Psychological counseling assisting method, device and equipment

Abstract

The invention is suitable for the technical field of psychological counseling, and particularly relates to a psychological counseling assisting method, device and equipment, and the method comprises the steps: obtaining basic data, behavior data, physiological signals and input data of a student; performing multi-modal data fusion on the basic data, the behavior data, the physiological signal and the input data to obtain a psychological state feature vector; based on the basic data, the behavior data, the physiological signals and the input data, whether psychological early warning is triggered or not is determined; if it is determined that the psychological early warning is triggered, generating a target dredging scheme based on the psychological state feature vector; and obtaining short-term data and long-term data of the students, and optimizing the generation process of the target guidance scheme based on the short-term data and the long-term data of the students. According to the method, the pertinence and effectiveness of psychological counseling are improved, reference can be provided for psychological teachers, and counseling strategies of students can be continuously optimized.

Description

Technical Field

[0001] The present application belongs to the field of psychological counseling technology, and in particular relates to psychological counseling auxiliary methods, devices and equipment. Background Art

[0002] Many students face psychological challenges such as academic pressure, social anxiety, identity confusion, and mood swings. Some may experience negative emotions due to test anxiety, low self-esteem, or family pressure, which can even affect their studies and daily lives. In these situations, psychological counseling can help students relieve anxiety, overcome social phobias, alleviate academic stress, and cultivate positive thinking, thereby improving their resilience and adaptability.

[0003] Currently, many schools have a relatively small number of psychological counselors who are responsible for the mental health of a large number of students. This makes it difficult to provide in-depth, one-on-one counseling. Instead, a single counselor may serve hundreds or even thousands of students, making it difficult to fully address each individual's mental state. Due to the limited number of counselors, to quickly understand and provide counseling, counselors typically use generic approaches, such as counseling courses and teacher-led conversations. These approaches are difficult to tailor to individual circumstances, potentially leading to varying student acceptance of the same intervention approach, and some approaches may be ineffective or even counterproductive.

[0004] To sum up, when providing psychological counseling to students, there is a problem that the psychological counseling for students is not targeted due to limited resources and the use of general psychological counseling programs. Summary of the Invention

[0005] The embodiments of the present application provide a psychological counseling auxiliary method, device and equipment, which can solve the problem in related technologies that when providing psychological counseling to students, the psychological counseling of students is not targeted due to limited resources and the use of general psychological counseling solutions.

[0006] In a first aspect, an embodiment of the present application provides a psychological counseling assistance method, comprising: Obtaining basic data, behavioral data, physiological signals, and input data of students; wherein the basic data includes academic performance data, psychological assessment results, family background information, and consultation records; the behavioral data includes classroom behavior and social data; the physiological signals include heart rate data and skin conductance data; and the input data includes voice data or text data; Performing multimodal data fusion on the basic data, the behavioral data, the physiological signal, and the input data to obtain a psychological state feature vector; Determining whether to trigger a psychological warning based on the basic data, the behavioral data, the physiological signal, and the input data; If it is determined that the psychological warning is triggered, generating a targeted counseling plan based on the psychological state feature vector; Obtain the short-term data and long-term data of the student, and optimize the generation process of the target counseling plan based on the short-term data and long-term data of the student; wherein the short-term data is the heartbeat data of the student after counseling, and the long-term data includes the academic performance data and social data of the student after counseling.

[0007] The above technical solutions in the embodiments of the present application have at least the following technical effects: The psychological counseling assistance method provided in this application first obtains students' basic data (grade data, psychological assessment results, family background information, and consultation records), behavioral data (classroom behavior and social data), physiological signals (heart rate data and skin conductance data), and input data (voice data or text data). It then performs multimodal data fusion on the basic data, behavioral data, physiological signals, and input data to obtain a psychological state feature vector. Based on the basic data, behavioral data, physiological signals, and input data, it then determines whether a psychological warning is triggered. If a psychological warning is triggered, a targeted counseling plan is generated based on the psychological state feature vector. Finally, the student's short-term data (heart rate data after counseling) and long-term data (grade data and social data after counseling) are obtained, and the generation process of the targeted counseling plan is optimized based on the student's short-term and long-term data. This method uses multi-source data to provide richer contextual information, improves the accuracy and sensitivity of psychological problem identification, and can automatically determine whether a warning is triggered, reducing the burden of manual screening and improving response speed. This method considers students' psychological profiles, behavioral patterns, and family backgrounds to generate customized intervention plans. This approach can provide a reference for psychology teachers and improve the relevance and effectiveness of counseling, avoiding one-size-fits-all interventions. This method provides technical support for intelligent mental health monitoring and intervention, contributing to the construction of an intelligent psychological counseling support system with a closed-loop perception-decision-feedback system, enabling the continuous optimization of student counseling strategies.

[0008] In a second aspect, an embodiment of the present application provides a psychological counseling auxiliary device, comprising: an acquisition unit, configured to acquire basic data, behavioral data, physiological signals, and input data of the student; wherein the basic data includes academic performance data, psychological assessment results, family background information, and consultation records; the behavioral data includes classroom behavior and social data; the physiological signals include heartbeat data and skin conductance data; and the input data includes voice data or text data; a multimodal data fusion unit, configured to perform multimodal data fusion on the basic data, the behavioral data, the physiological signal, and the input data to obtain a psychological state feature vector; a psychological warning triggering unit, configured to determine whether to trigger a psychological warning based on the basic data, the behavioral data, the physiological signal, and the input data; a target guidance scheme generating unit, configured to generate a target guidance scheme based on the psychological state feature vector if it is determined that the psychological warning is triggered; An optimization unit is used to obtain the short-term data and long-term data of the student, and optimize the generation process of the target counseling plan based on the short-term data and long-term data of the student; wherein the short-term data is the heartbeat data of the student after counseling, and the long-term data includes the academic performance data and social data of the student after counseling.

[0009] In a third aspect, an embodiment of the present application provides a psychological counseling auxiliary device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in any one of the embodiments of the first aspect when executing the computer program.

[0010] It can be understood that the beneficial effects of the second to third aspects mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0012] Figure 1 This is a flowchart of a psychological counseling assistance method provided by an embodiment of the present application; Figure 2 This is a schematic diagram of the implementation process of generating a target counseling solution in the psychological counseling auxiliary method provided in the embodiment of the present application; Figure 3 It is a structural diagram of the psychological counseling auxiliary equipment provided in an embodiment of the present application. DETAILED DESCRIPTION

[0013] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0014] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0015] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0016] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0017] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0018] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0019] In related technologies, many schools have a relatively small number of psychological counselors who are responsible for the mental health of a large number of students. This makes it difficult to provide in-depth, one-on-one counseling. Instead, a single counselor may serve hundreds or even thousands of students, making it difficult to fully address each individual's mental state. Due to the limited number of counselors, to quickly understand and provide counseling, counselors typically use generic methods, such as counseling courses and teacher interviews. These methods are difficult to tailor to individual circumstances, and can result in different student responses to the same intervention, making some methods ineffective or even counterproductive.

[0020] To address the above-mentioned issues, embodiments of the present application provide a psychological counseling assistance method, device, and apparatus. This method first obtains a student's basic data (grade data, psychological assessment results, family background information, and consultation records), behavioral data (classroom behavior and social data), physiological signals (heart rate data and skin conductance data), and input data (voice data or text data). Multimodal data fusion is then performed on the basic data, behavioral data, physiological signals, and input data to obtain a psychological state feature vector. Based on the basic data, behavioral data, physiological signals, and input data, a determination is made as to whether a psychological warning has been triggered. If a psychological warning has been triggered, a targeted counseling plan is generated based on the psychological state feature vector. Finally, the student's short-term data (heart rate data after counseling) and long-term data (grade data and social data after counseling) are obtained, and the generation process of the targeted counseling plan is optimized based on the student's short-term and long-term data. This method uses multi-source data to provide richer contextual information, improves the accuracy and sensitivity of psychological problem identification, and can automatically determine whether a warning has been triggered, reducing the burden of manual screening and improving response speed. This method considers students' psychological profiles, behavioral patterns, and family backgrounds to generate customized intervention plans. This approach can provide a reference for psychology teachers and improve the relevance and effectiveness of counseling, avoiding one-size-fits-all interventions. This method provides technical support for intelligent mental health monitoring and intervention, contributing to the construction of an intelligent psychological counseling support system with a closed-loop perception-decision-feedback system, enabling the continuous optimization of student counseling strategies.

[0021] The psychological counseling assistance method provided in the embodiment of the present application can be applied to a psychological counseling assistance device. In this case, the psychological counseling assistance device is the executor of the psychological counseling assistance method provided in the embodiment of the present application. The embodiment of the present application does not impose any restrictions on the specific type of the psychological counseling assistance device.

[0022] For example, the psychological counseling auxiliary device can be a cellular phone, a mobile phone, a tablet computer, a wearable device, an augmented reality (AR) / virtual reality (VR) device, a laptop computer, an ultra-mobile personal computer (UMPC), a netbook, a desktop computer, a handheld device with wireless communication capabilities, a computing device or other processing device connected to a wireless modem, a computer, a laptop computer, customer premise equipment (CPE) and / or other devices for communicating on a wireless system and a next-generation communication system, such as a mobile terminal in a 5G network or a mobile terminal in a future evolved public land mobile network (PLMN), etc.

[0023] In order to better understand the psychological counseling assistance method provided in the embodiment of the present application, the specific implementation process of the psychological counseling assistance method provided in the embodiment of the present application is exemplarily introduced below.

