Method and system for generating psychological intervention programs for minors

Abstract

The present invention relates to the field of natural language processing technology, and in particular to a method and system for generating a psychological intervention plan for minors, comprising: using a natural language processing model to classify and extract information of minors, receiving analysis requirements and extracting keywords from the classified information; creating decision trees for the classified information, setting a discrimination criterion for keywords, and adding the keywords to the decision tree according to the discrimination criterion; calculating the information entropy of the decision tree, and determining whether to rebuild the decision tree according to whether the information entropy is greater than a first threshold; calculating whether the correlation between decision results is greater than a second threshold, and if so, taking the decision result as an information processing result, otherwise rebuilding the decision tree; using the natural language processing model to process the information analysis result and outputting an intervention plan; analyzing the information of minors based on the natural language processing model and the decision tree, and then outputting an intervention plan to assist staff in conducting psychological intervention on minors.

Description

Technical Field

[0001] The present invention relates to the technical field of natural language processing, and in particular to a method and system for generating a psychological intervention program for minors. Background Art

[0002] Protecting the healthy development of minors is a consensus of all sectors of society. However, in recent years, more and more mental health problems have emerged among minors. There are many teenagers suffering from depression and autism due to mental health problems. Some even develop violent tendencies and become destabilizing factors in society. They are likely to hurt themselves and others when they grow up. Therefore, in the process of growth, minors need external mental health guidance, especially for those with criminal tendencies or even those who have a record of breaking the law. However, at this stage, the number of professionals in the field of minors' mental health is too small relative to the number of minors. Therefore, a method is needed to generate personalized psychological intervention plans for each minor to assist relevant staff to intervene in minors more professionally and effectively to maintain their mental health.

[0003] With the development and breakthroughs of natural language processing technology, there is a technical foundation for generating a psychological intervention plan for minors, thereby assisting relevant staff in providing effective psychological counseling for minors. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for generating a psychological intervention plan for minors in view of the deficiencies of the prior art, aiming to assist relevant staff in providing effective psychological counseling for minors.

[0005] The present invention achieves the above-mentioned purpose through the following technical solutions:

[0006] A method for generating a psychological intervention program for minors, comprising:

[0007] Information classification step: Use a natural language processing model to classify the information of minors and extract information on their growth experience, living environment, interpersonal relationships, language, behavior, and self-assessment;

[0008] Information processing step: receiving the analysis requirements submitted by the user, and using the natural language processing model to extract keywords from the classified information according to the analysis requirements;

[0009] Creating a decision tree step: creating a decision tree for each of the classified information, using a natural language processing model to set a discrimination criterion for the keyword, and adding the keyword as a child node to the decision tree according to the discrimination criterion;

[0010] Adjustment step: calculating the information entropy of the decision tree. If the information entropy is higher than the first threshold, executing the decision tree creation step again. If the information entropy is lower than the first threshold, calculating the probability of the category corresponding to each child node in the decision tree, comparing the category corresponding to the maximum probability to obtain the decision result.

[0011] Calculate the correlation between the decision result corresponding to each decision tree and the decision results corresponding to other decision trees. If the correlations are all greater than a second threshold, use the decision result as the information analysis result. Otherwise, execute the decision tree creation step again.

[0012] Output plan step: After processing the information analysis results using the natural language processing model, output the intervention plan.

[0013] Furthermore, the step of creating a decision tree includes:

[0014] Create root node sub-step: Create the root node of the decision tree according to the type of information;

[0015] Setting a discrimination criterion sub-step: inputting the keyword into the natural language processing model, and allowing the natural language processing model to set a discrimination criterion for the keyword according to the analysis requirements;

[0016] Add child node sub-step: determine whether the keyword meets the discrimination criteria, if so, add the keyword as a child node to a subtree of the decision tree, if not, add the keyword to another subtree of the decision tree.

[0017] Furthermore, the step of creating a decision tree includes:

[0018] The decision tree is a ternary tree;

[0019] Any one of the left subtree, the middle subtree and the right subtree of the decision tree is used as an irrelevant subtree, and the irrelevant subtree is used to add keywords that cannot be classified by the discrimination criteria, and the irrelevant subtree does not participate in the calculation of information entropy.

