is usually low, and narrative topological collapse may occur.

[0058] Figure 4 and Figure 5 The comparison illustrates the differences in cognitive structure between a healthy mind and a topologically collapsed mind. For example... Figure 4 As shown, the characteristics of a healthy state are high global structural entropy and multidimensional connectivity, where the core node "self" can freely connect to multiple independent core nodes such as "work," "family," and "achievement." Figure 5 As shown, the characteristics of the collapsed state are low global structural entropy and deadlock. The edge weights between core nodes such as "self", "work" and "future" and the negative core node "I am not worthy" are relatively high, and the degree centrality of the negative core node "I am not worthy" is also relatively high. As a result, the cognitive paths of core nodes such as "self", "work" and "future" are all locked by a single negative core node, losing the connectivity to other cognitive paths.

[0059] S4. As dialogue text is input in each dialogue window, the weighted cognitive topology graph, global structural entropy, and degree centrality of each core node are updated in real time. Based on the global structural entropy and the degree centrality of the core nodes, it is determined whether narrative topology collapse has occurred. If it has occurred, the psychological state is determined to be abnormal.

[0060] make , This represents the global structural entropy before and after the weighted cognitive topology map is updated. ; in, After the weighted cognitive topology graph is updated, the core node set consists of newly added core nodes and core nodes whose weighting degree has changed.

[0061] If only one dialogue relationship edge is added (its two core nodes are...) , ),but The computational complexity is from Down to .

[0062] This invention uses an incremental update method to calculate the global structural entropy. Based on the local approximate update formula of the first-order Taylor expansion, it only makes local corrections for newly added nodes and nodes with changed weights, without recalculating the entire graph. This significantly reduces the computation time in real-time dialogue scenarios, improves the real-time performance of graph updates and global structural entropy calculations, and adapts to the needs of online continuous dialogue evaluation.

[0063] In this embodiment, narrative topology collapse is determined to have occurred when both of the following conditions are met simultaneously: Condition 1: The global structural entropy continuously decreases over T consecutive time windows, and the global structural entropy in each of these T time windows is lower than the preset healthy baseline threshold. .

[0064] Introducing a continuous decline over T consecutive time windows is to filter out short-term fluctuations and capture long-term trends; a healthy baseline threshold is introduced. This is to establish objective standards to distinguish between normal fluctuations and pathological collapses.

[0065] Condition 2: In the cognitive topology graph, the proportion of the number of connected edges of the maximum degree node to the total number of edges in the graph is greater than a preset proportion threshold.

[0066] This invention quantifies the degree of cognitive black hole by the proportion of edges in the maximum-degree node. When the proportion of edges in a certain core node (such as "I am not worthy") exceeds 80% of the total number of edges in the graph (this is only an example and not a limitation of this invention), it indicates that almost all cognitive paths point to this negative core node, forming a state of "mental blockage." It's like a person who, no matter what topic they discuss, will eventually circle back to the negative belief "I am useless," with all thoughts sucked in by this single belief, unable to extend outward.

[0067] This invention employs a dual-condition approach—"continuous decrease in global structural entropy + over-centralization of negative core nodes"—to jointly determine narrative topology collapse. This approach fully considers the overall flexibility and connectivity of cognitive networks, reducing the risk of misjudgment based on a single indicator. The judgment criteria are unified, quantifiable, and reproducible, improving the stability and accuracy of abnormal psychological state identification.

[0068] like Figure 2 As shown, in one embodiment of the present invention, a diagnostic report will also be output based on the determination result.

[0069] In summary, this invention can automatically convert unstructured dialogue text into a weighted cognitive topology graph, and dynamically adjust the topological structure of this graph as the dialogue window updates. By accurately characterizing the connectivity, flexibility, and structural complexity of an individual's mind through this cognitive topology graph, and further using the global structural entropy and degree centrality of this weighted cognitive topology graph, it achieves an objective determination of narrative topological collapse phenomena. This invention establishes a scientifically rigorous and unified quantitative standard for psychological state assessment, greatly improving the accuracy of assessment results. This invention quantifies changes in psychological state at the cognitive network level, breaking through the limitations of traditional methods that rely solely on superficial emotional polarity analysis, and enhancing the objectivity and quantifiability of psychological assessment.