[0024] Figure 1 A schematic flow chart of a psychological counseling assistance method provided in an embodiment of the present application is shown. The psychological counseling assistance method includes: S100: Acquire the student's basic data, behavioral data, physiological signals, and input data. Basic data includes academic performance, psychological assessment results, family background information, and counseling records; behavioral data includes classroom behavior and social interaction data; physiological signals include heart rate data and skin conductance data; and input data includes voice data or text data.

[0025] Understandably, grade data can come from a school's internal grade management system or student information system, which records midterm and final grades, assignment scores, and classroom performance ratings. Grade data can reflect a student's academic ability and progress. In psychological early warning, grade data can help determine whether a student is experiencing psychological burdens due to factors such as academic pressure.

[0026] Psychological assessments can be conducted by professional psychologists or school psychologists using standardized psychological assessment tools (e.g., Beck Depression Inventory, Self-Rating Anxiety Inventory, Social Adjustment Scale, etc.). The results of psychological tests can be scale scores or reports, and can be used to reflect students' emotional state and mental health levels, such as anxiety and depression.

[0027] Family background information can include family economic status, parents' education level, family structure, etc. Family background information is related to students' psychological state. For example, factors such as family pressure and parents' education methods may have an impact on students' mental health.

[0028] Counseling records are records of consultations between students and school psychologists or psychological professionals. These records can include the student's emotions, concerns, problem analysis, and proposed solutions. These records help understand the student's past psychological issues, emotional fluctuations, and coping processes, and can help predict potential psychological crises.

[0029] Classroom behavior data can be collected from teachers, classroom management systems, learning management systems, etc. Classroom behavior data can reflect students' concentration and engagement, and can reveal whether students show psychological signs such as anxiety, depression, and powerlessness in class.

[0030] Social data includes students' social interactions with classmates, social media, and participation in school activities. Social data can reflect students' social status. Isolation and social difficulties may be precursors to psychological problems, and high-quality social behavior is associated with a good psychological state.

[0031] Heart rate data is monitored by wearable devices (such as smart watches and fitness trackers) or hospital monitoring equipment. Heart rate data can reflect students' physiological response states. When students feel anxious, stressed, or emotionally unstable, their heart rate will change.

[0032] Skin conductance data is collected through galvanic skin response sensors and wearable devices (such as smartwatches and skin sensors). Skin conductance data can reflect students' emotional fluctuations. For example, when they are anxious or nervous, their skin conductance increases, providing physiological information about their psychological state.

[0033] Input data consists of student confessions, either voice or text. Information such as tone, speed, and intonation in the voice can be used to assess the student's emotional state. Emotions such as anxiety and anger are often expressed in voice. Text input by students can also be used to assess emotional tendencies and psychological states. Negative emotions, feelings of isolation, or anxiety may be evident in their text, providing support for psychological early warning.

[0034] For example, student performance data can be directly obtained by interfacing with the school's academic affairs system or performance management system. This data can be obtained through regular mental health questionnaires, online psychological assessment systems, or face-to-face psychological assessments. Family background information can be obtained through student registration forms, family interviews, questionnaires, and other methods. Consultation records can be obtained through offline consultation records or online psychological counseling platforms.

[0035] Classroom behavior can be obtained through teacher observation and recording of student behavior in class, or through classroom management systems (such as attendance, classroom interaction, and homework submission data). Social data can be obtained through social platforms (such as WeChat groups and Facebook groups), school social activity records, and student interaction systems. Social network analysis tools can be used to analyze students' social circles, interaction frequency, and emotional interactions.

[0036] Heart rate data and skin conductance data can be obtained through wearable devices worn by students (such as Apple Watch, Fitbit, etc.). The devices monitor data such as heart rate and changes in skin conductance.

[0037] Voice data is recorded by voice recording devices (such as mobile phones, computers, voice assistants, etc.). Voice data can be converted into text for analysis using speech recognition technology. Natural language processing (NLP) technology can be used to perform sentiment analysis on the text input by students.

[0038] The data obtained in this step can provide a basis for subsequent psychological warning and generation of counseling plans.

[0039] S200, performing multimodal data fusion on basic data, behavioral data, physiological signals and input data to obtain a psychological state feature vector.

[0040] For example, grades, psychological assessment results, family background information and consultation records are structured data (numerical or categorical), and the data can be processed by filling in missing values, standardization or normalization, etc., to facilitate the comparability of the data in terms of numerical range.

[0041] Classroom behavior and social data are time series data (such as students' participation in class, number of interactions, speaking time, etc.). The sliding window method can be used to sample, smooth, or denoise time series data.

[0042] Heartbeat data and skin conductance data are continuous time series data, which may require noise removal (such as using filtering technology), feature extraction (such as peak value, average value, standard deviation, etc.) and normalization.

[0043] Natural language processing (NLP) technology can be used to perform word segmentation and word embedding (such as Word2Vec or BERT) on text; voice data can be converted into numerical representation through audio processing (such as feature extraction, MFCC, etc.).

[0044] Features can be extracted from various types of data. For example, emotional state features (such as anxiety, depression, stress, etc.) can be extracted from input data; physiological condition features (such as heartbeat and skin conductance fluctuations) can be extracted from physiological signals; social state features (such as social interaction frequency and isolation) can be extracted from social data; and academic status features (such as grades and learning participation) can be extracted from grade data.

[0045] Early fusion can be used to fuse the features of various data types at the data level (that is, directly concatenating the feature vectors of all data sources into a single large vector). This method is suitable for situations where the features of various data types are relatively consistent and the dimensions of different data sources are consistent. Late fusion can be used to model each type of data separately (for example, using a classifier or regression model to process each type of data separately), and then combining the respective prediction results to obtain a psychological state feature vector. This method is suitable for situations where the feature extraction process for each data source is complex and the correlation between data sources is weak. For complex multimodal data, methods such as multilayer perceptrons (MLPs), convolutional neural networks (CNNs), or long short-term memory networks (LSTMs) can be used to extract features from each data point. Information from different modalities can then be integrated through shared layers or cross-modal learning layers to obtain a psychological state feature vector.

[0046] The implementation process of this step involves the processing and feature extraction of multiple data sources, and integrating these features into a psychological state feature vector through multimodal fusion technology. This vector will provide a basis for psychological early warning and support subsequent psychological intervention and counseling plans.

[0047] In one possible implementation, see Figure 2 , S200, performs multimodal data fusion on basic data, behavioral data, physiological signals and input data to obtain a psychological state feature vector, including: S210, constructing a basic portrait vector based on the basic data.

[0048] For example, the slope and standard deviation of the score sequence in the score data can be calculated, and the error rate can be calculated based on the wrong questions in the score data. The slope, standard deviation and error rate values ​​are combined to form a score feature vector. For example, a student's math score sequence is [75, 82, 68, 90] (four exams). , The slope calculation formula is The standard deviation calculation formula is According to the slope calculation formula, the slope of the student's math score series is 3.1, the standard deviation is 8.16, the student's error rate for knowledge point A is 30%, and the error rate for knowledge point B is 5%. Then the student's score eigenvector is [3.1, 8.16, 0.3, 0.05].

[0049] The student's anxiety and depression scores can be extracted from the psychological assessment results, and the anxiety and depression scores can be normalized based on the normal score. The normalized anxiety and depression scores are combined to form an assessment feature vector. For example, if the student's anxiety score is 2.8 and the depression score is 1.2, the normal value is 1.5, and the anxiety score is normalized: , normalized depression scores: , then the student's evaluation feature vector is [2.6, -0.6].

[0050] Family structures can be categorized into two-parent families, single-parent families, and left-behind families. Family structures can be coded, with students assigned a 1 for one family structure and a 0 for another. A student's income data can be normalized based on the highest and lowest family incomes reported by the school. For example, if a student's family structure is single-parent with a monthly income of 5,000 yuan, and the school's income range is 2,000 to 20,000 yuan, then the structure vector is [0, 1, 0]. The normalized income is (5,000-2,000) / (20,000-2,000) = 0.166, and the student's family eigenvector is [0, 1, 0, 0.166].

[0051] The number of consultations in the consultation record can be directly used as the consultation feature vector. The performance feature vector, assessment feature vector, family feature vector, and consultation feature vector can be concatenated to obtain a basic profile vector. For example, if the number of consultations is 3, then the basic profile vector for the student is [3.1, 8.16, 0.3, 0.05, 2.6, -0.6, 0, 1, 0, 0.166, 3].

[0052] Through these steps, a student's academic performance data, psychological assessment results, family background information, and counseling records can be converted into a standardized numerical vector, which is the basic profile vector. This basic profile vector can be widely used in subsequent tasks such as predicting student psychological status and providing personalized interventions.

[0053] S220 , extracting statistical features of classroom behaviors from the behavioral data, converting social data from the behavioral data into sentiment intensity and social frequency through sentiment analysis, and integrating the statistical features, sentiment intensity, and social frequency to obtain a behavioral feature vector. The statistical features include mean and variance.

[0054] For example, we can calculate the average of all classroom behaviors, such as the average number of times a student speaks in several classes. We can also calculate the volatility of classroom behavior, that is, the degree of difference in performance across classes. For example, we can calculate the variance of a student's class participation to reflect whether the student's performance in class is stable. Assume that the number of times a student speaks in class over 5 classes is [2, 4, 5, 3, 6]. Then the mean is (2+4+5+3+6) / 5=4, the variance is ((2-4)^2+(4-4)^2+(5-4)^2+(3-4)^2+(6-4)^2) / 5=2, and the statistical characteristics are [4, 2].

