Method, system and program product for personality assessment

Abstract

The present application provides a method, system and program product for personality assessment. The method includes receiving, by a processor, a record of a psychological counseling conversation between a visitor and a psychological counselor, and an assessment scale for personality assessment; using a trained first learning model, based on the record of the psychological counseling conversation, predicting the visitor's answers to each scale question in the assessment scale; outputting the visitor's personality assessment result based on the predicted answers to each scale question in the assessment scale. In the method of the present application, the visitor does not need to fill in the assessment scale, thus avoiding unfavorable factors such as time-consuming, subjective bias and social expectation bias, and at the same time, can make full use of professional and mature assessment scales to make the assessment efficient, objective, accurate and credible, and with the support of the record of the psychological counseling conversation, reasonable explanations and analyses related to the personality assessment results can also be obtained.

Description

Technical Field

[0001] The present application belongs to the field of data analysis and prediction, and in particular, relates to a method, system and program product for personality assessment. Background Art

[0002] With the rapid development of my country's social economy, people's life pressure and psychological problems are increasing day by day. Improving the level of prevention and treatment of common mental disorders and identification and intervention of psychological and behavioral problems has become an important goal of social development. Psychological counseling, as an important form of mental health service, is of great significance in helping people solve psychological problems and improve their mental health. In the psychological counseling scenario, understanding the personality traits of the client is crucial to improving the quality and effectiveness of counseling services. At present, the assessment of personality traits mainly relies on asking the client to complete a self-assessment questionnaire. The self-assessment questionnaire is an individual's answer to questions about his or her own thinking, feelings and behavior. Taking the Big Five personality as an example, the Big Five Personality Questionnaire (BFI) is currently the most widely used personality assessment method in the industry with stable reliability and validity. It can measure the individual's personality traits in five aspects: openness, conscientiousness, extroversion, agreeableness and neuroticism. In general, the client will fill in the self-assessment questionnaire / scale before the start of the consultation, but the self-assessment scale has problems such as time-consuming, subjective bias, and social expectation bias, which may have an adverse effect on subsequent psychological counseling.

[0003] In addition, the personality characteristics / traits of the visitor can also be evaluated through peer-evaluation questionnaires or third-party peer-evaluation methods. Peer-evaluation questionnaires are used by psychologists or trained observers to evaluate the personality of an individual based on his or her behavior in specific situations. They can achieve independent personality assessment without the participation of the visitor; third-party peer-evaluation methods refer to hiring professional psychological counselors or trained psychological measurement personnel to complete the scale on behalf of the visitor based on the conversation records between the visitor and the counselor, and the visitor's words, deeds and other characteristics during the consultation process to achieve personality assessment. The above methods do not require the visitor to fill in the scale and can meet the relatively objective evaluation needs, but the manual evaluation process increases the time and economic cost of the evaluation, and the real-time and efficiency are not high. Some studies also use natural language processing and other technologies to perform language analysis, learn and extract language features such as keywords and sentence patterns from psychological counseling conversation records or related data, and directly use machine learning models to predict the personality characteristics of the visitor. However, this type of method relies on the construction of the model, and the accuracy, reliability and professionalism of the generated evaluation results are not convincing.

[0004] It can be seen from this that there is a need for a personality assessment method that can maintain a considerable level of professionalism without the need for the client himself or others to help answer the personality assessment scale, and is more objective, efficient and accurate, so as to help improve the quality, effectiveness and efficiency of psychological counseling services. Summary of the invention

[0005] In view of the above problems, the present application is proposed to solve the above problems existing in the prior art.

[0006] The purpose of this application is to provide a method, system and program product for personality assessment, which can automatically and efficiently conduct personality assessment on visitors, and the assessment results are more professional, objective, accurate and reliable.

[0007] According to the first scheme of the present application, a method for personality assessment is provided, comprising a processor: receiving a record of a psychological counseling conversation between a visitor and a psychological counselor, and an assessment scale for personality assessment; using a trained first learning model to predict the visitor's answers to each scale item in the assessment scale based on the record of the psychological counseling conversation; and outputting the visitor's personality assessment results based on the predicted answers to each scale item in the assessment scale.

[0008] According to a second scheme of the present application, a system for personality assessment is provided, the system comprising an interface and a processor, wherein the interface is configured to obtain a record of a psychological counseling conversation between a visitor and a psychological counselor, and an assessment scale for performing a personality assessment; and the processor is configured to execute the steps of the method for personality assessment described in the various embodiments of the present application.

[0009] According to a third aspect of the present application, a computer program product is provided, comprising a computer program / instruction, wherein when the computer program / instruction is executed by a processor, the steps of the method for personality assessment as described in the various embodiments of the present application are implemented.

[0010] According to the method, system and program product for personality assessment in each embodiment of the present application, by training the first learning model, it can automatically predict the client's answers to each scale question in the personality assessment scale based on the psychological counseling conversation record. Therefore, it is possible to give full play to the stability of the professional personality assessment scale widely used in the industry in terms of reliability and validity without requiring the client or other personnel to fill out the assessment scale, thereby avoiding the adverse effects of time-consuming, low efficiency, social expectation bias, cognitive bias, etc., thereby achieving an efficient, objective, accurate and professional and credible assessment of the client's personality. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In the drawings, which are not necessarily drawn to scale, the same reference numerals may describe similar parts in different views. The same reference numerals with letter suffixes or different letter suffixes may represent different instances of similar parts. The accompanying drawings generally illustrate various embodiments by way of example and not limitation, and together with the specification and claims, are used to illustrate the disclosed embodiments. When appropriate, the same reference numerals are used throughout the drawings to refer to the same or similar parts. Such embodiments are illustrative and are not intended to be exhaustive or exclusive embodiments of the present apparatus or method.