[0020] Furthermore, in the adjustment step, the information entropy is calculated using the following formula:

[0021]

[0022] Among them, H[X] represents information entropy, P(x) is the probability of the category corresponding to each child node, and A is the depth of the decision tree.

[0023] Furthermore, the calculation process of the correlation includes:

[0024] Performing feature extraction on the decision result to obtain a two-dimensional feature vector;

[0025] The Pearson correlation coefficient between the two-dimensional feature vector of each decision result and the two-dimensional feature vector of other decision results is calculated using the following formula:

[0026]

[0027] Where sim represents the Pearson correlation coefficient, x is the abscissa of the two-dimensional feature vector, and y is the ordinate of the two-dimensional feature vector. and are the average values ​​of the horizontal and vertical coordinates of the two-dimensional feature vectors of all decision results, i is the number of decision trees, and n is the number of decision trees.

[0028] Furthermore, the first threshold is set by the following formula:

[0029]

[0030] Where h[x] represents the first threshold, is the average probability of the category corresponding to each child node, and A is the depth of the decision tree.

[0031] Furthermore, the second threshold is negatively correlated with the number of decision trees and is set according to the following formula:

[0032]

[0033] Wherein, SIMTT is the second threshold, and B is the number of decision trees.

[0034] Furthermore, the output scheme step includes:

[0035] Preparation sub-step: select transformer as the natural language processing model, obtain successful cases of psychological intervention for minors as language templates, divide the language templates into a training set and a validation set, extract keywords from the training set and construct a training decision tree;

[0036] Model training sub-step: input the training decision tree into the natural language processing model for processing and then output the processing result, calculate the Pearson correlation coefficient between the processing result and the validation set, until the Pearson correlation coefficient is greater than 0.7, then determine that the training is completed;

[0037] Plan output sub-step: After the natural language processing model processes the information analysis results, it outputs the intervention plan.

[0038] A system for intelligently generating a psychological intervention plan for minors, applying the above-mentioned method for generating a psychological intervention plan for minors, comprises:

[0039] Information classification module: uses a natural language processing model to classify and extract information about minors to obtain information about their growth experience, living environment, interpersonal relationships, language, behavior, and self-assessment;

[0040] Information processing module: receives analysis requirements submitted by users and uses natural language processing models to extract keywords from classified information based on analysis requirements;

[0041] Creating a decision tree module: Creating decision trees for the classified information respectively, using a natural language processing model to set the discrimination criteria for the keywords, and adding the keywords as child nodes to the decision trees according to the discrimination criteria;

[0042] Adjustment module: calculate the information entropy of the decision tree. If the information entropy is higher than the first threshold, execute the decision tree creation module again. If the information entropy is lower than the first threshold, calculate the probability of the category corresponding to each sub-node in the decision tree, compare the category corresponding to the maximum probability and obtain the decision result.

[0043] Calculate the correlation between the decision result corresponding to each decision tree and the decision results corresponding to other decision trees. If the correlations are all greater than a second threshold, use the decision result as the information analysis result. Otherwise, execute the decision tree creation module again.

[0044] Output plan module: After processing the information analysis results using the natural language processing model, output the intervention plan.

[0045] Furthermore, the creation of a decision tree module includes:

[0046] Create root node submodule: Create the root node of the decision tree according to the type of information;

[0047] Setting a discrimination criterion submodule: inputting the keyword into the natural language processing model, and allowing the natural language processing model to set a discrimination criterion for the keyword according to the analysis requirements;

[0048] Add child node submodule: determine whether the keyword meets the discrimination criteria. If so, add the keyword as a child node to a subtree of the decision tree. If not, add the keyword to another subtree of the decision tree.

[0049] The beneficial effects of the present invention are as follows: by creating an analysis method based on a decision tree, the growth experience, living environment, interpersonal relationship, language, behavior, and self-evaluation of minors are analyzed to obtain the psychological analysis results of minors, and then the psychological analysis results are processed by a natural language processing model that learns language doubling, and a targeted intervention plan is obtained and output in the form of text suitable for communicating with minors, which effectively assists relevant staff in psychological intervention of minors; at the same time, three subtrees are set for the root node of the decision tree, one of which is used as an addition position for irrelevant nodes, which reduces the influence of interference information on the analysis results and makes the psychological analysis of minors more accurate; by setting an information entropy threshold, the decision tree is reconstructed when it is greater than a predetermined information entropy threshold, and the effect of adaptively creating a decision tree during the analysis process is achieved to improve the accuracy of the analysis results; by calculating the correlation between the analysis results of one type of information and the analysis results of other types of information, when the correlation is lower than the threshold, the analysis is re-performed to improve the accuracy of the analysis results; by using a natural language model to learn a language template, and then processing the psychological analysis results of minors, an intervention plan in the form of text suitable for communicating with minors is output, and the auxiliary effect of psychological intervention work on minors is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 The present invention is a flowchart of a method for generating a psychological intervention program for minors.