[0055] Sentiment analysis can be performed on students' social data (such as chat logs) to extract the emotional intensity of each social interaction. Higher emotional intensity indicates more positive student emotions; conversely, lower emotional intensity indicates more negative emotions. Existing sentiment analysis tools (such as VADER, TextBlob, and BERT) can be used to analyze social text. The sentiment scores returned by sentiment analysis can be used as emotional intensity features. The number of social activities a student participates in within a certain period (e.g., weekly or monthly) can be calculated, such as the number of interactions with classmates or teachers or the frequency of posts on social platforms. Suppose a student's social activity records for a week include social activity 1, with a chat content emotional intensity of 0.8; social activity 2, with a social media post emotional intensity of -0.3; and social activity 3, with a chat content emotional intensity of 0.5. Then, the emotional intensity for that week is (0.8 + (-0.3) + 0.5) / 3 = 0.33, and the social frequency is 3 (indicating that the student participated in three social activities that week). The student's behavioral feature vector is [4, 2, 0.33, 3].

[0056] S230 , calculating a heart rate variability value based on the heartbeat data of the physiological signal, and calculating a peak frequency of the skin conductance data in the physiological signal, and combining the heart rate variability value and the peak frequency to obtain a physiological feature vector.

[0057] It can be understood that Heart Rate Variability (HRV) refers to the fluctuation of the time interval between consecutive heartbeats. It is an indicator of the function of the autonomic nervous system. HRV reflects the heart's response to stress, emotions, etc. and can be calculated by analyzing heartbeat data.

[0058] For example, a Fourier transform can be performed on the heartbeat interval (RR interval, the time interval between two consecutive heartbeats) in the heartbeat data to calculate the power of different LF and HF frequency bands. The LF / HF ratio is then calculated as the heart rate variability value. HF (High Frequency) and LF (Low Frequency) reflect the frequency distribution of cardiac activity. The LF / HF ratio is an important indicator of the balance of the autonomic nervous system. A higher LF / HF ratio indicates a dominant sympathetic nervous system, while a lower LF / HF ratio indicates a dominant parasympathetic nervous system.

[0059] Skin conductance (SC) reflects changes in the skin's surface electrical resistance and is influenced by factors such as mood swings and stress. Fluctuations in SC often manifest as peaks at certain frequencies. The peak frequency can be determined by analyzing the frequency characteristics of the SC signal. The peak frequency of SC is the point in the data where the frequency component is strongest, reflecting the primary period of the signal's variation.

[0060] For example, a fast Fourier transform (FFT) can be used to perform frequency domain analysis on skin conductance data, obtaining frequency components and their corresponding amplitudes in the frequency domain. The frequency with the largest amplitude (i.e., the peak frequency) can then be found from the frequency components. Suppose the FFT analysis yields frequencies [0.5Hz, 1Hz, 1.5Hz, 2Hz] and amplitudes [0.2, 0.5, 0.4, 0.3]. In this case, the peak frequency is 1Hz because it corresponds to the largest amplitude.

[0061] The heart rate variability value (LF / HF) and the peak frequency of skin conductance data are combined into a physiological feature vector, which can be used to analyze mental health, mood fluctuations and stress levels.

[0062] S240 converts the input data into standard psychological data based on the educational metaphor dictionary. The standard psychological data is encoded using the BERT model to generate a high-dimensional semantic vector. This high-dimensional semantic vector is then reduced to generate a metaphor feature vector. The educational metaphor dictionary includes a mapping between students' metaphorical expressions and standard psychological descriptions.

[0063] It is understandable that students often use metaphorical language when expressing their mental states. One of the core tasks of the educational metaphor dictionary is to identify and interpret these metaphorical expressions. Metaphorical expressions are indirect descriptions of mental or emotional states through metaphors, analogies and other means.

[0064] BERT is a pre-trained language model that can understand the semantic information in text and capture the deep semantics of each word based on the context, thereby generating a high-dimensional semantic vector suitable for the situation.

[0065] For example, natural language processing (NLP) techniques (such as pre-trained models like BERT and GPT) or rule-based metaphor recognition methods can be used to process student input text (which can be converted to text if it's speech). Metaphors can be identified by searching for common metaphorical words, phrases, and structures. For example, "I felt like I was walking on the edge of a cliff before the exam" is a metaphor that expresses a psychological state of anxiety or stress. The meaning of a metaphor can be understood in context. By analyzing this context, NLP models can help accurately identify the psychological intent of the metaphor. For example, "I felt like I was walking on the edge of a cliff before the exam" may need to be interpreted in the context of the student's academic pressure and psychological state.

[0066] The educational metaphor dictionary stores mappings between metaphorical expressions and standard psychological states (or emotional states). Each metaphorical expression has one or more corresponding standard psychological states (such as anxiety and stress). A student's metaphorical expression can be converted into a standard psychological description by searching for the corresponding mapping in the educational metaphor dictionary. This can be done manually using a mapping table or automatically through machine learning model training.

[0067] Standard psychological data is input into the BERT model, and the standard psychological description is converted into a vector representation through the BERT encoder. Each psychological description corresponds to a high-dimensional vector (such as 384 dimensions), where each dimension represents a potential semantic feature.

[0068] Because the semantic vectors generated by the BERT model are high-dimensional, dimensionality reduction can be performed in practical applications to reduce computational complexity and extract the most useful features. Dimensionality reduction algorithms such as principal component analysis (PCA) can be used to reduce high-dimensional semantic vectors to lower dimensions. This dimensionality reduction preserves most information while reducing data redundancy. The reduced vectors have lower dimensions (e.g., 16 or 32 dimensions) but still capture the core semantics of standard psychological data, enabling more effective modeling and analysis of students' psychological states.

[0069] This step enables the quantification and analysis of students' psychological states. Based on the students' metaphorical characteristics at different time points, their emotional changes can be predicted, providing early warnings for teachers and psychological counselors, and providing a data basis for subsequent mental health intervention and support.

[0070] S250, performing multimodal data fusion on the basic portrait vector, the behavioral feature vector, the physiological feature vector, and the metaphor feature vector to obtain a psychological state feature vector.

[0071] For example, a separate sub-model can be designed for each modality (each vector), and the vectors in the intermediate layers of each sub-model can be concatenated or weighted averaged to form a mental state feature vector. The intermediate layer vectors are high-dimensional representations of each modality, incorporating deep abstraction and feature extraction of the modality. For example, a basic portrait vector is transformed through multiple layers of a fully connected network (FNN) to obtain a new, learned high-dimensional vector representation. This high-dimensional vector contains richer abstract information, not just the original features.

[0072] A fully connected neural network (FNN) or multi-layer perceptron (MLP) can be used to gradually extract high-dimensional representations of the basic image vector. A convolutional neural network (CNN) or RNN (such as LSTM) can be used to extract high-dimensional representations of the behavioral feature vector. A fully connected neural network (FNN) or a small convolutional neural network (CNN) can be used to extract high-dimensional representations of the physiological feature vector. A pre-trained language model such as BERT can be used for encoding to extract high-dimensional representations of the metaphorical feature vector. The high-dimensional representations of the basic image vector, the behavioral feature vector, the physiological feature vector, and the metaphorical feature vector can be fused through concatenation or weighted averaging. The fused vector is then passed to a fully connected neural network (FNN) for final processing to generate a unified psychological state feature vector.

[0073] This step makes full use of information from different data modalities (basic portraits, behavioral data, physiological signals, metaphorical expressions) to provide an accurate multi-dimensional representation for mental health assessment and personalized intervention.

[0074] In a possible implementation, the psychological counseling assistance method further includes: S201: Collect the expression texts of students on campus and annotate their psychological intentions through a semi-automatic annotation method to obtain a corpus, wherein the corpus includes a mapping between the expression texts and the psychological intentions.

[0075] It can be understood that psychological intent refers to the emotion or psychological state conveyed by a student's expression (such as anxiety, confusion, stress, joy, etc.). Semi-automatic annotation methods combine the advantages of manual annotation and machine learning, using a pre-trained natural language processing (NLP) model to initially classify and annotate student texts, and then manually correct or supplement the annotations.

[0076] For example, we can collect students' feedback in class or after the course, students' participation in mental health questionnaires, interviews or online discussions, text content posted by students on school social platforms, forums, etc., and collect students' speeches through speech-to-text or daily conversation records. These text contents can be used as basic corpus for further annotation and analysis.

[0077] Sentiment classification models (such as BERT and LSTM) can be used to automatically annotate student expressions. This automated annotation can then be manually reviewed and revised to ensure accuracy. Semi-automatic annotation methods can generate an annotated corpus containing each student expression and its corresponding psychological intent. This corpus will serve as the foundation for subsequent analysis and metaphor dictionary construction.

[0078] S202: Extract high-frequency expression texts from the corpus and process the expression texts using natural language processing methods to obtain metaphor texts; obtain the corrected metaphor texts to obtain an educational metaphor dictionary. The metaphor texts include mappings between students' metaphorical expressions and standard psychological descriptions.

[0079] For example, word frequency statistics can be performed on the expressions in the corpus to identify the most common expressions. Methods such as the bag-of-words model and TF-IDF can be used to extract high-frequency words from the expressions in the corpus. Metaphor recognition algorithms within natural language processing (NLP) can be used to automatically identify metaphorical expressions in text. Metaphorical expressions can be compared with standard psychological descriptions to clarify the psychological intent behind the metaphors. If any metaphors are unclear or non-standard, they can be manually corrected or supplemented to ensure that each metaphor corresponds to an accurate psychological state. This creates a dictionary of educational metaphors, which can then be applied to student input data recognition.