[0012] Figure 1 A schematic flow chart of a method for personality assessment according to an embodiment of the present application is shown.

[0013] Figure 2 Another flowchart of a method for personality assessment according to an embodiment of the present application is shown.

[0014] FIG3( a ) is a schematic diagram showing scale questions in a personality assessment scale according to an embodiment of the present application.

[0015] FIG3( b ) is a schematic diagram showing a scoring method of a personality assessment scale according to an embodiment of the present application.

[0016] FIG3( c ) shows an explanation of personality traits corresponding to different scores of a personality assessment scale according to an embodiment of the present application.

[0017] Figure 4 Another flowchart of a method for personality assessment according to an embodiment of the present application is shown.

[0018] Figure 5 A schematic diagram showing part of the components of a system for personality assessment according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand the technical solution of the present application, the present application is described in detail below in conjunction with the accompanying drawings and specific implementation methods. The embodiments of the present application are further described in detail below in conjunction with the accompanying drawings and specific implementation methods, but are not intended to limit the present application.

[0020] The words "first", "second" and similar terms used in this application do not indicate any order, quantity or importance, but are only used to distinguish different parts. The expressions "first" and "second" are only used to distinguish by numbering for the convenience of expression, and are not intended to imply that the "first component" and the "second component" must have different physical properties. In fact, the "first component" and the "second component" may have the same or different structures, and are not limited here, as long as the "first component" and the "second component" are discrete components. Furthermore, if the context provides sufficient explanation, the "first component" and the "second component" may not even be discrete components, but may be integrated into the same component, or may be interchangeable with each other.

[0021] In the present application, when a specific device is described as being located between a first device and a second device, there may or may not be an intermediate device between the specific device and the first device or the second device. When a specific device is described as being connected to other devices, the specific device may be directly connected to the other device without an intermediate device, or may not be directly connected to the other device but have an intermediate device.

[0022] "Include" or "comprising" and other similar words mean that the elements before the word include the elements listed after the word, and do not exclude the possibility of including other elements. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0023] It should also be understood that the term "and / or" in this application is only a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this application generally indicates that the associated objects before and after are in an "or" relationship.

[0024] Figure 1 A schematic flow chart of a method for personality assessment according to an embodiment of the present application is shown.

[0025] like Figure 1 As shown, the method for personality assessment according to the embodiment of the present application can be performed according to steps 101 to 103.

[0026] First, in step 101, a processor may receive a record of a psychological counseling conversation between a visitor and a psychological counselor, and an assessment scale for personality assessment.

[0027] FIG3 (a) shows a schematic diagram of the scale questions in a personality assessment scale according to an embodiment of the present application. In some embodiments, the personality assessment scale can be, for example, the Big Five Personality Questionnaire that defines five personality traits of openness, conscientiousness, extroversion, agreeableness, and neuroticism as shown in FIG3 (a). In other embodiments, the assessment scale can also be, for example, the Eysenck Personality Questionnaire, the Minnesota Multiphasic Personality Inventory, the Cattell 16 Personality Test Scale, etc. Different assessment scales can assess personality characteristics / traits from different angles, and the present application does not limit the type of assessment scale used for personality assessment, as long as it is recognized by the psychological counseling industry, has stable reliability and validity, and is a personality assessment method that is more commonly used in the psychological counseling process.

[0028] In order to facilitate the application to the first learning model in the present application, the above-mentioned psychological counseling dialogue record can be in text form, for example. In some embodiments, for example, it can be converted from the text of text psychological counseling such as online chat or dialogue between the counselor and the psychological counselor through channels such as the Internet, or it can also be converted from the audio record of the visitor's psychological counseling process or the video record of the psychological counseling process. In other embodiments, the psychological counseling dialogue record may not be limited to a specific form or source, for example, it is also based on including the above-mentioned multiple sources, or even converted and synthesized by multiple consultation records, and the present application does not make specific restrictions on this.

[0029] In step 102, the processor uses the trained first learning model to predict the visitor's answers to each scale question in the assessment scale based on the psychological counseling conversation record.

[0030] Take the Big Five Personality Questionnaire shown in Figure 3 (a) as an example. In this questionnaire, all 60 scale questions are in the form of multiple-choice questions. The choice here means "select one from a limited number of options (here five)". The options corresponding to each scale question are "1-strongly disagree; 2-disagree; 3-neutral; 4-somewhat agree; 5-strongly agree". When filling in the form, the words selected from 1 to 5 are used as the answers to the scale questions. In other types of assessment scales, there may be different ways to answer scale questions, and this application does not limit this.