[0051] Figure 2 This is a schematic diagram of a system for intelligently generating psychological intervention plans for minors proposed by the present invention.

[0052] Figure 3 This is a screenshot of a self-esteem scale in the present invention.

[0053] Figure 4 This is a screenshot of the software interface of the method for generating a psychological intervention plan for minors implemented in the present invention. DETAILED DESCRIPTION

[0054] The present invention is described in detail below with reference to the accompanying drawings.

[0055] like Figure 1 The figure shows a flow chart of a method for generating a psychological intervention program for minors, including: information classification step, information processing step, decision tree creation step, adjustment step and program output step.

[0056] The present invention creates an analysis method based on a decision tree, analyzes the growth experience, living environment, interpersonal relationship, language, behavior, and self-evaluation of minors to obtain psychological analysis results of minors, and then processes the psychological analysis results through a natural language processing model that learns language doubling, obtains targeted intervention plans, and outputs them in the form of text suitable for communicating with minors, effectively assisting relevant staff in psychological intervention of minors; at the same time, three subtrees are set for the root node of the decision tree, one of which is used as an addition position for irrelevant nodes, thereby reducing the influence of interference information on the analysis results and making the psychological analysis of minors more accurate; by setting an information entropy threshold, the decision tree is rebuilt when it is greater than a predetermined information entropy threshold, thereby achieving the effect of adaptively creating a decision tree during the analysis process and thus improving the accuracy of the analysis results; by calculating the correlation between the analysis results of one type of information and the analysis results of other types of information, when the correlation is lower than the threshold, the analysis is re-performed, thereby improving the accuracy of the analysis results; by using a natural language model to learn a language template, and then processing the psychological analysis results of minors, outputting an intervention plan in the form of text suitable for communicating with minors, thereby improving the technical effect of assisting psychological intervention work for minors.

[0057] like Figure 4 As shown, this is a screenshot of the interface for analyzing the mental health of a minor with a criminal record during the process of generating a psychological intervention plan for the minor's mental health in this embodiment. The cause of the case indicates the minor's psychological problem. The "intervention method" in the lower left corner of the figure is the part for generating the intervention plan.

[0058] Information classification step: Use a natural language processing model to classify the information of minors and extract information on their growth experience, living environment, interpersonal relationships, language, behavior, and self-assessment;

[0059] In one embodiment, the information of the minor comes from his / her personal files, as well as self-esteem scales, mental health scales, stress test scales, self-assessment test scales filled out by the minor, as well as living environment surveys, text descriptions from relatives and friends, and records of his / her personal language content, such as Figure 3 As shown, it is a screenshot of a self-esteem scale, from which the information needed for analyzing the psychology of minors is classified and extracted through a natural language model. In this embodiment, a transformer model is used to extract and classify growth experience information, living environment information, interpersonal relationship information, language information, behavior information and self-assessment information. By using multi-angle information types, the results of the psychological analysis of minors are made more accurate.

[0060] Information processing step: receiving the analysis requirements submitted by the user, and using the natural language processing model to extract keywords from the classified information according to the analysis requirements;

[0061] In the specific implementation, the analysis needs include analysis of the emotional state, personality analysis, value analysis, world view analysis, and emotional intelligence analysis of minors. According to the specific situation, the corresponding analysis needs are selected. The natural language processing model first understands the analysis needs, and then extracts keywords that are helpful to the analysis needs from the classified information based on its understanding; thereby achieving the technical effect of extracting effective keywords from a large amount of information for specific analysis needs.