[0080] The Educational Metaphor Dictionary not only maps metaphors to psychological descriptions but also includes common contexts and application areas of metaphorical expression, helping to understand the specific context of students' emotions. The Educational Metaphor Dictionary effectively maps students' metaphorical expressions to standard psychological descriptions, providing data support for subsequent psychological analysis, emotional assessment, and personalized interventions.

[0081] S300: Determine whether to trigger a psychological warning based on basic data, behavioral data, physiological signals, and input data.

[0082] For example, a suitable model can be used to assess the student's mental state using the mental state feature vector obtained in step S200 to determine whether the student is in a potential mental risk state. Models can include traditional machine learning models, such as support vector machines (SVMs), random forests (RFs), and gradient boosted decision trees (GBDTs), which can be used for classification problems (e.g., determining whether to trigger an alert); deep learning models, such as convolutional neural networks (CNNs), long short-term memory networks (LSTMs), or transformers, which can automatically learn complex relationships between features; and ensemble learning methods.

[0083] Models can be trained using labeled training datasets (e.g., to determine whether a student's mental state has been assessed as requiring a psychological warning). During training, hyperparameters can be tuned through methods such as cross-validation and grid search to select the optimal model and parameters. The model output can consider multiple dimensions of mental state (e.g., anxiety, depression, stress, etc.), outputting probability values ​​for each mental state, or combining the results into a comprehensive warning value, and determining whether a warning should be triggered based on a preset threshold.

[0084] Thresholds for psychological state assessment can be set based on historical data and business needs. If the psychological state score or probability of each psychological state output by the model exceeds the preset threshold, a psychological warning is triggered. Different warning strategies can also be defined for each situation based on different psychological state characteristics (such as anxiety, depression, loneliness, etc.). If multiple psychological indicators (such as anxiety, depression, stress, etc.) are abnormal at the same time, a more serious psychological intervention warning can be triggered.

[0085] When a psychological warning is triggered, timely feedback can be provided to educators, parents, or psychological counselors, who can then provide relevant psychological counseling solutions. In practical applications, by collecting student feedback data after each warning (such as counseling effectiveness, academic performance, and changes in psychological state), the model can be retrained and optimized, gradually improving the accuracy and effectiveness of the warnings.

[0086] This step can comprehensively assess students' mental states through data from multiple modalities and make timely early warning responses.

[0087] In one possible implementation, S300, based on basic data, behavioral data, physiological signals, and input data, determining whether to trigger a psychological warning includes: S310, determining basic risk based on the basic portrait vector.

[0088] For example, weights can be assigned to the performance eigenvector, assessment eigenvector, family eigenvector, and consultation eigenvector, and the basic risk value can be directly calculated by weighted summation. For example, each eigenvector can be set to be equally important, with a weight of 0.25. The basic risk value is 0.25 × (3.1 + 8.16 + 0.3 + 0.05) + 0.25 × (2.6 + (-0.6)) + 0.25 × (0 + 1 + 0 + 0.166) + 0.25 × 3 = 4.444.

[0089] S320, using the LSTM network to calculate the concentration corresponding to the classroom behavior in the behavioral data, and based on the attention mechanism, calculate the correlation between the social data and physiological signals in the behavioral data to obtain the correlation, and determine the behavioral risk based on the concentration and correlation.

[0090] It can be understood that LSTM (Long Short-Term Memory Network) is a recurrent neural network (RNN) suitable for processing time series data. Because it can capture long-term and short-term dependencies, it is very suitable for analyzing time series data such as classroom behavior.

[0091] The attention mechanism is a technology that enables the model to focus on the most relevant parts of the input data when learning. The attention mechanism can help the model learn important features in social data and physiological signals, thereby evaluating the correlation between them.

[0092] For example, classroom behavior data is fed into an LSTM network, which then outputs a concentration vector (e.g., a value between 0 and 1) representing the student's concentration level at each moment in class. The LSTM layer processes the time-series classroom behavior data, followed by a fully connected layer that outputs predicted concentration values. When training the LSTM network, the mean squared error (MSE) can be used as a loss function to evaluate the difference between the model's predicted concentration and the actual data.

[0093] The additive attention mechanism or the dot-product attention mechanism can be used to calculate the temporal importance of social data and physiological signals through the attention layer to obtain a weight value that reflects the correlation between social data and physiological signals at a certain moment.

[0094] Focus and relevance can be combined through a weighted summation to produce a behavioral risk value. The weights for focus and relevance can be set manually based on experience, or the model can be trained using real-world data to learn how to adjust the weights based on a student's focus and relevance, ensuring that the final behavioral risk value accurately reflects the student's overall risk.

[0095] This step effectively evaluates students' behavioral performance by combining classroom behavior time-series data and multimodal information (social data and physiological signals), and provides a data basis for subsequent personalized intervention and support.

[0096] S330: Compare the heart rate variability value with the peak frequency. If the heart rate variability value is greater than the peak frequency, determine the heart rate variability value as a psychological risk. If the peak frequency is greater than the heart rate variability value, determine the peak frequency as a psychological risk.

[0097] For example, the heart rate variability value calculated in step S230 may be compared with the peak frequency to determine which one has a larger value, and the larger value may be used as the psychological risk value.

[0098] S340, calculating the initial risk based on the metaphor feature vector, identifying whether there is a contradiction between the standard psychological data and the behavioral data based on the standard psychological data and the behavioral data, and if there is a contradiction, obtaining a penalty coefficient, and calculating the metaphor risk based on the initial risk and the penalty coefficient.

[0099] For example, the metaphor feature vector can be encoded using the BERT model to calculate a basic risk value (initial risk) between 0 and 1.

[0100] Before identifying inconsistencies, normalize the standard psychological and behavioral data to align them temporally and facilitate comparison on the same scale. Simple rules can be developed to initially identify inconsistencies. For example, if anxiety scores are high (e.g., exceeding a critical threshold) and social interactions are high, this can be considered a contradiction—in other words, a student's psychological health data is inconsistent with their social behavior. Low anxiety scores and low social interactions suggest that the student may be experiencing social isolation, another manifestation of inconsistency.

[0101] Machine learning models can be used for contradiction detection. A training dataset is constructed containing standardized student psychological and behavioral data. Each record includes both data and a label (indicating whether a contradiction exists). Features are extracted from these data and fed into the model for training. Contradiction detection can be treated as a binary classification task. A classification model (such as a random forest, support vector machine (SVM), or neural network) is used to determine whether a contradiction exists. The model outputs a contradiction (1 for the presence of a contradiction, 0 for the absence of a contradiction) based on the input data (standard psychological and behavioral data).

[0102] When a contradiction is detected between the standard psychological data and the behavioral data, a penalty coefficient (γ) can be obtained. The role of the penalty coefficient is to adjust the initial risk value according to the intensity of the contradiction. For example, the penalty coefficient for a strong contradiction (such as high anxiety but high social activity) can be set to 0.5; the penalty coefficient for a weak contradiction (such as low anxiety and low social interaction frequency) can be set to 0.3; and the penalty coefficient for no contradiction can be set to 0. The metaphor risk value can be calculated using the initial risk and the penalty coefficient according to the metaphor risk calculation formula. The metaphor risk calculation formula is: ,in, represents the metaphorical risk value, represents the initial risk value, represents the penalty coefficient, It is a constant that indicates the intensity of the contradiction (for example, if it is set to 0.3, it indicates the degree of punishment). Suppose a student uses the metaphor of the abyss, and the BERT model calculates the metaphor risk as 0.9. However, the student is very socially active, so the contradiction is identified and the penalty coefficient is 0.5. The metaphor risk value is .

[0103] This step can identify the contradictions between standard psychological data and behavioral data, and adjust the risk according to the penalty coefficient to obtain metaphorical risk. It combines metaphorical characteristics, standard psychological data and behavioral data to dynamically adjust students' metaphorical risk, providing a more accurate and detailed tool for students' psychological state assessment and personalized intervention plan formulation.

[0104] S350 calculates the overall risk based on the basic risk, behavioral risk, psychological risk, metaphorical risk, synergy gain coefficient, and exponential function, and determines whether to trigger a psychological warning based on the overall risk. The synergy gain coefficient and exponential function are used to increase additional risk when multiple risks exceed corresponding thresholds.

[0105] As you can understand, the synergy gain coefficient is used to adjust the additional risk in the risk synthesis process. It can increase the risk value when multiple risk values ​​exceed the threshold simultaneously. The synergy gain coefficient can be set to a small value (such as 0.15). The exponential function determines whether multiple risk values ​​exceed the threshold simultaneously. When multiple risk values ​​exceed their respective thresholds simultaneously, the exponential function returns 1, otherwise it returns 0. The combined effect of the exponential function and the synergy gain coefficient is that when all risk values ​​exceed their respective thresholds simultaneously, the indicator function value is 1, and the synergy gain coefficient acts on the calculation of the combined risk, adding additional risk.

[0106] For example, assuming that the basic risk value is 1.8 (threshold 1.5, weight 0.3), the behavioral risk value is 1.3 (threshold 1.2, weight 0.25), the psychological risk value is 2.1 (threshold 2.0, weight 0.25), and the metaphorical risk value is 0.75 (threshold 0.8, weight 0.2), and the synergy gain coefficient is 0.2, the basic risk value, behavioral risk value, and psychological risk value exceed the corresponding thresholds, while the metaphorical risk value does not exceed them. Therefore, the comprehensive risk is =0.3×1.8+0.25×1.3+0.25×2.1+0.2×0.75+0.15×(1×1×1×0)=1.54, where 、 、 、 Basic risk value , behavioral risk value , psychological risk value Metaphorical Risk Value The corresponding weight.