[0031] In the embodiments of the present application, the visitor does not need to fill out the assessment scale in person, nor does the psychologist or other third party providing psychological counseling services to the visitor need to fill out the assessment scale manually. Instead, the first learning model can be pre-trained to simulate the visitor's personality characteristics, so that the trained first learning model can realistically predict the most likely answer or option that the visitor will give when facing each scale question in the assessment scale based on the psychological counseling conversation record. The present application does not impose any specific restrictions on the implementation of the first learning model, as long as it is trained to enable it to give answers to the assessment scale questions with a higher accuracy instead of the visitor based on the psychological counseling conversation record.

[0032] In step 103, the processor outputs the personality assessment result of the visitor based on the predicted answers to the various scale questions in the assessment scale. That is, when the trained first learning model completes the predicted answers to the various scale questions in the assessment scale, the scores of the visitor in various personality characteristics / traits / features can be further calculated and summarized based on the scoring or other methods defined in the assessment scale, and the scores of various aspects or the judgment conclusions based on the scores of various aspects are used as the assessment results and output. In other embodiments, personality assessment based on the answers to the scale questions can also adopt other calculation or judgment methods, and the specific methods are not listed one by one in this application.

[0033] Fig. 3 (b) shows a schematic diagram of a scoring method of a personality assessment scale according to an embodiment of the present application. Fig. 3 (c) shows an explanation of personality traits corresponding to different scores of a personality assessment scale according to an embodiment of the present application.

[0034] Figure 3 (b) still takes the Big Five Personality Questionnaire as an example, and provides detailed scoring methods for the Big Five personality dimensions and the Big Five personality sub-dimensions. Therefore, the processor can, for example, score separately according to the predicted options of each scale question in the evaluation scale, and finally provide the scores of each Big Five personality dimension and the Big Five personality sub-dimension as the evaluation result of the visitor's personality traits.

[0035] In other embodiments, the explanation of personality traits corresponding to different scores in FIG. 3 (c) may be provided to the client and / or the psychological counselor as part of the assessment results for reference based on the client's scores on each dimension of the Big Five personality.

[0036] According to the method for personality assessment in each embodiment of the present application, by training the first learning model, it can automatically predict the visitor's answers to each scale question in the personality assessment scale based on the psychological counseling conversation record, thereby, the visitor can be unaware and thus have a lighter psychological burden, while also avoiding the social expectation bias, self-perception bias and other problems that may be introduced when actively filling out the scale. The method of the present application also does not require other personnel to fill out additional forms, thereby avoiding the adverse effects of long time consumption and low efficiency. At the same time, it can also make full use of the stability of professional personality assessment scales widely used in the industry in terms of reliability and validity, thereby achieving an efficient, objective, accurate and professional and credible assessment of the visitor's personality.

[0037] Figure 2 Another flowchart of a method for personality assessment according to an embodiment of the present application is shown.

[0038] like Figure 2 As shown, step 201 and Figure 1 Similar to step 101 in , the psychological counseling dialogue record text between the visitor and the psychological counselor and the assessment scale for personality assessment are received.

[0039] In step 202, the trained first learning model is used to predict the client's answers to each scale item in the assessment scale based on the psychological counseling conversation record, as well as the explanatory information associated with the answer to each scale item as the context information of the corresponding scale item.

[0040] In step 203, the personality of the visitor is evaluated based on the predicted answers to the various scale items in the evaluation scale, and the visitor's personality evaluation result, the predicted answers to the various scale items, and the context information of the corresponding scale items are output together.

[0041] According to the method of the embodiment of the present application, since the first learning model predicts the answer to each scale question based on the psychological counseling dialogue record, in this process, the first learning model can collect and summarize the relevant evidence for its prediction and corresponding answer (such as the options of each scale question in the selected Big Five Personality Questionnaire), thereby generating explanatory information associated with the answer to each scale question, wherein such explanatory information can be the original text of the content explained by the visitor in the psychological counseling, or it can be the first learning model that combines information from multiple sources and regenerates the descriptive information about why the answer option is selected, and the causal reasoning process of selecting the option based on the dialogue content. In addition, the first learning model can not only extract the corresponding event information from the dialogue record, but also capture the visitor's multi-dimensional information such as language style, speaking style, dialogue context, emotional expression, etc., and give an explanation of the correlation between such information and the visitor's personality dimension, further improving the accuracy and reliability of the prediction. Therefore, compared with the technical solution of directly using a specific model to predict the personality assessment results of the client without using a personality assessment scale, the method of the embodiment of the present application not only makes full use of the professionalism of the mature assessment scale, but also can provide more fine-grained, more detailed and convincing context information of each scale question, thereby making the assessment results more interpretable, thereby helping to establish a more effective and trustworthy communication channel between the psychological counselor and the client, which is conducive to improving the efficiency and service quality of psychological counseling, and can provide more comprehensive and personalized services for the counselor. Furthermore, it can also provide reference direction guidance for formulating the client's follow-up consultation or diagnosis and treatment plan.

[0042] Figure 4 Another flowchart of a method for personality assessment according to an embodiment of the present application is shown. Figure 4 An example of applying the method of the present application to a psychological consultation process in real time is shown.

[0043] like Figure 4 As shown, first, in step 401, the processor dynamically receives the psychological counseling dialogue record between the visitor and the psychological counselor during the psychological counseling process.

[0044] In step 402, upon receiving an updated record of the psychological counseling conversation between the client and the psychological counselor, the processor uses the trained first learning model to predict the client's updated answers to each scale item in the assessment scale based on the updated record of the psychological counseling conversation.