[0062] Creating a decision tree step: creating a decision tree for each of the classified information, using a natural language processing model to set a discrimination criterion for the keyword, and adding the keyword as a child node to the decision tree according to the discrimination criterion;

[0063] In one embodiment, the step of creating a decision tree includes:

[0064] Create root node sub-step: Create the root node of the decision tree according to the type of information;

[0065] Setting a discrimination criterion sub-step: inputting the keyword into the natural language processing model, and allowing the natural language processing model to set a discrimination criterion for the keyword according to the analysis requirements;

[0066] Add child node sub-step: determine whether the keyword meets the discrimination criteria, if so, add the keyword as a child node to a subtree of the decision tree, if not, add the keyword to another subtree of the decision tree.

[0067] Specifically, taking the analysis requirement "whether the minor has depression or depression tendency" as an example, it is classified as sentiment analysis, and the root node of the decision tree is created. The root node represents "the emotional state of the minor", and then a child node is created to represent "normal emotional state" and added to the root node, and a child node "abnormal emotional state" is created and added to the root node. Then, subnodes representing various abnormal emotional conditions are added to the subnode representing "abnormal emotional state" to form a subtree. The newly added child node on another node of normal emotional condition and the node together constitute another subtree. Finally, keywords are added to the corresponding subtree according to the judgment criteria. For example, the keyword "don't want to do anything" may be judged by the criteria that if this psychological state appears for more than one week continuously, it is judged as depression, otherwise it is judged as normal. By creating a decision tree, the keywords in each type of information are analyzed, and the analysis of the psychological state of minors is realized.

[0068] In another embodiment, the step of creating a decision tree includes:

[0069] The decision tree is a ternary tree;

[0070] Any one of the left subtree, the middle subtree and the right subtree of the decision tree is used as an irrelevant subtree, and the irrelevant subtree is used to add keywords that cannot be classified by the discrimination criteria, and the irrelevant subtree does not participate in the calculation of information entropy.

[0071] Specifically, when creating a decision tree, when determining which subtree a keyword should be added to, there are ambiguous or undecidable situations. At this time, it is added to an "irrelevant subtree", thereby improving the accuracy of the decision tree analysis.

[0072] Adjustment step: calculating the information entropy of the decision tree, if the information entropy is higher than the first threshold, executing the decision tree creation step again, if the information entropy is lower than the first threshold, calculating the probability of the category corresponding to each sub-node in the decision tree, comparing the category corresponding to the maximum probability and obtaining the decision result;

[0073] In one embodiment, in the adjustment step, the information entropy is calculated using the following formula:

[0074]

[0075] Among them, H[X] represents information entropy, P(x) is the probability of the category corresponding to each child node, and A is the depth of the decision tree.

[0076] Specifically, the depth of the decision tree is incorporated into the calculation process of information entropy, so that the calculation result of information entropy is more suitable for the corresponding decision tree, so as to improve the effectiveness of judging the accuracy of the analysis result of the decision tree through information entropy.

[0077] In one embodiment, the first threshold is set by the following formula:

[0078]

[0079] Where h[x] represents the first threshold, is the average probability of the category corresponding to each child node, and A is the depth of the decision tree.

[0080] Specifically, when the decision tree is too deep, that is, the number of layers of the decision tree is too many, the analysis of the information is more accurate, and the corresponding threshold is set lower; when the decision tree is too shallow, that is, the number of layers of the decision tree is too few, the analysis of the information is more accurate, and the corresponding threshold is set higher; by incorporating the depth of the decision tree into the setting process of the first threshold, the first threshold is more adapted to the corresponding decision tree, and the accuracy of the first threshold in determining whether the decision tree is accurate in the analysis is improved. When the information entropy is lower than the first threshold, the decision tree is valid, and the analysis result is calculated; when the information entropy is higher than the first threshold, it is determined that the decision tree processes the keywords too chaotically, and the decision tree is recreated for analysis, thereby improving the accuracy of the final analysis result.

[0081] Calculate the correlation between the decision result corresponding to each decision tree and the decision results corresponding to other decision trees. If the correlations are all greater than a second threshold, use the decision result as the information analysis result. Otherwise, execute the decision tree creation step again.

