Intelligent interpretation system for result of psychological evaluation meter

Abstract

The invention provides a psychological evaluation table result intelligent interpretation system which specifically comprises a psychological data collection module used for collecting a psychological evaluation text data set, an interpretation model training module used for training an initial interpretation model according to the psychological evaluation text data set, and a psychological evaluation table result intelligent interpretation module. Obtaining an interpretation text corresponding to each piece of psychological evaluation data on the psychological evaluation theme; the psychological state evaluation module is used for calculating psychological state evaluation indexes according to the interpretation text and selecting the initial interpretation model corresponding to the optimal psychological state evaluation index as a final interpretation model; and the psychological evaluation result interpretation module is used for calling the final interpretation model to interpret the psychological evaluation table result to be interpreted to obtain a corresponding interpretation result. Through the method and the device, the problems of high calculation overhead and low efficiency under a specific task due to the fact that the psychological evaluation table result is interpreted by only depending on a large language model in the prior art are solved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an intelligent interpretation system for psychological assessment scale results. Background Art

[0002] With the prevalence of mental health problems, psychological assessments are increasingly used in the fields of medicine, education, and business. As an important tool for quantifying psychological states, psychological assessment scales can help mental health professionals assess an individual's mental health, emotional state, and behavioral tendencies. However, the results of psychological assessment scales are mostly presented in the form of scores, and the interpretation process relies on the experience and subjective judgment of psychology experts. There is a lack of standardized interpretation paths, and the professional background of the interpreters is required to be high. This may lead to inconsistencies and limitations in the interpretation results, especially for non-professionals, it is more difficult to understand and apply the scale results.

[0003] At the same time, faced with the high demand and diverse characteristics of scale results, manual interpretation methods are difficult to meet the needs of fast and large-scale analysis. Although some automated report generation tools have been developed, they are mostly presented in the form of simple rule matching or score comparison tables. They lack in-depth analysis of individual psychological states and are difficult to provide accurate mental health interpretations. This has limited the accessibility of mental health services, especially in areas where there is a lack of professional mental health practitioners, making it even more difficult to popularize the professional application of psychological assessments.

[0004] In recent years, large language models (LLMs) have made great progress in the field of natural language understanding and generation. Through training on large-scale corpora, LLMs have the ability to understand the semantics of complex texts and can adapt to the fine-tuning needs of different fields. They are suitable for specialized tasks such as psychological counseling and assessment interpretation. However, relying solely on large language models may lead to high computational overhead and inefficiency in specific tasks. Summary of the invention

[0005] The present invention provides a system for intelligently interpreting the results of a psychological assessment scale, which is used to solve the problem in the prior art that only relying on a large language model to interpret the results of a psychological assessment scale will bring high computational overhead and be inefficient under specific tasks.

[0006] The present invention provides a system for intelligently interpreting the results of a psychological assessment scale, comprising a psychological data collection module, an interpretation model training module, a psychological state assessment module, and a psychological assessment result interpretation module, wherein the psychological data collection module is used to collect a psychological assessment text data set, wherein the psychological assessment text data set is classified according to psychological assessment topics, and each psychological assessment data item in the psychological assessment text data set is annotated with an emotional state label; The interpretation model training module is used to train an initial interpretation model according to the psychological assessment text data set to obtain an interpretation text corresponding to each psychological assessment data on the psychological assessment topic, wherein the initial interpretation model is integrated by a large language model and a natural language processing model; The psychological state evaluation module is used to calculate the psychological state evaluation index according to the interpretation text, and select the initial interpretation model corresponding to the optimal psychological state evaluation index as the final interpretation model, wherein the psychological state evaluation index includes the BLEU score, the MSB score, the emotion accuracy rate of the emotion state label, the conformity score of the psychological evaluation scale, and the emotion deviation degree of the emotion state label; The psychological assessment result interpretation module is used to call the final interpretation model to interpret the psychological assessment scale results to be interpreted, and obtain corresponding interpretation results.

[0007] Furthermore, each psychological assessment data in the psychological assessment text data set also includes a pre-interpreted reference text, and the BLEU score is calculated based on the reference text and the interpreted text. The process of the psychological state assessment module calculating the BLEU score includes: Calculating the n-gram overlap probability between the reference text and the interpreted text as an interpretation accuracy score; Determine the geometric mean of the interpretation accuracy scores corresponding to each n-gram; The geometric mean is weighted according to a BP penalty coefficient to obtain a BLEU score, wherein the BP penalty coefficient is determined according to a size relationship between a text length of the reference text and a text length of the interpreted text.

[0008] Furthermore, each psychological assessment scale in the psychological assessment data is provided with a different quantitative weight, the interpretation text includes the scale assessment score of each psychological assessment scale, and the process of the psychological state assessment module calculating the MSB score includes: Determining the overall score of the psychological assessment data according to the quantitative weight and the scale assessment score; Determining the standardized score of the psychological assessment scale according to the mean and standard deviation of the scale assessment scores; Determine the time interval between each psychological assessment scale, and determine the score change value of the corresponding scale assessment score within the time interval; An MSB score is calculated based on the overall score, the standardized score, and the score change value.

[0009] Further, the calculating of the MSB score based on the overall score, the standardized score and the score change value includes: The standardized scores of each psychological assessment scale are summed, and the control outliers of the psychological assessment data are calculated based on the summation results; Calculating the psychological state fluctuation value of the psychological evaluation data according to the score change value of the scale evaluation score; The overall score, the control abnormality value and the psychological state fluctuation value are weighted according to preset weights, and the weighted results are summed to obtain the MSB score.