[0107] You can set a warning threshold. When the calculated comprehensive risk value exceeds the warning threshold, a psychological warning is triggered. For example, if the warning threshold is 1.5, then 1.54 is greater than 1.5, triggering a psychological warning.

[0108] Triggering psychological warnings through multi-dimensional data can eliminate the noise interference or randomness of a single data source (for example, short-term HRV abnormalities may be caused by exercise), capture hidden crises under the synergistic effect of multiple dimensions, and eliminate false positives by detecting contradictions between data (for example, metaphors show anxiety but physiological data are stable).

[0109] S400: If it is determined that a psychological warning is triggered, a target guidance plan is generated based on the psychological state feature vector.

[0110] For example, when a student's psychological state score or the probability value of each psychological state exceeds a preset threshold, the system triggers a psychological warning. The student's psychological state feature vector can be input into the counseling solution generation model, and the model can output a psychological counseling solution for the student.

[0111] The training process of the counseling plan generation model: collect the basic data, behavioral data, physiological signals and input data of each student, use the method of step S200 to fuse these data into the psychological state feature vector corresponding to each student (the input of the model), and mark the specific psychological counseling plan (the output of the model) to form a data set. Before the data set is multimodal data fused, the data can be preprocessed. For missing values, you can choose to fill (such as filling with the mean or mode) or discard relevant data; standardize or normalize data of different ranges (such as heart rate, grade data, etc.) to facilitate comparison on the same scale; categorical variables such as students' gender and grade can be converted into numerical values ​​through one-hot encoding (One-Hot Encoding) or label encoding (Label Encoding); you can select key features that affect the counseling effect and remove redundant or irrelevant features.

[0112] You can choose an appropriate machine learning model. If the target counseling solution falls into predefined categories (such as emotional counseling, social counseling, academic counseling, etc.), it can be considered a multi-classification problem, and a classification algorithm (such as decision tree, random forest, support vector machine, etc.) can be selected. If the counseling solution is based on a specific score or continuous variable (such as counseling effectiveness rated from 1 to 5), a regression model (such as linear regression, Lasso regression, etc.) can be used for prediction. For large data volumes and complex features, deep learning models (such as multi-layer neural networks, LSTM, etc.) can be used to handle complex pattern recognition tasks. They are particularly suitable for time-series input data (such as physiological signals and speech data).

[0113] The dataset can be divided into a training set and a test set, with 80% of the data used for training and 20% for testing. The model is trained using the training set, and model parameters are adjusted through backpropagation (for neural networks) or other optimization methods (for regression, decision trees, etc.). The model's performance is evaluated on the test set using metrics such as mean squared error (MSE), accuracy, F1 score, and AUC. Methods such as cross-validation and grid search are used to adjust the model's hyperparameters and find the optimal model configuration.

[0114] After the model is trained, it can be accessed through an API or embedded into a school's mental health management system. The system automatically generates a targeted counseling plan based on the student's mental state eigenvector. Once the targeted counseling plan is generated, it can be provided to teachers, psychological counselors, and other relevant personnel, enabling them to quickly understand the student's mental state and the risks they face. Psychological counselors can use the targeted counseling plan as a reference for further personalized counseling for that student, reducing the time psychological counselors spend assessing students' mental states and developing intervention measures, thereby improving their work efficiency.

[0115] This step can not only effectively identify students' mental health problems, but also take intervention measures before problems occur, improve students' mental health levels, and help teachers provide more efficient and personalized support.

[0116] In one possible implementation, see Figure 2 In step S400, a target guidance plan is generated based on the psychological state feature vector, including: S410: Searching for a set of candidate counseling strategies that match the psychological state feature vector from the psychological knowledge graph, wherein the psychological knowledge graph is constructed based on the mapping between psychological problems and psychological counseling strategies.

[0117] As you can understand, a psychological knowledge graph is a graphical tool for organizing and representing knowledge in the field of psychology. It aims to connect and structure various psychological concepts, psychological problems, counseling strategies, emotional states, and other information, and to represent the relationships between these information through the nodes and edges of the graph. The psychological knowledge graph can be used to help understand psychological problems, recommend personalized counseling strategies, and provide structured queries on psychological knowledge.

[0118] Exemplarily, psychological knowledge graph construction: knowledge in the field of psychology can be collected by reading psychology literature and journal articles to collect relevant knowledge such as psychological problems, counseling strategies and emotional states; by interviewing psychology experts (such as psychologists, psychiatrists, etc.), practical experience and theoretical knowledge can be collected; by using psychology textbooks and manuals, standard psychological concepts and theories can be obtained; and clinical mental health data (such as treatment records, medical records, etc.) can be used to supplement and verify the information in the knowledge graph.

[0119] NLP methods can be used to extract psychological entities (such as psychological problems and counseling strategies) from collected text. Pattern recognition or deep learning methods can then be used to automatically identify relationships between entities (e.g., anxiety → meditation). Based on these relationships, rules can be generated to describe the relationships between different entities. For example, if a person has anxiety, meditation can be recommended. As the field of psychology continues to develop, the discovery of new theories, counseling strategies, and issues can continuously update the knowledge graph. This can be achieved through regular literature review, clinical data collection, and expert feedback.

[0120] Machine learning models (such as neural networks, support vector machines, etc.) or similarity metrics (such as cosine similarity) can be used to calculate the similarity between the psychological state feature vector and the nodes in the graph, and select the nodes most similar to the psychological state feature vector. These nodes represent potential psychological problems and corresponding counseling strategies.

[0121] By constructing a psychological knowledge graph, we can efficiently structure professional knowledge in the field of psychology and provide solid knowledge support for practical applications (such as personalized psychological intervention, counseling strategy recommendations, etc.).

[0122] S420: Perform multi-objective optimization on the candidate diversion strategy set to select the optimal diversion strategy combination.

[0123] As you can understand, NSGA-II is a non-dominated sorting genetic algorithm suitable for multi-objective optimization problems. The NSGA-II algorithm transforms multiple objectives into a holistic evaluation metric and selects the optimal solution using non-dominated sorting and crowding comparison. Optimization objectives can include minimizing time cost (minimizing the sum of the time required to implement all candidate strategies), minimizing privacy exposure risk (minimizing the sum of the privacy exposure risks of all candidate strategies), and maximizing the expected effect coefficient (maximizing the weighted sum of the expected effect coefficients of all candidate strategies).

[0124] For example, multiple grooming strategy combinations are randomly generated based on each optimization objective. Each combination contains multiple strategies, each with a corresponding time cost, privacy exposure risk, and expected effectiveness coefficient. Each individual in the initial population is subjected to a non-dominated sorting algorithm. This non-dominated sorting algorithm divides the population into multiple levels. Individuals in the first level are optimal (they are not dominated by other individuals on all objectives), individuals in the second level are sub-optimal, and so on. Crowding comparison is used to select solutions that are sparse in the target space, thereby maintaining population diversity. Individuals with higher crowding levels occupy a larger area in the target space. A tournament selection method (based on non-dominated sorting and crowding) can be used to select individuals with good fitness. Crossover is performed on these selected individuals to generate new strategy combinations. Through small-probability mutation operations, some of the individual characteristics are changed to explore more of the solution space. An elite selection strategy is used to retain the current optimal individual and update the population. When the population reaches a certain number of generations or the optimization converges, the algorithm terminates and outputs the optimal solution. Ultimately, the NSGA-II algorithm outputs a Pareto frontier solution set, consisting of multiple alternative diversion strategy combinations that are independent of each other in terms of time cost, privacy exposure risk, and expected effectiveness. The three output alternative strategy combinations can be ranked based on their acceptability and their combined effectiveness to select the optimal strategy combination.

[0125] S430: Based on the optimal guidance strategy combination, basic data and behavioral data, the optimal guidance strategy combination is adjusted to generate a target guidance plan.

[0126] For example, the frequency and duration of each strategy can be adjusted by calculating the student's time constraints and the strategy's time requirements. The frequency of a strategy indicates the number of times a counseling strategy is implemented within a specific time period. Adjusting the frequency of a strategy can dynamically increase or decrease the frequency of a strategy based on the student's psychological assessment results and behavioral performance. For example, if a student's anxiety score is high, the frequency of progressive muscle relaxation training and meditation exercises can be automatically increased (such as 3 or 4 times a week); if the student's social interaction frequency is low, the frequency of social skills training can be increased; if the student's classroom participation is low, the frequency of counseling strategies for improving classroom behavior (such as classroom discussions, participatory exercises, etc.) can be increased.

[0127] Each counseling strategy has a duration, which can be adjusted to extend or shorten the strategy's implementation time based on the student's needs and responses. For example, for students with severe anxiety, cognitive behavioral therapy sessions might need to be adjusted from the standard 60 minutes to 90 minutes. For students with low motivation, social skills training sessions could be extended to strengthen their social confidence and communication skills. For students with limited time, the duration of the strategy can be reduced to ensure effective implementation.

[0128] The intensity of strategies can be dynamically adjusted based on student psychological assessment results. Strategy intensity is another key factor in adjustment. The intensity of each strategy can be automatically adjusted based on student responses (e.g., mood changes, class participation, social improvement, etc.). For example, for students whose anxiety has eased, the frequency or intensity of progressive muscle relaxation training can be reduced. If a student's social interactions increase, the intensity or frequency of social skills training can be reduced in favor of more relaxing exercises or activities.