[0070] like Figure 1 As shown, in this embodiment, the psychological state assessment method further includes the following steps: S5. The counselor modifies the weighted cognitive graph based on the intervention strategy, calculates the new global structural entropy and node degree centrality for the modified weighted cognitive topology graph, and verifies the effectiveness of the intervention strategy based on the calculation results.

[0071] In step S5, the counselor modifies the weighted cognitive map, including: Modify the edge weights of the weighted cognitive graph, and / or add an intervention node (which is a positive node, such as a small thing that was recently accomplished) to the weighted cognitive topology graph, connect the intervention node to the first core node (whose corresponding cognitive entity is "self") and set the corresponding edge weights.

[0072] After determining the psychological state, this invention supports counselors in modifying the cognitive topology map according to the intervention strategy and recalculating the global structural entropy and degree centrality. It can quantitatively compare the changes in the network structure of the map before and after the intervention, intuitively verify the effectiveness of the intervention strategy, provide quantifiable and traceable feedback for psychological intervention, and improve the scientificity and repeatability of the intervention plan.

[0073] This invention also provides a psychological state assessment system for implementing the psychological state assessment method described in this invention, such as... Figure 6 As shown, it includes: entity extraction module 10, core node mining module 20, graph construction module 30, calculation module 40, diagnosis module 50, graph rendering and display module 60, and interface interaction and editing module 70.

[0074] The entity extraction module 10 uses a large language model to extract cognitive entities and attribution relationships between cognitive entities from the dialogue text, and generates corresponding feature vectors for the cognitive entities.

[0075] The core node mining module 20 is used to convert cognitive entities into corresponding nodes and cluster all nodes to obtain core nodes.

[0076] If there is an attribution relationship between two core nodes, then there is a relation edge between the two core nodes. The graph construction module 30 calculates the edge weight for the relation edge to realize the construction of a weighted cognitive topology graph for the core nodes (each vertex in the graph corresponds to a core node).

[0077] The calculation module 40 is used to calculate the global structural entropy for the weighted cognitive topology graph and to calculate the degree centrality of each core node. In this embodiment, the graph construction module 30 calculates and provides the structural data (including edge set, vertex set, and edge weight) of the weighted cognitive topology graph to the cognitive graph database, and the cognitive graph database provides the above structural data to the calculation module 40.

[0078] The diagnostic module 50 determines whether narrative topological collapse has occurred based on global structural entropy and the degree centrality of core nodes.

[0079] The graph rendering and display module 60 is connected to the graph construction module 30 and is used to render the weighted cognitive topology graph into a three-dimensional force-guided graph and display it in the visualization interface.

[0080] In this embodiment, a three-dimensional force-directed graph is rendered using a force-directed graph algorithm. Degree centrality is represented by the size of core nodes, and edge weight is represented by the thickness of edges, achieving intuitive visualization of the cognitive topology. Simultaneously, based on the calculation results, negative core nodes that are locked to other core nodes are automatically highlighted. In the three-dimensional force-directed graph, simulated "electrostatic repulsion" exists between vertices, while relational edges are equivalent to "spring attraction."

[0081] The interface interaction editing module 70 receives editing operations performed by the counselor on the three-dimensional force guidance map in the visualization interface, and provides the editing operations to the atlas construction module 30 to update the structural data of the cognitive topology map. The calculation module 40 recalculates the global structural entropy and degree centrality based on the updated atlas topology. The diagnosis module 50 re-evaluates whether the psychological state is abnormal based on the recalculated global structural entropy and degree centrality. The atlas rendering and display module 60 obtains the updated structural data from the atlas construction module 30 and simultaneously refreshes the three-dimensional force guidance map in the visualization interface, allowing the counselor to intuitively observe and adjust the cognitive structure and see quantitative feedback immediately, thus improving the system's interactivity, interpretability, and clinical usability.

[0082] like Figure 7As shown in the interactive example, when "collapse" is detected, the negative core node representing "I am not worthy" on the interface, due to its extremely high weight, exhibits enormous mass and gravity, tightly attracting core nodes such as "Self" and "Future" together (manifesting as a deadlocked black hole). The consultant clicks "Simulate Intervention" on the interface, manually adds a new core node "Recent Successful Small Things," and connects it to the core node "Self." The calculation module 40 calculates the structural entropy increment in real time based on the newly added edge, and the visualization interface then shows the process of the core node "Self" breaking free from the "black hole's" gravity and the cognitive topology network structure re-expanding, thereby helping the consultant verify the effectiveness of the intervention strategy.