[0010] Furthermore, the interpretation text also includes the emotion interpretation label of the psychological assessment data and the corresponding emotion interpretation score. The process in which the psychological state evaluation module calculates the emotion accuracy rate of the emotion state label and the emotion deviation degree of the emotion state label includes: Determining the number of accurate interpretations of the emotional state label in the interpretation text, wherein the number of accurate interpretations is the number of times the emotional interpretation label completely matches the emotional state label corresponding to the psychological assessment data; The ratio of the number of accurate interpretations to the total number of psychological assessment data is used as the emotion accuracy rate of the emotion state label; The difference between the reference emotion score and the emotion interpretation score is used as the emotion deviation of the emotion state label, wherein the reference emotion score is the evaluation score for the emotion state label in the reference text corresponding to the psychological assessment data.

[0011] Furthermore, the psychological state assessment module calculates the compliance score of the psychological assessment scale, including: Determining the number of times the interpretation content of the interpretation text for the psychological assessment scale is in compliance, the number of times the interpretation content is in compliance is the expected number of times the interpretation content of the interpretation text for the psychological assessment scale is in compliance with the psychological assessment content of a specific field; The ratio of the number of times the interpretation content is consistent to the total number of the psychological assessment content is used as the compliance score of the psychological assessment scale.

[0012] The present invention also provides a method for intelligently interpreting the results of a psychological assessment scale, the method comprising the following steps: Collecting a psychological assessment text data set, wherein the psychological assessment text data set is classified according to psychological assessment topics, and each psychological assessment data in the psychological assessment text data set is annotated with an emotional state label; Training an initial interpretation model according to the psychological assessment text data set to obtain an interpretation text corresponding to each psychological assessment data on the psychological assessment topic, wherein the initial interpretation model is an integration of a large language model and a natural language processing model; Calculating a psychological state evaluation index according to the interpretation text, and selecting an initial interpretation model corresponding to the optimal psychological state evaluation index as the final interpretation model, wherein the psychological state evaluation index includes a BLEU score, an MSB score, an emotion accuracy rate of the emotion state label, a conformity score of the psychological evaluation scale, and an emotion deviation degree of the emotion state label; The final interpretation model is called to interpret the psychological assessment scale results to be interpreted, and the corresponding interpretation results are obtained.

[0013] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for intelligently interpreting the results of the psychological assessment scale as described above is implemented.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for intelligently interpreting the results of the psychological assessment scale as described above.

[0015] The present invention also provides a computer program product, comprising a computer program, which, when executed by a processor, implements the method for intelligently interpreting the results of the psychological assessment scale as described above.

[0016] The intelligent interpretation system for the results of the psychological assessment scale provided by the present invention integrates the interpretation model through a large language model and a natural language processing model. The large language model is specifically responsible for general semantic understanding and generation tasks, and the natural language processing model focuses on the detailed interpretation of emotional states and specific psychological assessment scales. The collaborative interpretation of the two models reduces the computational overhead, and enables the system to have both deep understanding and rapid response during the interpretation process. In addition, when evaluating the psychological state, the system selects the optimal final interpretation model through the psychological state evaluation indicators under a variety of specific tasks, so that the system can achieve efficient interpretation and focused analysis under multiple specific tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced one by one below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 It is a schematic diagram of the architecture of the intelligent interpretation system of the psychological assessment scale results provided by the present invention.

[0019] Figure 2 Schematic diagram of interpreting the results of the psychological assessment scale provided by the embodiment of the present invention Figure 3 It is a flow chart of the method for intelligently interpreting the results of a psychological assessment scale provided by the present invention.

[0020] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] The intelligent interpretation system of the psychological assessment scale results provided by the present invention can be integrated into an online psychological assessment platform (such as a psychological counseling simulation system) specifically used for psychological assessment through an API interface. The intelligent interpretation method of the psychological assessment scale results provided by the present invention is applied to the intelligent interpretation system of the psychological assessment scale results, which is used to assist the psychological assessment platform in interpreting the psychological assessment scale results to be interpreted input by the user, and to feed back the corresponding interpretation results to the user through the psychological assessment platform.

[0023] The following describes the intelligent interpretation system of psychological assessment scale results provided by the present invention in conjunction with the accompanying drawings. Figure 1 Schematic diagram of the architecture of the intelligent interpretation system for psychological assessment scale results provided by the present invention. Figure 1 As shown, the intelligent interpretation system for psychological assessment scale results (hereinafter referred to as the system) includes a psychological data collection module, an interpretation model training module, a psychological state assessment module, and a psychological assessment result interpretation module, a total of four modules, which are explained one by one below.

[0024] First, there is the psychological data collection module of the system, which is mainly used to collect psychological assessment text data sets. Since the system mainly relies on the interpretation model to interpret the results of the psychological assessment scale, it is necessary to collect psychological assessment text data sets to fine-tune the interpretation model. Therefore, for the data set, the embodiment of the present invention collects a large amount of psychological consultation and psychology-related text, audio, and video source data, and performs processing operations such as sorting and cleaning on these data.