[0045] In step 403, the processor outputs the personality assessment result of the visitor based on the updated answer of each scale question in the predicted assessment scale.

[0046] In step 404, when the answers to the questionnaire questions and / or the personality assessment results of the client being assessed change, the processor identifies the changed content and presents it to the psychological counselor.

[0047] As mentioned above, since the method according to the embodiment of the present application can use the trained first learning model to perform fully automatic and highly real-time predictions on the answers to the various scale questions in the evaluation scale, it is fully applicable to the scenario of real-time application of psychological counseling. In the practice process of proposing this application, a large number of experimental verifications have shown that the richness of the psychological counseling conversation records and the prediction accuracy of the answers to the scale questions, as well as the accuracy of the final personality assessment results, show a positive correlation trend within a certain range. In addition, when a specific scale question fails to obtain sufficient support from the conversation records, its prediction result may fluctuate due to failure to converge. Therefore, Figure 4 The embodiment in the embodiment identifies the changes in the answers to the scale questions and / or the personality of the client being assessed, and presents them to the psychological counselor in real time, which can provide a reference for the psychological counselor, for example, to further determine whether it is necessary to have a more in-depth conversation with the client in terms of certain characteristics, etc. The specific application of the identified changes is not described in detail in this application, and can be applied as needed.

[0048] In the method for personality assessment according to the embodiment of the present application, the first learning model for predicting the answers to the scale questions can be constructed based on a probability model, for example. In some embodiments, its specific implementation method can, for example, select a data-driven sequence (sequence-to-sequence, seq2seq) model, and its symbolic representation is shown in the following formula (1).

[0049] (1)

[0050] Among them, Model represents the first learning model built based on the probability model. BFI question is any scale item in the assessment scale. Dialogue It indicates the record of the psychological consultation dialogue between the client and the counselor. BFI choice It is the visitor's answer to the scale questions predicted by the first learning model.

[0051] Just as an example, when the infrastructure of the first learning model is a sequence model, it can first perform basic training / pre-training as follows: first, on a large amount of general text data, the probability distribution of the following word in the vocabulary is predicted through the provided context, and then the cross entropy loss is calculated with the real probability distribution, and the model parameters are updated through the back propagation algorithm. Here, it is not limited to specific models such as Transformer and RNN, and a suitable sequence model architecture can be selected according to the actual situation, and this application does not limit this.

[0052] After the first learning model is constructed and pre-trained, in order to make it more suitable for psychological counseling application scenarios, psychological counseling corpus, including psychological counseling dialogue data, various scale questions, etc., can be further used for supplementary training to make the word probability generated by it more consistent with the probability distribution in the context of psychological dialogue, so as to make up for its inherent deficiencies in capturing and expressing relevant semantic information of psychological state, thereby obtaining the first learning model after scene adaptation, improving the validity and efficiency of the first learning model in personality prediction tasks and related context generation in psychological counseling scenarios.

[0053] In some embodiments, in order to make the psychological counseling dialogue record meet the input information requirements of the first learning model, the processor may also pre-process the psychological counseling dialogue record. The specific pre-processing steps may include, for example:

[0054] Step S1, data cleaning: removing special symbols, punctuation marks and information irrelevant to psychological counseling in the psychological counseling dialogue record;

[0055] Step S2, generating a conversation history: converting the psychological counseling conversation record after word segmentation into a multi-round conversation format through a given template so as to meet the format requirements of the input information of the first learning model;

[0056] Step S3, word segmentation: segment the psychological counseling conversation record into word units consisting of one or more characters, and then map the word units to numerical values ​​through a word table.

[0057] In some embodiments, each scale question in the evaluation scale can be template matched according to {question content}+{options provided by the scale}+{task instructions}, where the question content is the specific content of the scale question, the options are the optional answer items corresponding to the scale question, and the task instructions are the type of prediction task.

[0058] Before inputting into the first learning model, the scale questions after template matching can also be spliced ​​with the complete pre-processed psychological counseling dialogue record, and then the spliced ​​scale questions and psychological counseling dialogue record are used as the input of the trained first learning model. An exemplary mode of input information and output information of the first learning model is given as follows:

[0059] / * Input information * /

[0060] Conversation history:

[0061] Consultant: {Chat record 1}

[0062] Visitor: {Chat record 2} ...

[0064] Consultant: {Chat record n-1}

[0065] Visitor: {chat record n}

[0066] Scale title: {title content}+{options provided by the scale}+{task instructions}

[0067] / *Output information* /

[0068] Selection result: {the scale option predicted by the first learning model}

[0069] The conversation history is the conversation history in the form of a multi-round conversation generated after the psychological counseling conversation record is processed in the preprocessing step S2; the topic content is the specific content of the scale topic, such as "I am an extrovert and sociable person" in the Big Five Personality Questionnaire; and in the Big Five Personality Questionnaire, for each scale topic, the options provided by the scale are "1. Strongly disagree", "2. Disagree", "3. Neutral", "4. Somewhat agree", "5. Strongly agree"; the type of prediction task given by the task instruction can at least include only predicting the answer to the scale topic and predicting the answer to the scale topic and the explanation information of the corresponding scale topic. For the Big Five Personality Questionnaire, it can be transformed into: "Select the one that best matches yourself" and "Select the one that best matches yourself and explain it". This input mode can provide effective prediction conditions for the first learning model as a probability model, thereby realizing the prediction of the visitor's personality.