[0083] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus 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 apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0084] The artificial intelligence-based psychological assessment and medical auxiliary diagnosis solutions involved in the embodiments of the present invention strictly comply with national regulations on data security, personal information protection and medical industry management in terms of data collection, processing, model training and application, and meet the requirements of medical ethics review.

[0085] The diagnostic data involved in the embodiments of this invention (such as, but not limited to, behavioral observation data, clinical consultation records, physiological and biochemical indicators, etc.) are all derived from legal and compliant publicly available de-identified datasets, or legally collected with the subject's (or their legal guardian's) explicit consent after full disclosure. During the data collection process, this invention does not involve concealment, deception, or unauthorized acquisition of sensitive personal information.

[0086] In the embodiments of this invention, when collecting personal biometric information such as facial micro-expressions, eye gaze trajectories, voice acoustic features, or body movements of subjects for auxiliary assessment in psychological-related diagnoses, the collection is conducted in a specific environment where the subjects are fully informed and voluntarily cooperate. The subjects have been clearly informed of the specific purpose and scope of use of the data collection and have obtained their specific consent. This invention does not contain any covert, non-intrusive, or unauthorized image and sound capture methods.

[0087] If the embodiments of this invention involve multimodal medical data fusion (e.g., simultaneously inputting text questionnaire data, facial expression videos, and voice recordings of subjects), this system adopts a strict permission isolation and encrypted transmission mechanism when processing such data. The original data of different modalities are physically or logically isolated during the feature extraction stage. The artificial intelligence algorithm model only performs fusion calculations on the extracted abstract feature vectors. The entire calculation process and the final output diagnostic results cannot be reversed to reconstruct the original multimodal joint information of the subject, thus eliminating the privacy leakage risk that may be caused by cross-modal data association.

[0088] Before using the aforementioned data for model training or actual diagnosis, all data underwent rigorous de-identification and anonymization processes. This eliminated identity-related biases while retaining the disease characteristic dimensions required for psychological assessment. Furthermore, the algorithmic decision-making mechanism of this invention is built upon objective psychopathological characteristics. During model training, it fully considers the diversity, representativeness, and balance of the samples, and avoids setting biased, stereotypical, or discriminatory decision rules based on non-disease factors such as the subject's gender, age, occupation, or social status.

[0089] The systems, methods, and devices claimed in this invention for psychological assessment and medical auxiliary diagnosis, etc., whose output analysis results, risk probabilities, or classification labels are intended solely for clinical reference by legally qualified psychiatrists, psychotherapists, and other medical professionals. This invention serves only as a clinical auxiliary assessment tool and cannot be used directly as the final basis for clinical diagnosis, nor can it automatically generate drug treatment prescriptions or psychological intervention instructions. The final psychological state assessment and treatment plan must be made by a professional physician through interviews and comprehensive clinical examinations. The implementation of this invention does not replace the diagnostic and treatment actions of professional medical personnel.

[0090] Although the present invention has been described in detail through the preferred embodiments above, it should be understood that the above description should not be considered as a limitation of the present invention. Various modifications and substitutions to the present invention will be apparent to those skilled in the art after reading the above description. Therefore, the scope of protection of the present invention should be defined by the appended claims.

Claims

1. A method for assessing psychological state, characterized in that, Including the following steps: S1. Input the dialogue text of the dialogue window into the pre-trained language model. The language model extracts cognitive entities and attribution relationships between the cognitive entities from the dialogue text, and maps the cognitive entities to a high-dimensional vector space to obtain the feature vectors of the cognitive entities. S2. Using the cognitive entity as a node, extract the core node from all nodes using the density peak clustering algorithm; S3. Construct a weighted cognitive topology graph with the core nodes as vertices and the attribution relationships between the core nodes as edges; calculate the global structural entropy of the weighted cognitive topology graph and the degree centrality of the core nodes; S4. As the dialogue text is input in each dialogue window, the weighted cognitive topology graph, the global structural entropy, and the degree centrality are updated in real time. Based on the global structural entropy and the degree centrality, determine whether narrative topological collapse has occurred; If this occurs, it indicates an abnormal psychological state.