[0025] The sources of data include psychological scale database, clinical psychological assessment reports, public psychology journals and books, Internet psychological support platform, education industry psychological assessment platform, etc. Among them, the psychological scale database comes from mental health research institutions, mainly including text data of psychological assessment scales, which involve multiple psychological assessment topics, including depression, anxiety, emotion management, self-esteem scales and other topics. Clinical psychological assessment reports come from anonymous research reports of mental health research institutions, and the data volume of both is 591. Public psychology journals and books come from academic research results or theoretical knowledge of some psychology journals, and the data volume is 8870. The Internet psychological support platform comes from the psychological support platform data recorded in some psychological support websites on the Internet, specifically anonymous user mental health data, with a data volume of 5699. The education industry psychological assessment platform comes from some educational psychological assessment data recorded in the school or enterprise mental health platform, specifically the psychological assessment results and trend data of teachers, students or staff of schools or enterprises, with a data volume of 7912.

[0026] The psychological data collection module of the system of the embodiment of the present invention relies on a large amount of high-quality psychological health data sources and professional knowledge bases to ensure that the system can generate high-level interpretation results. The collected data specifically comes from multiple psychological assessment channels, including processed psychological assessment scale data, clinical psychological health reports, psychology professional literature, teaching and training materials, etc., covering multiple specific psychological assessment topics such as depression, anxiety, emotional management, self-esteem, etc., providing rich interpretation scenarios and complete semantic knowledge for the interpretation model.

[0027] After the data set is collected, it is sorted and cleaned. Specifically, the collected data is desensitized to remove all personal identification information (such as name, address, etc.), and specific fields such as age, gender, psychological assessment time, etc. are standardized. Using NLP preprocessing methods, all text is converted to lowercase, stop words are removed, and extra spaces and symbols are deleted.

[0028] In the embodiment of the present invention, the psychological assessment text data set is classified according to the psychological assessment theme, that is, each psychological assessment data in the psychological assessment text data set has a corresponding psychological assessment theme, which is used to characterize the type of psychological assessment. After the collected data is integrated and cleaned, it needs to be classified according to the psychological assessment theme, which includes depression theme, anxiety theme, emotion management theme, self-esteem scale theme, etc.

[0029] In addition, each psychological assessment data in the psychological assessment text dataset is annotated with an emotional state label. Here, each psychological assessment data includes multiple psychological assessment scales and corresponding psychological text descriptions, and the psychological text descriptions are annotated with emotional state labels, specifically positive emotions, negative emotions, and neutral emotions. These emotional state labels provide clear label classification for the system's interpretation of the emotion management topic during the emotion recognition task. In addition, each emotional state label has a corresponding reference emotional score as the evaluation score of each emotional state.

[0030] In order to fine-tune the interpretation model, the embodiment of the present invention also adds pre-interpreted reference texts to each psychological assessment data in the psychological assessment text data set. These reference texts are obtained by pre-interpreting the psychological assessment data through psychological experts or other psychological assessment systems and platform software, and are used as fine-tuning standards for the interpretation model. When fine-tuning the interpretation model, the interpretation model can be optimized or the interpretation effect of the interpretation model can be evaluated by constructing a loss function.

[0031] In addition, each psychological assessment scale in the psychological assessment data is set with a different quantitative weight, because the psychological assessment scale is quantitatively evaluated according to different psychological items, including emotional state, behavioral tendency, self-cognition, interpersonal relationship and coping strategy. These are all items that express human psychological performance, and each item has a corresponding psychological assessment scale. The evaluation results of these items have different contributions to the final quantitative evaluation. Therefore, the embodiment of the present invention sets different quantitative weights for each psychological assessment scale in the psychological assessment data, specifically: emotional state setting accounts for 30%, behavioral tendency setting accounts for 25%, self-cognition setting accounts for 20%, interpersonal relationship setting accounts for 15%, and coping strategy setting accounts for 10%.

[0032] Through the collection of the above data, the psychological text description in each psychological assessment data can be exemplarily represented as: "Text: In the past week, I often felt anxious and even had difficulty falling asleep; Label: {Topic: Anxiety; Emotional state: Negative emotion; Quantification weight: 0.3; Emotional score: 4}.

[0033] In the embodiment of the present invention, the collected data is strictly cleaned and labeled to ensure accurate labeling of information such as emotional state and behavioral tendency. In order to enable the system to identify emotional characteristics such as emotional deviation and compliance score, multi-layer labels such as topics and emotions and quantitative weight labels are used to facilitate the model to generate interpretation results that are more in line with mental health standards. By anonymizing the data, it meets strict user data protection standards and provides users with safe mental health interpretation services.

[0034] The second is the interpretation model training module of the system, which is used to train the initial interpretation model based on the psychological assessment text data set to obtain the corresponding interpretation text of each psychological assessment data on the psychological assessment topic. The initial interpretation model is integrated by the large language model and the natural language processing model. The large language model is responsible for performing general semantic understanding and generation tasks, which can be the GLM4-9b model. The natural language processing model (Natural Language Processing, NLP) is a small NLP model that focuses on the detailed interpretation of emotional states and specific psychological assessment scales, which can be the TinyBERT model, to achieve accurate interpretation of different psychological measurement dimensions.

[0035] The training process of the initial interpretation model is also the process of model fine-tuning. When training the initial model, the corresponding psychological assessment text dataset is input according to the preset sample batch size, and then training is performed according to the preset initial learning rate (for example, 5e-5), the number of gradient accumulation steps (for example, 4), and the SFT fine-tuning parameters. The training rounds are set to 10 rounds. In each round of training, 80% of the input data set is randomly selected as the training set, 10% as the validation set, and 10% as the test set.