[0070] Still taking the example of the Big Five Personality Questionnaire where each scale question corresponds to multiple options, the prediction process of the first learning model can be expressed by the following formula (2).

[0071] (2)

[0072] in, BFI questionIt is the scale title. history is the conversation history in a multi-round conversation format generated in the aforementioned step S2, answer is the visitor’s predicted option. The original output of the first learning model is answer The probability distribution of the word units contained in the entire vocabulary can be further used using methods such as greedy search or beam search to select the word unit with the largest probability as the prediction result from all possible answers.

[0073] It is worth noting that when the task instruction is to predict only the answers to the scale questions, the output information of the first learning model is the predicted results of the visitor's possible answers to each scale question in the assessment scale. Taking the Big Five Personality Questionnaire as an example, the output information is the predicted scale options; when the task instruction is to predict the answers to the scale questions and the explanation information of the corresponding scale questions, after the first learning model outputs the predicted results of the scale questions, it further automatically splices the predicted results to the scale questions (that is, BFI question ) After that, the first learning model is input again, and the explanation of the option content is generated in an autoregressive manner. In other words, when generating the explanation information of the answer to the scale question, it is also a fully automatic mode without human intervention, which provides the greatest possible convenience for users such as consultants, and the consulting service is more efficient.

[0074] According to the above formula (1) and formula (2), after traversing all the scale questions in the assessment scale and obtaining the client's answers to each scale question, the client's personality assessment result can be further obtained according to the calculation method specified in the corresponding assessment scale. Taking the Big Five Personality Questionnaire as an example again, the weighted average of the scores of the predicted options of each scale question in the five personality traits of openness, conscientiousness, extraversion, agreeableness and neuroticism defined in the Big Five scale can be obtained to obtain the client's score on each personality trait. This process is shown in the following formula (3).

[0075] (3)

[0076] in, score trait is the corresponding score on a certain personality trait. items trait is the number of scale questions corresponding to this personality trait, response i Is the client's first i In this way, a quantitative measure can be obtained, which reflects the visitor's performance on each personality trait. For other types of personality assessment scales, there may be other personality trait / trait score calculation methods, which can be calculated according to the rules, and this application does not limit this.

[0077] According to formula (1), formula (2) and formula (3), it can be understood that the validity of the personality assessment method according to the embodiment of the present application is consistent with the probability of the first learning model as a probability model modeling the psychological counseling dialogue. In other words, if the first learning model can accurately predict the option , the evaluation result of personality characteristics / traits will be more accurate; on the contrary, if the probability distribution calculated by the first learning model is wrong, the predicted option for the specific scale question will also be wrong, so that the personality characteristics / trait score calculated by formula (3) will also lose accuracy and become unreliable. In addition, for different evaluation scales, the probability model expected to be constructed is also different. Therefore, in other embodiments, the method of the present application can also use the alignment training data set associated with the corresponding evaluation scale for specific personality evaluation to perform further scale alignment training on the first learning model after scene adaptation. That is to say, for different personality evaluation scales, the alignment training of the corresponding evaluation scales should be carried out in a targeted manner on the basis of the first learning model after scene adaptation obtained previously, so that when the score of personality characteristics / traits is calculated using formula (3), each scale question has as accurate a prediction value as possible, so that the final evaluation result is more accurate and reliable.

[0078] In other embodiments, an alignment training data set associated with a specific corresponding assessment scale for personality assessment may be used to further perform scale alignment training on the first learning model after scene adaptation. That is to say, for different personality assessment scales, alignment training of the corresponding assessment scales should be performed in a targeted manner based on the first learning model after scene adaptation obtained previously.

[0079] In some embodiments, an aligned training data set associated with a corresponding assessment scale is obtained by randomly sampling in a first conversation record data set using a first learning model that has not been trained for scale alignment, and samples in which the first learning model that has not been trained for scale alignment predicts the visitor's personality assessment results correctly can be used as positive samples in the aligned training data set, and samples in which the first learning model that has not been trained for scale alignment predicts the visitor's personality assessment results incorrectly can be used as negative samples in the aligned training data set.

[0080] In some embodiments, the first conversation record data set includes a plurality of conversation records related to psychological counseling, wherein each conversation record is labeled with the true personality of the counselor obtained by evaluating according to the corresponding evaluation scale as a true value. More specifically, the conversation record data with true value annotations can be, for example, calculated by the counselor filling out a scale on behalf of the counselor after the counseling conversation, or can be obtained by asking the counselor to fill out an evaluation scale at the same time as the counseling conversation. In other embodiments, such annotated data can also be obtained based on the counseling records of counselors known to have specific personalities. It is worth noting that since psychological counseling usually needs to be conducted continuously for multiple times, such conversation record data with true value annotations are relatively large in number and easy to obtain.

[0081] Furthermore, based on obtaining the positive samples and the negative samples, a DPO (Direct Preference Optimization) strategy may be used to perform scale alignment training.