[0082] In one embodiment, the calculation process of the correlation includes:

[0083] Performing feature extraction on the decision result to obtain a two-dimensional feature vector;

[0084] The Pearson correlation coefficient between the two-dimensional feature vector of each decision result and the two-dimensional feature vector of other decision results is calculated using the following formula:

[0085]

[0086] Where sim represents the Pearson correlation coefficient, x is the abscissa of the two-dimensional feature vector, and y is the ordinate of the two-dimensional feature vector. and are the average values ​​of the horizontal and vertical coordinates of the two-dimensional feature vectors of all decision results, i is the number of decision trees, and n is the number of decision trees.

[0087] Specifically, a natural language processing model is used to extract feature vectors of decision results, and then the Pearson correlation coefficient between the decision result corresponding to one type of information and the average value of all decision results is calculated to obtain the correlation between the decision result of each type of information and other decision results;

[0088] In one embodiment, the second threshold is negatively correlated with the number of decision trees and is set according to the following formula:

[0089]

[0090] Wherein, SIMT is the second threshold, and B is the number of decision trees.

[0091] Specifically, when creating a decision tree to analyze the information of minors, each type of information corresponds to a decision tree. The more decision trees there are, the smaller the correlation between the decision results generated by each decision tree may be. Therefore, when the number of decision trees, that is, the number of information types, is greater than 5, the second threshold is set to 0.5; when the number of decision trees is less than 5, the second threshold is set to Through a large number of tests, the second threshold is set in the above manner, and the more accurate the calculation result of the correlation is, the more accurate the decision result can be controlled.

[0092] Output plan step: After processing the information analysis results using the natural language processing model, output the intervention plan.

[0093] In one embodiment, the output scheme step includes:

[0094] Preparation sub-step: select transformer as the natural language processing model, obtain successful cases of psychological intervention for minors as language templates, divide the language templates into a training set and a validation set, extract keywords from the training set and construct a training decision tree;

[0095] Model training sub-step: input the training decision tree into the natural language processing model for processing and then output the processing result, calculate the Pearson correlation coefficient between the processing result and the validation set, until the Pearson correlation coefficient is greater than 0.7, then determine that the training is completed;

[0096] Plan output sub-step: After the natural language processing model processes the information analysis results, it outputs the intervention plan.

[0097] Specifically, the transformer model used in keyword extraction is adopted. On the one hand, the transformer has an excellent network structure and has a good processing effect on natural language. On the other hand, after the transformer extracts keywords and understands the analysis requirements, its network nodes generate relevant memories. When outputting the plan, it is more suitable for specific minors and their psychological states, and it is also convenient for subsequent iterations to further improve the ability to analyze the psychological states of minors. When implementing this embodiment, the transformer is trained, and the data used are successful cases of psychological intervention for minors by professionals, including various information of the minors being intervened, such as growth experience, living environment, interpersonal relationships, and methods and wording adopted by professionals, thereby realizing that the natural language model can output a more effective intervention plan with text that is more suitable for communication. For example, in this embodiment, part of the intervention plan output for a minor who has suffered domestic violence and school bullying is as follows:

[0098] 1. Opening and establishing connection:

[0099] “Hi, I’m XXX, a psychologist who specializes in helping people with these issues. I understand this is a difficult time for you, but please know that we can overcome these challenges together.”

[0100] "I want to know what is the main problem you are facing now? Are you more troubled by family problems or school problems?"

[0101] 2. Understand the background and needs

[0102] "Domestic violence is a very difficult and sensitive topic for anyone to discuss. I would like to understand what is it about the situation that is so troubling to you?"

[0103] "Are you often targeted by other students in school? Why do they bully you?"

[0104] “Have you tried to communicate this issue with your family or teachers? How have they responded?”

[0105] 3. Provide support and suggestions

[0106] "For domestic violence, it is recommended that you have an honest conversation with your family and tell them how you feel and what is bothering you. At the same time, you may also consider calling the police or seeking other social support."

[0107] "In terms of school, it is recommended that you communicate with your teachers and school counselors to let them know about your difficulties. In addition, you can also join some interest groups or clubs to build friendships and support networks with other classmates."

[0108] "For emotion management, I can teach you some simple techniques such as deep breathing, meditation and relaxation training. These techniques can help you relieve tension and stress."

[0109] 4. Subsequent communication and follow-up

[0110] "I understand this may take some time, but remember, you are not alone. I am here to support you. We can meet regularly over the next few weeks to discuss your progress and any new concerns."

[0111] "I recommend keeping a journal of what's going on each day, how you're feeling, and anything else that's important to you. This will help you better understand your emotions and needs."