[0036] In each round of training, the initial interpretation model will obtain the corresponding interpretation text of each psychological assessment data on the psychological assessment topic, which will be used for verification or evaluation of the subsequent psychological state assessment module.

[0037] In an embodiment of the present invention, an interpretation model is integrated through a large language model and a natural language processing model. The large language model is specifically used for general semantic understanding and generation tasks, and the natural language processing model focuses on the detailed interpretation of emotional states and specific psychological assessment scales. The dual models work together to interpret, which can reduce computing overhead and enable the system to have both deep understanding and rapid response during the interpretation process.

[0038] Next is the psychological state evaluation module of the system, which calculates the psychological state evaluation index based on the interpreted text, and selects the initial interpretation model corresponding to the optimal psychological state evaluation index as the final interpretation model. In the embodiment of the present invention, in order to enable the interpretation model to be competent for the interpretation of psychological results under a variety of specific tasks, psychological state evaluation indicators under a variety of specific tasks are set for the interpretation model during training to measure the training effect of the model. The psychological state evaluation indicators include BLEU (bilingual evaluation understudy) score, MSB (Mental State Benchmark) score, emotional accuracy of emotional state labels, conformity score of psychological assessment scales, and emotional deviation of emotional state labels.

[0039] The following is a description of the process of calculating each psychological state evaluation indicator in the psychological state evaluation module. The first is the BLEU score, which is used to measure the text similarity between the interpreted text and the pre-interpreted reference text. First, the n-gram algorithm is used to calculate the n-gram overlap probability between the reference text and the interpreted text as the interpretation accuracy score, denoted as , expressed as the following formula (1): (1) In the above formula (1), n ​​represents the value of n set by the n-gram algorithm, which can be reasonably set according to actual needs. Indicates the count of overlapping content text between the interpreted text and the reference text under the gram segment. Indicates the total content text count in the reference text under the gram segment.

[0040] By setting different n values ​​through the n-gram algorithm, the interpretation accuracy score of each n-gram can be calculated according to the above formula (1): , and then determine the geometric mean of the interpretation accuracy scores for each n-gram. Here, for each n-gram, the interpretation accuracy score , set the corresponding weight , when calculating the geometric mean, first use the weights Interpretation accuracy score After weighting and calculating the geometric mean, the geometric mean is weighted according to the BP penalty system to obtain the BLEU score. The calculation formula is as follows: (2) In the above formula (2), exp is the exponential function and N is the interpretation accuracy score. The total number of n-grams, that is, the total number of different n values ​​set by the n-gram algorithm, where n represents the interpretation accuracy score of the nth n-gram , BP is the BP penalty coefficient, which is determined based on the relationship between the text length of the reference text and the text length of the interpreted text. It is used to eliminate the difference in text length between the reference text and the interpreted text. The value of the BP penalty coefficient can be expressed as the following formula (3): (3) In the above formula (3), f represents the text length of the reference text, c represents the text length of the interpreted text, and exp represents the exponential function.

[0041] In an embodiment of the present invention, when training an interpretation model, the BLEU scores of the interpreted text and the reference text are calculated to measure the text similarity between the interpreted text and the reference text. The text similarity is used as one of the psychological state evaluation indicators to evaluate the effect of the interpretation model training, which helps to improve the accuracy of the interpretation model in generating the interpretation text.

[0042] The embodiment of the present invention is aimed at the psychological state evaluation index of the interpreted text, which also includes calculating the MSB score. The MSB score is also called the comprehensive mental health index of the psychological assessment scale, which is used to measure the comprehensive mental health state evaluation of the psychological assessment scale in the interpreted text, and train the interpretation model to perform detailed interpretation and in-depth evaluation of the psychological assessment scale.

[0043] The calculation of the MSB score is divided into three parts. The first part is the overall score for the psychological assessment data, the second part is the anomaly detection score, and the third part is the score change value of the psychological state trend. The specific calculation process of each part is explained below.

[0044] First is the overall score of the first part. This overall score is calculated based on the quantitative weights set in the psychological assessment scale, and is used to reflect the comprehensive scores on multiple psychological assessment scales. When calculating, the overall score of the psychological assessment data is determined based on the quantitative weights and the scale assessment scores. As mentioned earlier, each item has a corresponding psychological assessment scale, and each item has a corresponding quantitative weight. That is, each psychological assessment scale in the psychological assessment data is set with different quantitative weights. After being interpreted by the initial interpretation model, the interpretation text includes the scale evaluation score of each psychological assessment scale as the mental health indicator of the corresponding project, recorded as , and the overall score of the psychological assessment data The calculation formula can be expressed as follows: (4) In the above formula (4), n represents the total number of psychological assessment scales of the corresponding items in the psychological assessment data, i represents the i-th psychological assessment scale, represents the quantitative weight corresponding to the i-th psychological assessment scale, Represents the scale evaluation score of the i-th psychological evaluation scale in the interpretation text.