[0082] Compared with the traditional RLHF (reinforcement learning from human feedback) algorithm, the DPO algorithm no longer needs to train a ranker or reward model, but directly provides positive and negative samples to the model, which are calculated by the difference in entropy between the model to be trained and the reference model. The DPO algorithm aims to directly optimize the model's preferences without explicit reward modeling. The key to the DPO algorithm is to map the probability distribution of human preferences to the text generated by the model to achieve direct optimization of the optimal strategy. We derive the objective function of DPO from the KL (Kullback-Leibler) constrained reward maximization objective of RLHF. The KL constrained reward maximization objective is defined as follows (4-1):

[0083] (4-1)

[0084] in, is the model’s policy, representing the probability distribution of generating output y given input x, is a reward model that evaluates the reward value of the output given an input x and an output y. is the reference model, which represents the probability distribution of the reference strategy generating output y given input x, is a parameter that controls the deviation from the reference strategy and is used to adjust the degree of difference between the current model and the reference model. is the data distribution, indicating the input data The distribution of The KL divergence measures the difference between two probability distributions. The optimization objective is achieved by maximizing the reward and minimizing the KL divergence (Kullback-LeiblerDivergence) with the reference model.

[0085] To simplify the above optimization problem, DPO reparameterizes the reward function and maps it to the optimal strategy. , the optimal strategy It can be expressed as the following formula (4-2).

[0086] (4-2)

[0087] in, is a normalizing constant that ensures that the probability distribution sums to 1.

[0088] By taking the logarithm of the above equation (4-2), we can reparameterize the reward function into equation (4-3).

[0089] (4-3)

[0090] In DPO, we omit the explicit reward model training and directly optimize the policy model , and its objective function is formula (4-4).

[0091] (4-4)

[0092] in, is the objective function of DPO, which means optimizing the logarithmic difference between positive and negative samples. σ is the sigmoid function, which is used to map the logarithmic difference to the probability value. y w and y t They represent positive samples and negative samples respectively, that is, the correct prediction results and incorrect prediction results of the model.

[0093] The DPO strategy is used to optimize the first learning model. The process is simple and there is no need to explicitly train the reward model. The strategy is directly optimized, which reduces the computational complexity and instability, avoids the high variance and instability in the traditional machine learning process, and improves the training efficiency and model validity.

[0094] In addition, the above positive and negative sample sampling strategies described in the embodiment of the present application are sampling strategies that are closer to the real distribution than other sampling methods: in the traditional DPO algorithm, the results directly obtained from the probability P generated by the model are usually used as negative samples, and the text obtained by rewriting the negative samples by humans or strong models is used as positive samples. This sampling method causes the problem that the positive and negative samples do not obey the same distribution P, which makes the result of calculating the KL divergence by formula (4-1) inaccurate. In the present application, both positive and negative samples are sampled from the same P, and then compared with the known visitor evaluation scale results. The generated results consistent with the visitor's real results are used as positive samples, and the inconsistent generated results are used as negative samples, so that the positive and negative samples obey the same distribution, the calculated KL divergence is more accurate, and the final optimized model effect is better. Therefore, the embodiment of the present application can make the first learning model better aligned with the corresponding evaluation scale through the application of the DPO algorithm, which not only makes the predicted personality more accurate, but also makes the context of the explanatory information provided by it more accurate and the reasoning complete and reasonable, which improves the validity and efficiency of the personality prediction task in the psychological counseling dialogue.

[0095] In some embodiments, when it is necessary to use multiple different assessment scales to conduct personality assessment on the visitor, the processor may further perform the following steps.

[0096] First, the first learning model that has completed scale alignment training for the corresponding assessment scale can be used to predict the client's answers to each scale question in each assessment scale based on the psychological counseling conversation record. Then, based on the predicted answers to each scale question in each assessment scale, the personality assessment results of the client calculated according to each assessment scale are output and presented to the psychological counselor, so that the psychological counselor can conduct multiple assessments and comprehensive analysis of the client from different angles and / or different dimensions.

[0097] According to an embodiment of the present application, a system for personality assessment is also provided. Figure 5 A schematic diagram showing part of the components of a system for personality assessment according to an embodiment of the present application is shown.

[0098] like Figure 5 As shown, the system 500 at least includes an interface 501 and a processor 502. The interface 501 can be configured to obtain, for example, a record of a psychological consultation conversation between a visitor and a psychological counselor, and an assessment scale for personality assessment.

[0099] In some embodiments, the interface 501 can be configured to obtain a long text in the source language of the source language to be generated by the user input. In some embodiments, the interface 501 may include a network adapter, a cable connector, a serial connector, a USB connector, a parallel connector, a high-speed data transmission adapter (such as optical fiber, USB 3.0, Thunderbolt interface, etc.), a wireless network adapter (such as a WiFi adapter), a telecommunications (3G, 4G / LTE, etc.) adapter, etc., and this application does not limit this. The system 500 can transmit the acquired psychological counseling dialogue record between the visitor and the psychological counselor to other parts such as the processor 502 through the interface 501. In some embodiments, the interface 501 can also receive, for example, a trained first learning model from a learning network training device (not shown), which is not listed one by one in this application.