[0112] "I believe in your ability to overcome these challenges and emerge a stronger, more confident person. Please believe in yourself, you are not alone."

[0113] like Figure 2 As shown, it is a schematic diagram of a system for intelligently generating a psychological intervention plan for minors proposed by the present invention, which applies the above-mentioned method for generating a psychological intervention plan for minors, including: an information classification module, an information processing module, a decision tree creation module, an adjustment module and a plan output module;

[0114] Information classification module: uses a natural language processing model to classify and extract information about minors to obtain information about their growth experience, living environment, interpersonal relationships, language, behavior, and self-assessment;

[0115] Information processing module: receives analysis requirements submitted by users and uses natural language processing models to extract keywords from classified information based on analysis requirements;

[0116] Creating a decision tree module: Creating decision trees for the classified information respectively, using a natural language processing model to set the discrimination criteria for the keywords, and adding the keywords as child nodes to the decision trees according to the discrimination criteria;

[0117] Adjustment module: calculate the information entropy of the decision tree. If the information entropy is higher than the first threshold, execute the decision tree creation module again. If the information entropy is lower than the first threshold, calculate the probability of the category corresponding to each sub-node in the decision tree, compare the category corresponding to the maximum probability and obtain the decision result.

[0118] Calculate the correlation between the decision result corresponding to each decision tree and the decision results corresponding to other decision trees. If the correlations are all greater than a second threshold, use the decision result as the information analysis result. Otherwise, execute the decision tree creation module again.

[0119] Output plan module: After processing the information analysis results using the natural language processing model, output the intervention plan.

[0120] The present invention creates an analysis system based on a decision tree, analyzes the growth experience, living environment, interpersonal relationship, language, behavior, and self-evaluation of minors to obtain psychological analysis results of minors, and then processes the psychological analysis results through a natural language processing model that learns language doubling, obtains targeted intervention plans, and outputs them in the form of text suitable for communicating with minors, effectively assisting relevant staff in psychological intervention on minors; at the same time, three subtrees are set for the root node of the decision tree, one of which is used as an addition position for irrelevant nodes, thereby reducing the influence of interference information on the analysis results and making the psychological analysis of minors more accurate; by setting an information entropy threshold, the decision tree is rebuilt when it is greater than a predetermined information entropy threshold, thereby achieving the effect of adaptively creating a decision tree during the analysis process and thus improving the accuracy of the analysis results; by calculating the correlation between the analysis results of one type of information and the analysis results of other types of information, when the correlation is lower than the threshold, the analysis is re-performed, thereby improving the accuracy of the analysis results; by using a natural language model to learn a language template, and then processing the psychological analysis results of minors, outputting an intervention plan in the form of text suitable for communicating with minors, thereby improving the technical effect of assisting psychological intervention work on minors.

[0121] In one embodiment, the creating a decision tree module includes:

[0122] Create root node submodule: Create the root node of the decision tree according to the type of information;

[0123] Setting a discrimination criterion submodule: inputting the keyword into the natural language processing model, and allowing the natural language processing model to set a discrimination criterion for the keyword according to the analysis requirements;

[0124] Add child node submodule: determine whether the keyword meets the discrimination criteria. If so, add the keyword as a child node to a subtree of the decision tree. If not, add the keyword to another subtree of the decision tree.

[0125] Specifically, taking the analysis requirement "whether the minor has depression or depression tendency" as an example, it is classified as sentiment analysis, and the root node of the decision tree is created. The root node represents "the emotional state of the minor", and then a child node is created to represent "normal emotional state" and added to the root node, and a child node "abnormal emotional state" is created and added to the root node. Then, subnodes representing various abnormal emotional conditions are added to the subnode representing "abnormal emotional state" to form a subtree. The newly added child node on another node of normal emotional condition and the node together constitute another subtree. Finally, keywords are added to the corresponding subtree according to the judgment criteria. For example, the keyword "don't want to do anything" may be judged by the criteria that if this psychological state appears for more than one week continuously, it is judged as depression, otherwise it is judged as normal. By creating a decision tree, the keywords in each type of information are analyzed, and the analysis of the psychological state of minors is realized.

[0126] In summary, it can be seen that the present invention has the above-mentioned excellent characteristics, which can enhance the performance unprecedented in the prior art and become a product with great practical value.