[0045] The second part is the anomaly detection score, which is used to balance the impact of outliers in the scale evaluation scores. Since there may be outliers in the evaluation scores of each psychological evaluation scale, for example, the score of one item is much larger than the other scores. Therefore, it is necessary to perform anomaly detection on the scale evaluation scores to calculate the standardized scores of the psychological evaluation scale. The mean and standard deviation of the psychological assessment scale are used to determine the standardized scores. First, multiple scale assessment scores are calculated. The mean and standard deviation , and then calculate the standardized score of the psychological assessment scale based on the mean and standard deviation, recorded as , expressed as the following formula: (5) The last third part is the score change value of the psychological state trend. The score change value is to determine the trend of psychological health changes and capture the fluctuation of psychological state. For long-term tracking of psychological state data, the scale evaluation scores of the psychological assessment scale at different time points are calculated to identify the trend of changes in psychological health status, which helps to monitor the progress or deterioration of the individual's psychological state. When calculating, determine the time interval between each psychological assessment scale, such as arrive , this time interval can be determined according to the preset time node, such as the 5th to 15th day or the 1st to 3rd month. Next, the score change value of the corresponding scale evaluation score is determined within the score interval, which can be recorded as , expressed as the following formula: (6) In the above formula (6), At any point in time The following are the scale evaluation scores corresponding to the psychological evaluation scale. It means that at any point in time The following are the scale evaluation scores corresponding to the psychological assessment scale.

[0046] Finally, the MSB score is calculated based on the overall score, standardized score, and score change value, that is, the score values ​​calculated in the three parts are combined to obtain the MSB score.

[0047] In the embodiment of the present invention, when training the interpretation model, the MSB score of the psychological assessment data is obtained by calculation, which is used to measure the comprehensive mental health status assessment of the psychological assessment scale in the interpreted text, improve the interpretation model's ability to interpret and deeply assess the psychological assessment scale in detail, and achieve the effect of detailed interpretation.

[0048] The following describes the process of calculating the MSB score based on the overall score, standardized score, and score change value. First, the standardized score of each psychological assessment scale is calculated. Sum the values ​​and calculate the control abnormal value of the psychological evaluation data based on the summation result, and then calculate the score change value of the scale evaluation score , calculate the psychological state fluctuation value of the psychological assessment data, and finally weight the overall score, control abnormal values ​​and psychological state fluctuation value according to the preset weights, and sum the weighted results to obtain the MSB score, recorded as .

[0049] Here, different weights are assigned to the overall score, control abnormal values, and psychological state fluctuation values ​​to adjust the contribution of each part to the MSB score. The score values ​​are weighted by the weights, and then the three weighted results are summed to finally obtain the corresponding comprehensive mental health index, that is, the MSB score, which is expressed as the following formula: (7) In the above formula (7), represents the weight corresponding to the overall score, Represents the weight corresponding to the control outliers of the psychological evaluation data, represents the weight corresponding to the fluctuation value of the psychological state, and f represents the calculated function.

[0050] In the embodiment of the present invention, when the psychological assessment scale is carefully interpreted and deeply evaluated, the overall average score, abnormal value score and psychological state fluctuation value of the psychological assessment scale are integrated to reflect the evaluation status of the psychological assessment scale. The overall score is used to interpret specific scale items, the evaluation error caused by the scale score is reduced by controlling the abnormal value, and the dynamic evaluation of the psychological state change is achieved through the psychological state fluctuation value. Overall, the interpretation model's ability to carefully interpret and deeply evaluate the psychological assessment scale is improved.

[0051] In the field of psychological assessment, the most critical thing is the prediction of emotional state, so the interpretation model must have the ability to interpret emotional state. Therefore, in the specific task of emotion prediction, its psychological state evaluation indicators must include the emotion accuracy and emotion deviation of the emotion state label to achieve accurate prediction of the emotion state by the interpretation model. The following describes the process of calculating the emotion accuracy and emotion deviation of the emotion state label by the psychological state evaluation module.

[0052] When training the initial interpretation model, it is necessary to determine the accurate number of interpretations of the emotional state labels in the interpretation text. The accurate number of interpretations is the number of times the emotional interpretation labels completely match the emotional state labels corresponding to the psychological assessment data. Because the interpretation text generated by the initial interpretation model also includes the emotional interpretation labels of the psychological assessment data and the corresponding emotional interpretation scores. That is, the initial interpretation model is required to classify the emotional state, distinguish between positive emotions, negative emotions, and neutral emotions, and generate corresponding emotional interpretation labels, and then match them with the emotional state labels corresponding to the psychological assessment data to determine whether they match. If they completely match, it means that the emotional state classification is correct. If they do not match, it means that the emotional state classification is wrong.

[0053] In this way, each psychological assessment data is matched and then statistically analyzed. The number of complete matches is taken as the number of accurate interpretations, that is, the number of successful classifications. Then the proportion of the number of successful classifications in the total number of samples is calculated, that is, the ratio of the number of accurate interpretations to the total number of psychological assessment data, as the emotion accuracy of the emotional state label, recorded as Emotion Accuracy, expressed as: (8) As mentioned before, each emotional state label has a corresponding reference emotional score, denoted as In the interpretation text generated by the initial interpretation model, each sentiment interpretation label also has a corresponding sentiment interpretation score, recorded as , both of which can be used to calculate the emotional deviation of the initial interpretation model for the emotional state label, which is used to measure the deviation between the interpreted emotional state generated by the initial interpretation model and the actual state. Here, the difference between the reference emotional score and the emotional interpretation score is used as the emotional deviation of the emotional state label. , where the reference emotion score is the evaluation score for the emotion state label in the reference text corresponding to the psychological assessment data. The calculation process of emotion deviation is expressed as the following formula: (9) In the above formula (9), N is the total number of samples of psychological assessment data, and i represents the i-th psychological assessment data.