2. The psychological state assessment method as described in claim 1, characterized in that, The method further includes the following steps: S5. The counselor modifies the weighted cognitive graph based on the intervention strategy; calculates the new global structural entropy and node degree centrality for the modified weighted cognitive topology graph, and verifies the effectiveness of the intervention strategy based on the calculation results.

3. The psychological state assessment method as described in claim 1, characterized in that, The nodes include the first node to the Nth node, where N is the total number of cognitive entities extracted by the language model. Step S2 includes: S21, for the first m The local density is calculated from the eigenvectors of the nodes. and relative distance ; ； ； in, For the first m The eigenvector of the node and the first n Cosine distance between the feature vectors of nodes This is the cutoff distance threshold; For the first n Local density of the eigenvectors of a node; ; ； S22, if and If the product of the first and second digits is greater than the set first threshold, then the first digit... m The node is the core node.

4. The psychological state assessment method as described in claim 1, characterized in that, The weighted cognitive topology graph is a directed graph, which includes M core nodes, namely the first core node to the Mth core node; make Indicates from the first i The core node points to the first j The weights of the edges in the core nodes. ； in, For the first i Core nodes and the first j The core nodes exchange information within the dialog window. ； For the first i Core nodes and the first j The co-occurrence probability of core nodes in the dialog window; , For the first i Core node, the first j The probability that a core node appears alone in the dialog window.

5. The psychological state assessment method as described in claim 1, characterized in that, make This represents the weighted cognitive topology map. , , Let represent the set of core nodes and the set of edges in the weighted cognitive topology graph, respectively; This represents the global structural entropy. ； in, core node v The weighting degree, The total edge weights of the weighted cognitive topology graph are given.

6. The psychological state assessment method as described in claim 5, characterized in that, make , This represents the global structural entropy before and after the weighted cognitive topology map is updated. ； in, After the weighted cognitive topology graph is updated, the core node set consists of newly added core nodes and core nodes whose weighting degree has changed.

7. The psychological state assessment method as described in claim 1, characterized in that, In step S4, narrative topological collapse is determined to have occurred when both of the following conditions are met simultaneously: Condition 1: The global structural entropy continuously decreases over T consecutive time windows, and the global structural entropy of each of these T time windows is lower than a preset healthy baseline threshold. ; Condition 2: In the cognitive topology graph, the proportion of the number of connected edges of the maximum degree node to the total number of edges in the graph is greater than a preset proportion threshold.

8. The psychological state assessment method as described in claim 2, characterized in that, The core nodes include a first core node, whose corresponding cognitive entity is "self"; In step S5, the counselor modifies the weighted cognitive map, including: Modify the edge weights of the weighted cognitive graph, and / or add intervention nodes to the weighted cognitive topology graph, connect the intervention nodes to the first core node and set the corresponding edge weights; The intervention node is a positive node.

9. A psychological state assessment system for implementing the psychological state assessment method as described in any one of claims 1 to 8, characterized in that, include: The entity extraction module is used to extract cognitive entities and attribution relationships between cognitive entities from the dialogue text, and generate corresponding feature vectors for the cognitive entities. The core node mining module is used to convert the cognitive entities into corresponding nodes and cluster all nodes to obtain core nodes. The graph construction module is used to calculate edge weights for the core nodes and construct a weighted cognitive topology graph. The computation module is used to calculate the global structural entropy of the weighted cognitive topology graph and the degree centrality of the core nodes; The diagnostic module determines whether narrative topological collapse has occurred based on the global structural entropy and the degree centrality of the core nodes.

10. The psychological state assessment system as described in claim 9, characterized in that, Also includes: The graph rendering and display module is used to render the weighted cognitive topology graph into a three-dimensional force-guided graph and display it in a visualization interface. The interface interaction editing module is used to receive the editing operation performed by the user on the three-dimensional force guidance map in the visualization interface, and provide the editing operation to the map construction module to update the cognitive topology map; the map rendering and display module refreshes the three-dimensional force guidance map in the visualization interface based on the updated cognitive topology map.

Images