[0100] The processor 502 may be configured to execute the steps of the method for personality assessment according to various embodiments of the present application. As an example only, the method may include: receiving a record of a psychological counseling conversation between a visitor and a psychological counselor, and an assessment scale for personality assessment; using a trained first learning model, based on the record of the psychological counseling conversation, predicting the visitor's answers to each scale question in the assessment scale; and outputting the visitor's personality assessment result based on the predicted answers to each scale question in the assessment scale.

[0101] In some embodiments, the processor 502 may be a processing device including one or more general-purpose processing devices, such as a microprocessor, a central processing unit (CPU), a graphics processing unit (GPU), etc. More specifically, the processor may be a complex instruction set computing (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, a very long instruction word (VLIW) microprocessor, a processor running other instruction sets, or a processor running a combination of instruction sets. The processor may also be one or more special-purpose processing devices, such as an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), a digital signal processor (DSP), a system on a chip (SoC), etc.

[0102] In some embodiments, the system 500 may also include a display (not shown), which may be configured to, for example, present the personality assessment results of the visitor calculated according to each assessment scale to the psychological counselor when there are multiple assessment scales; and present the personality assessment results with a mark of the changed content to the psychological counselor when the answers to the scale questions and / or the personality assessment results of the visitor change, etc., which are not listed here one by one. In some embodiments, the display may include a liquid crystal display (LCD), a light emitting diode display (LED), a plasma display, or any other type of display, and provide a graphical user interface (GUI) presented on the display for user input and image / data display, etc.

[0103] According to an embodiment of the present application, a computer program product is also provided, including a computer program / instruction, and when the computer program / instruction is executed by a processor, the steps of the method for personality assessment described in each embodiment of the present application are implemented.

[0104] Those skilled in the art can understand that all or part of the processes in the method for personality assessment in each embodiment of the present application can be achieved by driving the relevant hardware through a computer program / instruction. In addition, the computer program / instruction can also be stored in a non-volatile computer-readable storage medium. When the computer program / instruction stored in the medium is executed by a processor or other hardware, the process of the method for personality assessment in each embodiment of the present application can be achieved.

[0105] In addition, according to an embodiment of the present application, a non-transitory computer-readable medium is provided, on which instructions are stored. When executed by a processor, the instructions execute the steps of the method for personality assessment as described in the various embodiments of the present application.

[0106] In some embodiments, the above-mentioned non-transitory computer-readable medium can be a read-only memory (ROM), random access memory (RAM), phase change random access memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), electrically erasable programmable read-only memory (EEPROM), other types of random access memory (RAM), flash disk or other forms of flash memory, cache, register, static memory, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, cassette tape or other magnetic storage device, or any other possible non-transitory medium used to store information or instructions that can be accessed by a computer device.

[0107] According to the method, system and program product for personality assessment in each embodiment of the present application, by training the first learning model, it can automatically predict the visitor's answers to each scale question in the personality assessment scale based on the psychological counseling conversation record, thereby, the visitor can be unaware and thus have a lighter psychological burden, and at the same time, it can also avoid the social expectation bias, self-cognition bias and other problems that may be introduced when actively filling out the scale. In addition, there is no need for other personnel to fill out the form additionally, thereby avoiding the adverse effects of long time consumption and low efficiency. At the same time, it can also make full use of the stability of the professional personality assessment scale widely used in the industry in terms of reliability and validity, thereby achieving an efficient, objective, accurate and professional and credible assessment of the visitor's personality. Furthermore, since the psychological counseling conversation record is used as a condition, and the answer to each scale question can generate the corresponding option / answer, therefore, it is also possible to generate more fine-grained, more detailed and convincing context information of the selected option of each scale question associated with each scale question and personality assessment result, thereby making the assessment result more interpretable, thereby helping to establish a more effective and trustworthy good communication channel between the psychological counselor and the visitor, which is conducive to improving the efficiency and service quality of psychological counseling, and can provide the counselor with more comprehensive and personalized services. Under certain conditions, it can even provide reference direction guidance for formulating the follow-up consultation or diagnosis and treatment plan for the visitor. In addition, since the method in this application is fully automated and efficient, it can dynamically update the assessment results as the psychological counseling process proceeds, and mark the changes worthy of attention. Therefore, it can provide real-time support and reference for the psychological counselor during the psychological counseling process, so that it can provide better services for the visitor.

[0108] In addition, although exemplary embodiments have been described herein, the scope includes any and all embodiments based on the present disclosure with equivalent elements, modifications, omissions, combinations (e.g., various embodiments intersecting schemes), adaptations or changes. The elements in the claims are to be interpreted broadly based on the language adopted in the claims, and are not limited to the examples described in this specification or during the enforcement of this application, which examples are to be interpreted as non-exclusive. Therefore, this specification and examples are intended to be considered as examples only, and the true scope and spirit are indicated by the claims and the full scope of their equivalents.

[0109] The above description is intended to be illustrative rather than restrictive. For example, the above examples (or one or more schemes thereof) may be used in combination with each other. For example, a person of ordinary skill in the art may use other embodiments when reading the above description. In addition, in the above-mentioned specific embodiments, various features may be grouped together to simplify the present disclosure. This should not be interpreted as an intention that a disclosed feature that is not claimed for protection is necessary for any claim. On the contrary, the subject matter of the present invention may be less than all the features of a particular disclosed embodiment. Thus, the claims are incorporated herein into the specific embodiments as examples or embodiments, wherein each claim is independently a separate embodiment, and it is considered that these embodiments may be combined with each other in various combinations or arrangements. The scope of the present invention should be determined with reference to the claims and the full scope of equivalent forms granted by these claims.