[0129] Through the above adjustment process, a personalized target guidance plan is finally generated. The plan can include each guidance strategy and the implementation method of each guidance strategy (implementation frequency, implementation duration, and strategy intensity).

[0130] These steps go beyond simply identifying counseling strategies; they also involve comprehensive adjustments based on students' actual needs, psychological states, and external factors, maximizing the effectiveness of counseling programs. This provides a systematic tool for personalized psychological intervention and education, precisely tailoring the most appropriate counseling solutions for students.

[0131] S500: Obtain the student's short-term data and long-term data, and optimize the process of generating a target counseling plan based on the student's short-term data and long-term data. The short-term data includes the student's heartbeat data after counseling, and the long-term data includes the student's academic performance data and social data after counseling.

[0132] Understandably, short-term data refers to data collected within a short period of time after a student receives counseling from a counselor. It reflects the impact of counseling on the student's immediate physiological reactions and emotional state. The optimization goal of short-term data is to adjust intervention measures in a timely manner to effectively alleviate the student's physiological stress. For example, changes in a student's heart rate (such as heart rate and heart rate variability) after counseling can reflect the student's stress level and emotional fluctuations.

[0133] Long-term data refers to data collected over a long period of time after a student receives counseling from a counselor. This data reflects the ongoing impact of counseling on the student's overall behavior and academic performance, and can provide feedback on the effectiveness of counseling. For example, data on students' academic performance after counseling, including test scores and homework grades, and changes in their social interactions, such as frequency of participation in school activities, interactions with classmates, and social network activity, can be included.

[0134] For example, heart rate data after each counseling session can be tagged with the counseling timestamp to ensure data timeliness. After counseling sessions, students' heart rates can be continuously monitored using wearable devices (such as smartwatches and health monitoring devices). Student performance data after counseling sessions can be collected through platforms such as learning management systems (LMS). After counseling sessions, students' social behavior can be tracked through social platforms or school activity records, or social network analysis tools can be used to analyze student interactions on social media (e.g., frequency of participation in group activities, number of speeches, etc.).

[0135] Each student's psychological state feature vector can be combined with the counseling plan and feedback data (both short-term and long-term) to form labeled training data. The target variable is the outcome the model needs to predict: whether the student benefited from the counseling plan. This can be a binary classification problem (e.g., whether counseling was effective) or a regression problem (e.g., a rating of counseling effectiveness). For classification problems, the target variable can be a label indicating whether counseling was effective or ineffective; for regression problems, the target variable can be a rating of counseling effectiveness (e.g., whether the student's anxiety level decreased, or the percentage of grade improvement).

[0136] You can choose a suitable machine learning algorithm based on the task requirements: if the goal is to determine whether the guidance plan is effective, you can use a classification algorithm, such as decision tree, random forest, and support vector machine (SVM); if the goal is to predict the specific numerical value of the guidance effect, you can use a regression algorithm, such as linear regression, support vector regression (SVR), and neural network regression.

[0137] The model can be trained using labeled training data, and the model's effectiveness can be evaluated using cross-validation methods to avoid overfitting. The model can be optimized by adjusting model hyperparameters (such as tree depth, regularization terms, etc.). The accuracy rate can be used to evaluate whether the grooming effect predicted by the model in classification problems is correct. The mean square error can be used to evaluate the gap between the grooming effect predicted by the model in regression problems and the actual effect. The final evaluation results can be used to further adjust the grooming solution generation model and optimize the solution generation capability.

[0138] The newly collected feedback data can be added to the training set and used to retrain the counseling plan generation model. This allows the model to adjust its ability to predict counseling plan effectiveness. Incremental learning methods (such as online learning and reinforcement learning) enable the model to continuously absorb new data and adjust model parameters in real time. As student data accumulates, the counseling plan generation model can be personalized and optimized, generating more precise counseling plans for each student based on their unique psychological characteristics.

[0139] By collecting short-term and long-term data and optimizing supervised learning methods, we can continuously improve the accuracy of the model and the effectiveness of the counseling plan. Short-term feedback primarily reflects the immediate effects of the counseling plan, while long-term feedback helps the model assess its lasting impact. By continuously training and adjusting the model, we can provide each student with a targeted counseling plan, which is conducive to the continued effectiveness of psychological counseling.

[0140] In a possible implementation, the psychological counseling assistance method further includes: S501, calculating the heart rate variability value after the grooming based on the short-term data, and calculating the physiological change degree based on the heart rate variability value after the grooming and the heart rate variability value.

[0141] For example, the heart rate variability value (LF / HF) after grooming may be calculated based on the heartbeat interval of each heartbeat in the heartbeat data of the student after grooming using the same method as step S230 .

[0142] Physiological change reflects the degree of change in the student's physiological state after counseling. The physiological change can be calculated by comparing the HRV values ​​before and after counseling. The calculation formula is: ,in, Indicates physiological changes. Indicates the heart rate variability value after guidance, Indicates the heart rate variability value before counseling.

[0143] S502: Calculate the behavior improvement degree based on the long-term data, the performance data in the behavior data, and the social data in the behavior data.

[0144] For example, the behavior improvement degree can be calculated by comparing the performance before and after counseling and the number of social interactions before and after counseling. The behavior improvement degree calculation formula is: ,in, Indicates the degree of improvement in behavior. Indicates the results before counseling. Indicates the results after guidance. Indicates the number of social interactions before counseling. Indicates the number of social interactions after counseling.

[0145] S503, obtaining the student's satisfaction with the target counseling program, and determining a first effect score based on the degree of physiological change, behavioral improvement, and satisfaction.

[0146] It can be understood that student satisfaction is an important measure of the effectiveness of the target counseling program. Students can evaluate their feelings, emotional changes after counseling based on the target counseling program, and the effectiveness of the counseling program based on the teacher's score.

[0147] For example, the first effect score can be calculated by combining the physiological change degree, behavioral improvement degree and satisfaction through weighted sum. Assuming that the physiological change degree is 0.2, the behavioral improvement degree is 0.15, and the satisfaction degree is 0.9, the corresponding weights are 0.4, 0.3, and 0.3 (which can be adjusted according to actual needs, and the weight sum is 1), the first effect score is =0.4×0.2+0.3×0.15+0.3×0.9=0.395. Among them, represents the first effect score, Express satisfaction, 、 、 They represent the weights corresponding to the degree of physiological change, the degree of behavioral improvement, and the degree of satisfaction respectively.

[0148] S504: updating the student's psychological resilience index based on the first effect score and the historical effect score, wherein the historical effect score is used to represent the student's historical first effect score after each psychological counseling session.

[0149] It can be understood that the psychological resilience index indicates the students' adaptability after undergoing several psychological counseling sessions.

[0150] For example, the effect of each psychological counseling session will gradually decay over time, and a time decay factor can be used to simulate this phenomenon. The time decay factor is a decay factor less than 1, indicating that the effect of each intervention gradually weakens over time. Assuming that the time interval between each counseling session is fixed, the decay factor can be a fixed value, such as 0.9.

[0151] The product of the first effect score and the time decay factor and the product of the historical effect score and the time decay factor can be accumulated to obtain the student's psychological resilience index. ,in, represents the psychological resilience index, Indicates the The first effect score after psychological counseling, represents the time decay factor, The time decay factor Assuming that the first effect score after the first counseling session was 0.45, the first effect score after the second counseling session was 0.32, and the first effect score after the current counseling session (the third session) was 0.395, then the student's psychological resilience index is .

[0152] By taking these steps into account, time factors and changes in intervention effects can be taken into account, which can help educators and mental health experts provide long-term psychological support to students and effectively track changes in students' psychological adaptation.

[0153] In one possible implementation, S500, optimizing the process of generating a target guidance plan based on the student's short-term data and long-term data, includes: S510: Calculate the recovery rate based on the short-term data and the performance improvement slope based on the performance data in the long-term data. The recovery rate represents the speed at which the student's heart rate variability value returns to the baseline value, and the performance improvement slope represents the linear regression slope of the student's test score after the counseling.

[0154] For example, a HRV baseline value can be set, and the time for the student's HRV to return to the HRV baseline value after counseling is calculated. Based on this time, the HRV after counseling and the HRV baseline value, the recovery speed is calculated, that is, ,in, Indicates the recovery speed, Indicates the HRV baseline value, Indicates the time it takes for the student's HRV to return to the HRV baseline value after counseling.

[0155] It can be understood that the performance improvement slope is used to measure the linear regression slope of students' test scores within a certain period of time, indicating the speed of improvement of students' performance after counseling.

[0156] For example, the performance data in the long-term data include a series of test scores or homework scores, and linear regression can be used to fit the performance data to obtain the performance improvement slope.

[0157] S520: Obtain the program adoption rate and determine a second effectiveness score based on the recovery speed, grade improvement slope, program adoption rate, and psychological resilience index. The program adoption rate represents the proportion of students who have adopted the psychological counseling program.

[0158] For example, the second effect score can also be calculated by weighted summation. The weights corresponding to the recovery speed, performance improvement slope, program adoption rate and psychological resilience index can be determined according to actual conditions, and the sum of their weights is 1.

[0159] S530: Construct a reward function based on the satisfaction score and the second effect score, wherein the reward function is used to represent the feasibility of the target guidance solution.

[0160] For example, a Markov decision process (MDP) is a decision-making framework in which, at each time step, an agent maximizes cumulative rewards by selecting actions. The MDP consists of a state space, an action space, a reward function, and a transition probability. The state space can be represented by a psychological state eigenvector; the action space can be represented by an optimal combination of counseling strategies, each with different implementation frequencies, durations, and intensities; the reward function can be calculated based on the student's satisfaction and secondary effect scores. The purpose of the reward function is to evaluate the feasibility and effectiveness of each strategy combination; the transition probability describes how to transition from the current state to the next state after taking an action in a specific state. For the generation process of psychological counseling plans, the transition probability can be simulated based on the student's recovery speed and the slope of their performance improvement.