[0054] In an embodiment of the present invention, the emotional state label is used to enable the initial interpretation model to predict the emotional state. By calculating the emotion accuracy and emotion deviation, the prediction result of the emotional state can be effectively evaluated, thereby ensuring that the interpretation model is closer to the actual situation in emotion interpretation and achieving accurate prediction of the emotional state.

[0055] In addition, considering that the interpretation text is the text generated by the initial interpretation model for the psychological assessment scale, the generated text may deviate from the expectation, that is, the interpretation content of the psychological assessment scale in the generated interpretation text may not conform to the psychological assessment content in a specific field, for example, whether it conforms to the standards of emotion interpretation or the guiding principles of the psychological assessment scale in the specific field of emotion interpretation. The standards for emotion interpretation are generally determined from people's subjective experience, physiological arousal, and external and facial expressions, while the generated interpretation text may not identify emotions according to the standards for emotion interpretation. Furthermore, the guiding principles for emotions in the psychological assessment scale are also targeted guidance based on the standards for emotion interpretation, while the guidance proposed in the interpretation text may not conform to the guiding principles.

[0056] The occurrence of the above situation may cause the interpretation content of the psychological assessment scale in the interpretation text to be inconsistent with the expected psychological assessment content in a specific field. Therefore, in order to eliminate the generation deviation of the initial interpretation model for the psychological assessment content, the embodiment of the present invention also sets an indicator of the conformity score of the psychological assessment scale in the psychological state evaluation index, which is explained in detail below.

[0057] First, determine the number of times the interpretation content in the interpretation text meets the psychological assessment scale. The number of times the interpretation content in the interpretation text meets the psychological assessment scale is the expected number of times the interpretation content in the interpretation text meets the psychological assessment content in a specific field. Here, count the number of times the interpretation content generated in the interpretation text meets the standard, also known as the number of contents that meet the standard. The ratio of the number of times the interpretation content meets the standard to the total number of psychological assessment contents is used as the conformity score of the psychological assessment scale. , that is, calculate the ratio of the number of contents that meet the standards to the total number of contents, expressed as: (10) In this way, in the psychological state assessment module of the system, the psychological state evaluation index can be calculated according to the interpretation of the text during each round of training. In the embodiment of the present invention, the number of training rounds is set to 10 rounds. Therefore, after 10 training rounds, 10 groups of psychological state evaluation indicators can be obtained, each group of psychological state evaluation indicators includes BLEU score, MSB score, emotion accuracy of emotion state label, conformity score of psychological assessment scale, emotion deviation of emotion state label, and the specific training results can be seen in Table 1 below: Table 1:

[0058] It can be seen from Table 1 above that with each training round, the corresponding various psychological state evaluation indicators are gradually improved, and the effect is getting better and better. Therefore, after 10 training rounds, the psychological state evaluation module of the system selects the initial interpretation model corresponding to the optimal psychological state evaluation indicator as the final interpretation model.

[0059] The optimal psychological state evaluation index means that each psychological state evaluation index reaches the optimal value or the comprehensive optimal value of multiple psychological state evaluation indexes. Each psychological state evaluation index reaches the optimal value, which can be achieved by setting an index threshold for each psychological state evaluation index. When each psychological state evaluation index reaches the corresponding index threshold, the optimal psychological state evaluation index can be determined.

[0060] According to Table 1, some psychological state evaluation indicators will decrease in some training rounds, and the indicators in the 10th training round are still lower than those in the previous training rounds. Therefore, the comprehensive optimality of the psychological state evaluation indicators can be considered when most (60% or 80%) of the psychological state evaluation indicators reach the optimal level.

[0061] According to the above selection criteria, the initial interpretation model at the 10th training round can be selected from Table 1 as the final interpretation model.

[0062] The last part is the psychological assessment result interpretation module of the system, which calls the final interpretation model to interpret the results of the psychological assessment scale to obtain the corresponding interpretation results.

[0063] After the final interpretation model is selected by the psychological state assessment module of the system, the psychological assessment result interpretation module can deploy the final interpretation model in the system, and then integrate it with the psychological assessment platform (such as the psychological counseling simulation system) through the API interface, receive the psychological assessment scale results to be interpreted by the user input into the psychological assessment platform, and then call the final interpretation model to accurately interpret the input psychological assessment scale results to be interpreted, and generate the corresponding interpretation text. Finally, the interpretation text is fed back to the user through the psychological assessment platform as the interpretation result of the system.

[0064] The interpretation process of the system can be found in Figure 2 As shown, Figure 2 It is a schematic diagram of the interpretation of the psychological assessment scale results provided by an embodiment of the present invention. The figure shows that after the user inputs the psychological assessment scale results to be interpreted on the psychological assessment platform, the system monitors and obtains the psychological assessment scale results to be interpreted, and then calls the deployed final interpretation model to interpret them, and finally outputs the corresponding interpretation text as the interpretation result, and displays the interpretation text of the corresponding psychological assessment scale results through the psychological assessment platform for users to browse.

[0065] In the embodiment of the present invention, an interpretation model is integrated through a large language model and a natural language processing model. The large language model is specifically used for general semantic understanding and generation tasks, and the natural language processing model focuses on the detailed interpretation of emotional states and specific psychological assessment scales. The collaborative interpretation of the two models reduces the computational overhead, and enables the system to have both deep understanding and rapid response during the interpretation process. In addition, the system collects data sets of various psychological assessment topics to train the interpretation model, so that the system can provide multi-level interpretation reports for different psychological theme dimensions. The interpretation of each theme dimension is generated by multiple models in a collaborative manner, so that the final report not only covers the individual's current psychological state, but also provides comprehensive and forward-looking psychological analysis. When evaluating the psychological state, the optimal final interpretation model is selected through the psychological state evaluation indicators under a variety of specific tasks, so that the system can achieve efficient interpretation and focused analysis under multiple specific tasks.