Claims

1. A method for personality assessment, characterized in that This includes, by the processor: Receive assessment scales for personality assessment and dynamically receive records of psychological counseling conversations between clients and counselors during the psychological counseling process; Using the trained first learning model, based on the psychological counseling dialogue record, predicting the visitor's answer to each scale question in the assessment scale; Outputting the visitor's personality assessment result based on the predicted answers to each scale question in the assessment scale; Upon receiving updated records of the counseling conversation between the client and the counselor: Using the trained first learning model, based on the updated psychological counseling dialogue record, predict the visitor's updated answer to each scale question in the assessment scale; Outputting the visitor's personality assessment result based on the updated answers to the various scale questions in the predicted assessment scale; In the event that the responses to the questionnaire and / or the personality assessment results of the client change, the changes are marked and presented to the counselor; The first learning model is constructed based on a probability model and is obtained by performing scale alignment training using a psychological counseling corpus and an alignment training data set associated with a corresponding assessment scale; The alignment training data set associated with the corresponding assessment scale is obtained by randomly sampling in the first conversation record data set using the first learning model that has not been trained for scale alignment, wherein the samples in which the first learning model that has not been trained for scale alignment correctly predicts the visitor's personality assessment result are used as positive samples in the alignment training data set, and the samples in which the first learning model that has not been trained for scale alignment incorrectly predicts the visitor's personality assessment result are used as negative samples in the alignment training data set, and the scale alignment training is performed using the positive samples and the negative samples using a direct preference optimization strategy; and the first conversation record data set contains multiple conversation records related to psychological counseling, wherein each conversation record is labeled with the true personality of the counselor obtained by evaluating according to the corresponding assessment scale as a true value.

2. The method according to claim 1, characterized in that The step of using the trained first learning model to predict the visitor's answer to each scale question in the assessment scale based on the psychological counseling dialogue record further includes: Using the trained first learning model, based on the psychological counseling conversation record, predict the visitor's answer to each scale question in the assessment scale, and the explanation information associated with the answer to each scale question as the context information of the corresponding scale question; The step of outputting the visitor's personality assessment result based on the predicted answers to each scale question in the assessment scale further includes: The personality of the visitor is evaluated based on the predicted answers to the various scale questions in the evaluation scale, and the personality evaluation result of the visitor, the predicted answers to the various scale questions and the context information of the corresponding scale questions are output together.

3. The method according to claim 1, characterized in that The psychological counseling dialogue record is in text form, and is obtained by converting at least one of a text of a psychological counseling session, an audio recording of the psychological counseling session, and a video recording of the psychological counseling session.

4. The method according to claim 1, characterized in that: In the case where there are multiple assessment scales, the method further comprises: Using the first learning model that has completed the scale alignment training for the corresponding assessment scale, based on the psychological counseling conversation record, predict the visitor's answer to each scale question in each assessment scale; Based on the predicted answers to the questions in each assessment scale, the personality assessment results of the visitor calculated according to each assessment scale are output and presented to the psychological counselor.

5. The method according to claim 1, characterized in that The method further includes preprocessing the psychological counseling dialogue record by a processor so that the preprocessed psychological counseling dialogue record meets the input information requirement of the first learning model, specifically including the following steps: S1, data cleaning: removing special symbols, punctuation marks and information irrelevant to psychological counseling in the psychological counseling dialogue records; S2, generate dialogue history: convert the psychological counseling dialogue record after word segmentation into a multi-round dialogue format through a given template so as to meet the format requirements of the input information of the first learning model; S3, word segmentation: segment the psychological counseling conversation record into word units consisting of one or more characters, and then map the word units to numerical values ​​through a word table.

6. The method according to claim 1 or 2, characterized in that: Using the trained first learning model, based on the psychological counseling dialogue record, predicting the visitor's answer to each scale question in the assessment scale further includes: Template matching is performed on each scale question in the evaluation scale according to {question content}+{options provided by the scale}+{task instructions}, wherein the question content is the specific content of the scale question, the options are the optional answer items corresponding to the scale question, and the task instructions are the types of prediction tasks, wherein the types of prediction tasks at least include only predicting the answers to the scale questions and predicting the answers to the scale questions and the explanation information of the corresponding scale questions; After the template matching, each scale question is spliced ​​with the complete pre-processed psychological counseling dialogue record, and used as the input of the trained first learning model, so as to predict the visitor's answer to each scale question in the evaluation scale.

7. A system for personality assessment, characterized in that include: The interface is configured to: obtain an assessment scale for personality assessment and dynamically receive a record of a psychological counseling conversation between a visitor and a psychological counselor during a psychological counseling process; A processor configured to: execute the method for personality assessment according to any one of claims 1-6.

8. The system according to claim 7, characterized in that The system further comprises: A display configured to: when there are multiple assessment scales, present to the psychological counselor the personality assessment results of the visitor calculated according to each assessment scale; and In the event that the responses to the questionnaire and / or the client's personality assessment results change, the client's personality assessment results with an indication of the changes are presented to the counselor.

9. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method for personality assessment according to any one of claims 1 to 6 are implemented.

Images