[0161] A reward function can be constructed using the contribution of the secondary effect score and satisfaction. If students provide positive feedback on the intervention (high satisfaction) and the intervention is effective (high secondary effect score), the reward value is large, indicating that the intervention is a relatively successful one. If the intervention is ineffective or the student is dissatisfied, the reward value is small, indicating that the intervention needs to be adjusted or replaced.

[0162] S540 , based on the reward function and the psychological state feature vector, a proximal strategy optimization algorithm is used to optimize the multi-objective optimization process in the process of generating the target grooming solution.

[0163] As you can understand, Proximal Policy Optimization (PPO) is a reinforcement learning algorithm suitable for optimizing policies in Markov decision processes, especially when there are multiple objectives to balance. The core idea of ​​PPO is to avoid excessive policy updates by making moderate policy updates, thus maintaining policy stability.

[0164] For example, PPO will guide the optimization of NSGA-II based on the psychological state eigenvector and the reward function in the MDP. Specifically, the goal of the PPO algorithm optimization is to adjust the objective function, selection, crossover, mutation, and other operations of NSGA-II so that the generated grooming solution performs better on multiple objectives.

[0165] In step S420, NSGA-II generates an initial solution by optimizing the objective, selection, crossover, and mutation operations. PPO improves NSGA-II's performance by optimizing the current policy. PPO evaluates each generated solution using a reward function and updates NSGA-II's policy based on the evaluation results. PPO updates the policy by calculating an advantage function, which assesses which actions have the best impact on the current policy. PPO updates the policy based on the results of the reward and advantage functions. This update affects parameters such as NSGA-II's selection mechanism, crossover, and mutation. The updated policy improves the grooming solutions generated by NSGA-II across various objectives. For example, PPO may adjust NSGA-II's selection preferences to select more effective grooming solutions, or optimize crossover and mutation operations to generate more effective offspring solutions. This process repeats over multiple iterations. After each iteration, PPO updates the policy based on feedback data, ultimately enabling NSGA-II to generate grooming solutions that increasingly meet the requirements.

[0166] The above steps use the PPO algorithm to optimize NSGA-II, continuously improving the effectiveness and adaptability of the generated guidance plans, making the generated guidance plans more in line with students' needs and improving the overall effect.

[0167] In a possible implementation, the psychological counseling assistance method further includes: S10, visualizing the generation process of the target counseling program, obtaining a program generation logic chain, and sending the program generation logic chain to the teacher side so that the psychology teacher can view the generation process of the student's target counseling program.

[0168] For example, the generation process of the target guidance plan can be visualized by constructing a decision tree, wherein the process of converting the student's basic data, behavioral data, physiological signals, and input data into a psychological state feature vector through a multimodal data fusion method is used as the root node.

[0169] The process of searching for a set of candidate counseling strategies that matches the psychological state feature vector through the psychological knowledge graph is used as the first-level node (psychological state matching). The branches corresponding to the first-level nodes can be psychological problems and corresponding counseling strategies. For example, if the psychological state feature vector matches an anxiety-related strategy, the process proceeds to the anxiety counseling solution generation node; if it matches a social problem, the process proceeds to the social skills training node.

[0170] The multi-objective optimization process for the candidate grooming strategies is performed as the second-layer nodes (multi-objective optimization). The corresponding branches of the second-layer nodes can be the grooming strategies corresponding to each objective. For example, minimizing time cost -> selecting the strategy with lower time cost, minimizing privacy exposure risk -> selecting the strategy with lower privacy exposure risk, and maximizing expected effect -> selecting the strategy with higher expected effect based on historical data or the success rate of similar cases.

[0171] The process of further adjusting the guidance strategy based on the optimal guidance strategy combination, basic data, and behavioral data is considered the third-level node (optimal strategy adjustment). Ultimately, the leaf nodes represent the generated target guidance plan. Based on these steps, a decision tree (the solution generation logic chain) is constructed. This solution generation logic chain illustrates the decision path from various student data to the final selected plan.

[0172] The student's solution generation logic chain can be saved as a PDF file or text report, which can be sent to the teacher in the form of an email; the solution generation logic chain can be embedded in the teacher's teaching platform, and the teacher can view the guidance solution generation process of each student through the platform; notifications can be pushed to the teacher through the instant messaging system or the school's internal communication platform to remind the teacher to view the student's generation logic chain.

[0173] After the teacher receives the logical chain of solution generation, he or she can clearly understand how the guidance plan for each student is generated through images or text chains, including the student's psychological state, behavioral data, and the basis for selecting the guidance plan.

[0174] Through this step, psychological teachers can timely understand the target guidance plan generation process of each student and make necessary adjustments based on the plan generation logic, thereby providing students with more personalized psychological intervention.

[0175] S20, when the student rejects the target counseling plan generated three times, the manual counseling signal and the logic chain of each plan generation are sent to the teacher side, so that the psychological teacher can provide psychological counseling to the student.

[0176] For example, after each counseling plan is generated, students can be presented with an interface to choose whether to adopt it. If a student refuses to accept the plan, this behavior can be recorded, and each student's acceptance of each counseling plan can be tracked, with the number of student rejections accumulated. If a student refuses a generated target counseling plan continuously or three times cumulatively, a manual counseling signal will be triggered, which may indicate that the student is dissatisfied with the current plan or has some psychological resistance.

[0177] If a student rejects a proposal three times, an automated process can trigger a manual intervention signal, which can be sent to the teacher via SMS, email, or system notifications, prompting timely intervention. Simultaneously, the logical chain of each proposal generation is sent to the teacher. The student's proposal generation process and reasons for rejection can be displayed to the teacher via email, a teaching platform, or a web application. Based on this logical chain, the teacher can understand the student's psychological state and behavior and adjust the content or form of the counseling plan. The teacher can also conduct one-on-one communication with the student to understand the real reason for the rejection and adjust the counseling approach.

[0178] This step helps students get timely manual intervention when they encounter difficulties, and enables teachers to clearly understand the process of generating each student's target guidance plan, thereby helping them develop more personalized and effective guidance measures.

[0179] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0180] Corresponding to the psychological counseling assistance method described in the above embodiment, the embodiment of the present application also provides a psychological counseling assistance device, and each unit of the device can implement each step of the psychological counseling assistance method.

[0181] The device includes: The acquisition unit is used to obtain students' basic data, behavioral data, physiological signals, and input data. Basic data includes academic performance, psychological assessment results, family background information, and consultation records; behavioral data includes classroom behavior and social interaction data; physiological signals include heart rate data and skin conductance data; and input data includes voice data or text data.

[0182] The multimodal data fusion unit is used to perform multimodal data fusion on basic data, behavioral data, physiological signals and input data to obtain a psychological state feature vector.

[0183] The psychological warning triggering unit is used to determine whether to trigger a psychological warning based on basic data, behavioral data, physiological signals and input data.

[0184] The target guidance scheme generating unit is used to generate a target guidance scheme based on the psychological state feature vector if it is determined that the psychological warning is triggered.

[0185] The optimization unit is used to obtain the student's short-term and long-term data and optimize the generation process of the target counseling plan based on the student's short-term and long-term data. The short-term data includes the student's heart rate data after counseling, and the long-term data includes the student's academic performance data and social data after counseling.

[0186] It should be noted that the information interaction, execution process, etc. between the above-mentioned units are based on the same concept as the method embodiment of this application. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0187] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned device can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0188] The present application also provides a psychological counseling auxiliary device. Figure 3 This is a structural diagram of a psychological counseling auxiliary device provided in one embodiment of the present application. Figure 3 As shown, the psychological counseling auxiliary device 6 of this embodiment includes: at least one processor 60 ( Figure 3 Only one is shown), at least one memory 61 ( Figure 3 Only one is shown in the figure) and a computer program 62 stored in the at least one memory 61 and executable on the at least one processor 60. When the processor 60 executes the computer program 62, the psychological counseling assistance device 6 implements the steps in any of the above-mentioned psychological counseling assistance method embodiments, or enables the psychological counseling assistance device 6 to implement the functions of each unit in the above-mentioned device embodiments.

[0189] For example, the computer program 62 may be divided into one or more modules / units, which are stored in the memory 61 and executed by the processor 60 to implement the present application. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program 62 in the psychological counseling auxiliary device 6.

[0190] The psychological counseling auxiliary device 6 can be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The psychological counseling auxiliary device can include, but is not limited to, a processor 60 and a memory 61. It can be understood by those skilled in the art that Figure 3 It is merely an example of the psychological counseling auxiliary device 6 and does not constitute a limitation of the psychological counseling auxiliary device 6. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, it may also include input and output devices, network access devices, buses, etc.

[0191] The processor 60 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0192] In some embodiments, the memory 61 can be an internal storage unit of the psychological counseling auxiliary device 6, such as the hard drive or memory of the psychological counseling auxiliary device 6. In other embodiments, the memory 61 can also be an external storage device of the psychological counseling auxiliary device 6, such as a plug-in hard drive, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the psychological counseling auxiliary device 6. Furthermore, the memory 61 can also include both the internal storage unit of the psychological counseling auxiliary device 6 and an external storage device. The memory 61 is used to store an operating system, application programs, a boot loader, data, and other programs, such as the program code of the computer program. The memory 61 can also be used to temporarily store data that has been output or is about to be output.