[0066] The embodiment of the present invention also provides a method for intelligent interpretation of psychological assessment scale results, which is applied to the above-mentioned intelligent interpretation system for psychological assessment scale results. Figure 3 , Figure 3 is a flow chart of the method for intelligent interpretation of psychological assessment scale results provided by the present invention, such as Figure 3 As shown, the method includes the following steps 101 to 104.

[0067] Step 101: Collect a psychological assessment text data set. The psychological assessment text data set is classified according to psychological assessment topics. Each psychological assessment data in the psychological assessment text data set is annotated with an emotional state label.

[0068] Step 102: train an initial interpretation model based on the psychological assessment text data set to obtain the interpretation text corresponding to each psychological assessment data on the psychological assessment topic. The initial interpretation model is integrated by a large language model and a natural language processing model.

[0069] Step 103: Calculate the psychological state evaluation index according to the interpreted text, and select the initial interpretation model corresponding to the optimal psychological state evaluation index as the final interpretation model. The psychological state evaluation index includes BLEU score, MSB score, emotion accuracy of emotion state label, conformity score of psychological assessment scale, and emotion deviation of emotion state label.

[0070] Step 104: Call the final interpretation model to interpret the psychological assessment scale results to obtain corresponding interpretation results.

[0071] It should be noted that the implementation process of the above steps 101 to 104 can correspond to the psychological data collection module, the interpretation model training module, the psychological state evaluation module and the psychological assessment result interpretation module in the intelligent interpretation system of the psychological assessment scale results. In addition, the intelligent interpretation method of the psychological assessment scale results here and the beneficial effects of the intelligent interpretation system of the psychological assessment scale results mentioned above can also correspond to each other, so the specific implementation process and beneficial effects of the intelligent interpretation method of the psychological assessment scale results are not repeated here.

[0072] Figure 4 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 4 As shown, the electronic device may include: a processor (processor) 410 , a communication interface (Communications Interface) 420 , a memory (memory) 430 and a communication bus 440 , wherein the processor 410 , the communication interface 420 , and the memory 430 communicate with each other through the communication bus 440 . The processor 410 can call the logic instructions in the memory 430 to execute the intelligent interpretation method of the psychological assessment scale results, which includes: collecting a psychological assessment text data set, wherein the psychological assessment text data set is classified according to the psychological assessment topic, and each psychological assessment data in the psychological assessment text data set is marked with an emotional state label; training an initial interpretation model according to the psychological assessment text data set to obtain an interpretation text corresponding to each psychological assessment data on the psychological assessment topic, wherein the initial interpretation model is integrated by a large language model and a natural language processing model; calculating a psychological state evaluation index according to the interpretation text, and selecting the initial interpretation model corresponding to the optimal psychological state evaluation index as the final interpretation model, wherein the psychological state evaluation index includes a BLEU score, an MSB score, an emotional accuracy rate of the emotional state label, a conformity score of the psychological assessment scale, and an emotional deviation of the emotional state label; calling the final interpretation model to interpret the psychological assessment scale results to be interpreted to obtain corresponding interpretation results.

[0073] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0074] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the intelligent interpretation method of the psychological assessment scale results provided by the above methods, the method including: collecting a psychological assessment text data set, the psychological assessment text data set is classified according to the psychological assessment topic, and each psychological assessment data in the psychological assessment text data set is marked with an emotional state label; training an initial interpretation model according to the psychological assessment text data set to obtain an interpretation text corresponding to each psychological assessment data on the psychological assessment topic, the initial interpretation model is integrated by a large language model and a natural language processing model; calculating a psychological state evaluation index according to the interpretation text, and selecting the initial interpretation model corresponding to the optimal psychological state evaluation index as the final interpretation model, the psychological state evaluation index including a BLEU score, an MSB score, an emotional accuracy rate of the emotional state label, a conformity score of the psychological assessment scale, and an emotional deviation of the emotional state label; calling the final interpretation model to interpret the psychological assessment scale results to be interpreted to obtain corresponding interpretation results.

[0075] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the intelligent interpretation method of the psychological assessment scale results provided by the above-mentioned methods, the method comprising: collecting a psychological assessment text data set, wherein the psychological assessment text data set is classified according to the psychological assessment topic, and each psychological assessment data in the psychological assessment text data set is marked with an emotional state label; training an initial interpretation model according to the psychological assessment text data set to obtain an interpretation text corresponding to each psychological assessment data on the psychological assessment topic, wherein the initial interpretation model is integrated by a large language model and a natural language processing model; calculating a psychological state evaluation index according to the interpretation text, and selecting the initial interpretation model corresponding to the optimal psychological state evaluation index as the final interpretation model, wherein the psychological state evaluation index includes a BLEU score, an MSB score, an emotional accuracy rate of the emotional state label, a conformity score of the psychological assessment scale, and an emotional deviation of the emotional state label; calling the final interpretation model to interpret the psychological assessment scale results to be interpreted to obtain corresponding interpretation results.

[0076] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0077] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0078] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A system for intelligent interpretation of psychological assessment scale results, comprising a psychological data collection module, an interpretation model training module, a psychological state assessment module, and a psychological assessment result interpretation module, characterized in that: The psychological data collection module is used to collect a psychological assessment text data set, wherein the psychological assessment text data set is classified according to psychological assessment topics, and each psychological assessment data in the psychological assessment text data set is marked with an emotional state label; The interpretation model training module is used to train an initial interpretation model according to the psychological assessment text data set to obtain an interpretation text corresponding to each psychological assessment data on the psychological assessment topic, wherein the initial interpretation model is integrated by a large language model and a natural language processing model; The psychological state evaluation module is used to calculate the psychological state evaluation index according to the interpretation text, and select the initial interpretation model corresponding to the optimal psychological state evaluation index as the final interpretation model, wherein the psychological state evaluation index includes the BLEU score, the MSB score, the emotion accuracy rate of the emotion state label, the conformity score of the psychological evaluation scale, and the emotion deviation degree of the emotion state label; The psychological assessment result interpretation module is used to call the final interpretation model to interpret the psychological assessment scale results to be interpreted, and obtain corresponding interpretation results.

2. The intelligent interpretation system for psychological assessment scale results according to claim 1, characterized in that: Each psychological assessment data in the psychological assessment text data set also includes a pre-interpreted reference text, and the BLEU score is calculated based on the reference text and the interpreted text. The process of the psychological state assessment module calculating the BLEU score includes: Calculating the n-gram overlap probability between the reference text and the interpreted text as an interpretation accuracy score; Determine the geometric mean of the interpretation accuracy scores corresponding to each n-gram; The geometric mean is weighted according to a BP penalty coefficient to obtain a BLEU score, wherein the BP penalty coefficient is determined according to a size relationship between a text length of the reference text and a text length of the interpreted text.

3. The intelligent interpretation system for psychological assessment scale results according to claim 1, characterized in that: Each psychological assessment scale in the psychological assessment data is set with a different quantitative weight, and the interpretation text includes the scale assessment score of each psychological assessment scale. The process of calculating the MSB score by the psychological state assessment module includes: Determining the overall score of the psychological assessment data according to the quantitative weight and the scale assessment score; Determining the standardized score of the psychological assessment scale according to the mean and standard deviation of the scale assessment scores; Determine the time interval between each psychological assessment scale, and determine the score change value of the corresponding scale assessment score within the time interval; An MSB score is calculated based on the overall score, the standardized score, and the score change value.

4. The intelligent interpretation system for psychological assessment scale results according to claim 3, characterized in that: The calculating of the MSB score based on the overall score, the standardized score and the score change value comprises: The standardized scores of each psychological assessment scale are summed, and the control outliers of the psychological assessment data are calculated based on the summation results; Calculating the psychological state fluctuation value of the psychological evaluation data according to the score change value of the scale evaluation score; The overall score, the control abnormality value and the psychological state fluctuation value are weighted according to preset weights, and the weighted results are summed to obtain the MSB score.

5. The intelligent interpretation system for psychological assessment scale results according to claim 1, characterized in that: The interpretation text also includes the emotion interpretation label of the psychological assessment data and the corresponding emotion interpretation score. The process in which the psychological state evaluation module calculates the emotion accuracy rate of the emotion state label and the emotion deviation of the emotion state label includes: Determining the number of accurate interpretations of the emotional state label in the interpretation text, wherein the number of accurate interpretations is the number of times the emotional interpretation label completely matches the emotional state label corresponding to the psychological assessment data; The ratio of the number of accurate interpretations to the total number of psychological assessment data is used as the emotion accuracy rate of the emotion state label; The difference between the reference emotion score and the emotion interpretation score is used as the emotion deviation of the emotion state label, wherein the reference emotion score is the evaluation score for the emotion state label in the reference text corresponding to the psychological assessment data.

6. The intelligent interpretation system for psychological assessment scale results according to claim 1, characterized in that: The psychological status assessment module calculates the compliance score of the psychological assessment scale, including: Determining the number of times the interpretation content of the interpretation text for the psychological assessment scale is in compliance, the number of times the interpretation content is in compliance is the expected number of times the interpretation content of the interpretation text for the psychological assessment scale is in compliance with the psychological assessment content of a specific field; The ratio of the number of times the interpretation content is consistent to the total number of the psychological assessment content is used as the compliance score of the psychological assessment scale.

7. A method for intelligently interpreting the results of a psychological assessment scale, characterized in that: The method comprises: Collecting a psychological assessment text data set, wherein the psychological assessment text data set is classified according to psychological assessment topics, and each psychological assessment data in the psychological assessment text data set is annotated with an emotional state label; Training an initial interpretation model according to the psychological assessment text data set to obtain an interpretation text corresponding to each psychological assessment data on the psychological assessment topic, wherein the initial interpretation model is integrated by a large language model and a natural language processing model; Calculating a psychological state evaluation index according to the interpretation text, and selecting an initial interpretation model corresponding to the optimal psychological state evaluation index as the final interpretation model, wherein the psychological state evaluation index includes a BLEU score, an MSB score, an emotion accuracy rate of the emotion state label, a conformity score of the psychological evaluation scale, and an emotion deviation degree of the emotion state label; The final interpretation model is called to interpret the psychological assessment scale results to be interpreted, and the corresponding interpretation results are obtained.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the method for intelligently interpreting the results of a psychological assessment scale as described in claim 7 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for intelligently interpreting the results of a psychological assessment scale as described in claim 7 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for intelligently interpreting the results of a psychological assessment scale as described in claim 7 is implemented.

Images