[0193] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0194] An embodiment of the present application provides a computer program product. When the computer program product is run on a psychological counseling assistance device, the psychological counseling assistance device implements the steps in any of the above-mentioned method embodiments.

[0195] If the integrated unit is implemented as a software functional unit and sold or used as a standalone product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the process steps in the above-mentioned method embodiments by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to the psychological counseling assistance device, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signals, telecommunication signals, and software distribution media. Examples include USB flash drives, removable hard drives, magnetic disks, or optical disks. In some jurisdictions, due to legislation and patent practice, computer-readable media cannot be electric carrier signals or telecommunication signals.

[0196] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0197] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0198] In the embodiments provided in the present application, it should be understood that the disclosed psychological counseling auxiliary devices, equipment and methods can be implemented in other ways. For example, the psychological counseling auxiliary devices and equipment embodiments described above are merely schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0199] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0200] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A psychological counseling auxiliary method, characterized in that: include: Obtaining basic data, behavioral data, physiological signals, and input data of students; wherein the basic data includes academic performance data, psychological assessment results, family background information, and consultation records; the behavioral data includes classroom behavior and social data; the physiological signals include heart rate data and skin conductance data; and the input data includes voice data or text data; Performing multimodal data fusion on the basic data, the behavioral data, the physiological signal, and the input data to obtain a psychological state feature vector; Determining whether to trigger a psychological warning based on the basic data, the behavioral data, the physiological signal, and the input data; If it is determined that the psychological warning is triggered, generating a targeted counseling plan based on the psychological state feature vector; Obtain the short-term data and long-term data of the student, and optimize the generation process of the target counseling plan based on the short-term data and long-term data of the student; wherein the short-term data is the heartbeat data of the student after counseling, and the long-term data includes the academic performance data and social data of the student after counseling.

2. The psychological counseling auxiliary method according to claim 1, characterized in that: The performing multimodal data fusion on the basic data, the behavioral data, the physiological signal, and the input data to obtain a psychological state feature vector includes: Constructing a basic portrait vector based on the basic data; Extracting statistical features of classroom behaviors from the behavioral data, converting social data from the behavioral data into emotional intensity and social frequency through sentiment analysis, and integrating the statistical features, the emotional intensity, and the social frequency to obtain a behavioral feature vector; wherein the statistical features include a mean and a variance; Calculating a heart rate variability value based on the heartbeat data of the physiological signal, and calculating a peak frequency of skin conductance data in the physiological signal, and combining the heart rate variability value and the peak frequency to obtain a physiological feature vector; Based on an educational metaphor dictionary, the input data is converted into standard psychological data; the standard psychological data is encoded based on a BERT model to obtain a high-dimensional semantic vector; the high-dimensional semantic vector is reduced in dimensionality to obtain a metaphor feature vector; wherein the educational metaphor dictionary includes a mapping between students' metaphorical expressions and standard psychological descriptions; Performing multimodal data fusion on the basic portrait vector, the behavioral feature vector, the physiological feature vector, and the metaphor feature vector to obtain the psychological state feature vector; The step of constructing a basic portrait vector based on the basic data includes: Calculating the slope and standard deviation of the performance sequence in the performance data to obtain trend characteristics and fluctuation characteristics, and calculating the error rate based on the performance data, and performing vector concatenation of the trend characteristics, fluctuation characteristics, and error rate to obtain a performance feature vector; Normalizing the anxiety scores and depression scores in the psychological assessment results to obtain standardized anxiety scores and depression scores, and integrating the standardized anxiety scores and depression scores to obtain an assessment feature vector; Encoding the family structure in the family background information to obtain a structure vector; normalizing the income data in the family background information to obtain an income vector; performing vector concatenation of the structure vector and the income vector to obtain a family feature vector; Extracting the number of consultations in the consultation record as a consultation feature vector; The performance feature vector, the assessment feature vector, the family feature vector and the consultation feature vector are integrated to obtain the basic portrait vector.

3. The psychological counseling auxiliary method according to claim 2, characterized in that: The method further comprises: Collecting expression texts of students on campus and annotating psychological intentions using a semi-automatic annotation method to obtain a corpus; wherein the corpus includes a mapping between the expression texts and the psychological intentions; Extract high-frequency expression texts from the corpus, and process the expression texts in combination with natural language processing methods to obtain metaphor texts; obtain the corrected metaphor texts to obtain an educational metaphor dictionary; wherein the metaphor text includes a mapping between students' metaphorical expressions and standard psychological descriptions.

4. The psychological counseling auxiliary method according to claim 2, characterized in that: The determining whether to trigger a psychological warning based on the basic data, the behavioral data, the physiological signal, and the input data includes: Determining basic risk based on the basic profile vector; Calculating the concentration corresponding to the classroom behavior in the behavioral data using an LSTM network, and calculating the correlation between the social data in the behavioral data and the physiological signal based on an attention mechanism to obtain a correlation, and determining the behavioral risk based on the concentration and the correlation; Comparing the heart rate variability value with the peak frequency, if the heart rate variability value is greater than the peak frequency, determining the heart rate variability value as a psychological risk; if the peak frequency is greater than the heart rate variability value, determining the peak frequency as the psychological risk; calculating an initial risk based on the metaphor feature vector, identifying whether there is a contradiction between the standard psychological data and the behavioral data based on the standard psychological data and the behavioral data, and if there is a contradiction, obtaining a penalty coefficient, and calculating a metaphor risk based on the initial risk and the penalty coefficient; Based on the basic risk, the behavioral risk, the psychological risk, the metaphorical risk, the synergistic gain coefficient, and the exponential function, the comprehensive risk is calculated, and whether to trigger a psychological warning is determined based on the comprehensive risk; wherein the synergistic gain coefficient and the exponential function are used to increase additional risks when multiple risks exceed corresponding thresholds.

5. The psychological counseling auxiliary method according to claim 1, characterized in that: Generating a target guidance plan based on the psychological state feature vector includes: Searching for a collection of candidate counseling strategies that match the psychological state feature vector from a psychological knowledge graph; wherein the psychological knowledge graph is constructed based on a mapping between psychological problems and psychological counseling strategies; Performing multi-objective optimization on the candidate grooming strategy collection to select the optimal grooming strategy combination; Based on the optimal grooming strategy combination, the basic data and the behavioral data, the optimal grooming strategy combination is adjusted to generate the target grooming plan.

6. The psychological counseling auxiliary method according to claim 2, characterized in that: The method further comprises: Calculating a heart rate variability value after grooming based on the short-term data, and calculating a physiological variability based on the heart rate variability value after grooming and the heart rate variability value; Calculating a behavior improvement degree based on the long-term data, the performance data in the behavior data, and the social data in the behavior data; Obtaining the student's satisfaction with the target counseling program, and determining a first effect score based on the degree of physiological change, the degree of behavioral improvement, and the satisfaction; Based on the first effect score and the historical effect score, the student's psychological resilience index is updated; wherein the historical effect score is used to represent the historical first effect score of the student after each psychological counseling.

7. The psychological counseling auxiliary method according to claim 6, characterized in that: The process of optimizing the generation of the target guidance plan based on the short-term data and long-term data of the student includes: Calculating a recovery rate based on the short-term data, and calculating a performance improvement slope based on the performance data in the long-term data; wherein the recovery rate is used to represent the speed at which the student's heart rate variability value returns to a baseline value, and the performance improvement slope is used to represent the linear regression slope of the student's test score after the counseling; Obtaining a program adoption rate, and determining a second effect score based on the recovery speed, the grade improvement slope, the program adoption rate, and the psychological resilience index; wherein the program adoption rate is used to represent the proportion of students who adopt the psychological counseling program; Constructing a reward function based on the satisfaction and the second effect score; wherein the reward function is used to characterize the feasibility of the target guidance solution; Based on the reward function and the psychological state feature vector, a proximal strategy optimization algorithm is used to optimize the multi-objective optimization process in the process of generating the target grooming plan.

8. The psychological counseling auxiliary method according to any one of claims 1 to 5, characterized in that: The method further comprises: Visualizing the generation process of the target counseling program to obtain a program generation logic chain, and sending the program generation logic chain to the teacher end so that the psychology teacher can view the generation process of the student's target counseling program; When the student rejects the target counseling plan generated three times, an artificial counseling signal and a logic chain of each plan generation are sent to the teacher end so that the psychological teacher can provide psychological counseling to the student.

9. A psychological counseling auxiliary device, characterized in that: include: an acquisition unit, configured to acquire basic data, behavioral data, physiological signals, and input data of the student; wherein the basic data includes academic performance data, psychological assessment results, family background information, and consultation records; the behavioral data includes classroom behavior and social data; the physiological signals include heartbeat data and skin conductance data; and the input data includes voice data or text data; a multimodal data fusion unit, configured to perform multimodal data fusion on the basic data, the behavioral data, the physiological signal, and the input data to obtain a psychological state feature vector; a psychological warning triggering unit, configured to determine whether to trigger a psychological warning based on the basic data, the behavioral data, the physiological signal, and the input data; a target guidance scheme generating unit, configured to generate a target guidance scheme based on the psychological state feature vector if it is determined that the psychological warning is triggered; An optimization unit is used to obtain the short-term data and long-term data of the student, and optimize the generation process of the target counseling plan based on the short-term data and long-term data of the student; wherein the short-term data is the heartbeat data of the student after counseling, and the long-term data includes the academic performance data and social data of the student after counseling.

10. A psychological counseling auxiliary device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 8 is implemented.

Images