[0127] The above contents are only preferred embodiments of the present invention. For ordinary technicians in this field, according to the concept of the present invention, there will be changes in the specific implementation methods and application scopes. The content of this specification should not be understood as limiting the present invention.

Claims

1. A method for generating a psychological intervention program for minors, characterized in that: include: Information classification step: Use a natural language processing model to classify the information of minors and extract information on their growth experience, living environment, interpersonal relationships, language, behavior, and self-assessment; Information processing step: receiving the analysis requirements submitted by the user, and using the natural language processing model to extract keywords from the classified information according to the analysis requirements; The step of creating a decision tree: creating a decision tree for each of the classified information, using a natural language processing model to set the discrimination criteria of the keyword, and adding the keyword as a child node to the decision tree according to the discrimination criteria, specifically including: creating a root node sub-step: creating a root node of the decision tree according to the type of information; setting a discrimination criteria sub-step: inputting the keyword into the natural language processing model, and the natural language processing model sets the discrimination criteria for the keyword according to the analysis requirements; adding a child node sub-step: judging whether the keyword meets the discrimination criteria, if so, adding the keyword as a child node to a subtree of the decision tree, if not, adding the keyword to another subtree of the decision tree, wherein the decision tree is a ternary tree; any one of the left subtree, the middle subtree, and the right subtree of the decision tree is used as an irrelevant subtree, and the irrelevant subtree is used to add keywords that cannot be classified by the discrimination criteria, and the irrelevant subtree does not participate in the calculation of information entropy; Adjustment step: Calculate the information entropy of the decision tree. If the information entropy is higher than the first threshold, execute the step of creating a decision tree again. If the information entropy is lower than the first threshold, calculate the probability of the category corresponding to each child node in the decision tree, compare the category corresponding to the maximum probability to obtain the decision result, calculate the correlation between the decision result corresponding to each decision tree and the decision results corresponding to other decision trees, and if the correlation is greater than the second threshold, use the decision result as the information analysis result. Otherwise, execute the step of creating a decision tree again. Specifically, the first threshold is set by the following formula: ， Where h[x] represents the first threshold, is the average probability of the category corresponding to each of the child nodes, A is the depth of the decision tree, and the second threshold is negatively correlated with the number of decision trees and is set according to the following formula: ， Wherein, SIMTT is the second threshold, and B is the number of decision trees; Output plan step: after processing the information analysis results using the natural language processing model, output the intervention plan; Wherein, the output scheme step includes: Preparation sub-step: select transformer as the natural language processing model, obtain successful cases of psychological intervention for minors as language templates, divide the language templates into a training set and a validation set, extract keywords from the training set and construct a training decision tree; Model training sub-step: input the training decision tree into the natural language processing model for processing and then output the processing result, calculate the Pearson correlation coefficient between the processing result and the validation set, until the Pearson correlation coefficient is greater than 0.7, then determine that the training is completed; Plan output sub-step: After the natural language processing model processes the information analysis results, it outputs the intervention plan.

2. The method for generating a psychological intervention program for minors according to claim 1, characterized in that: In the adjustment step, the information entropy is calculated using the following formula: ， Among them, H[X] represents information entropy, P(x) is the probability of the category corresponding to each child node, and A is the depth of the decision tree.

3. The method for generating a psychological intervention program for minors according to claim 1, characterized in that: The calculation process of the correlation includes: Performing feature extraction on the decision result to obtain a two-dimensional feature vector; The Pearson correlation coefficient between the two-dimensional feature vector of each decision result and the two-dimensional feature vector of other decision results is calculated using the following formula: ， Where sim represents the Pearson correlation coefficient, x is the abscissa of the two-dimensional feature vector, and y is the ordinate of the two-dimensional feature vector. and are the average values ​​of the horizontal and vertical coordinates of the two-dimensional feature vectors of all decision results, i is the number of decision trees, and n is the number of decision trees.

4. A system for intelligently generating a psychological intervention plan for minors, applying the method for generating a psychological intervention plan for minors as described in any one of claims 1 to 3, characterized in that: include: Information classification module: uses a natural language processing model to classify and extract information about minors to obtain growth experience information, living environment information, interpersonal relationship information, language information, behavior information, and self-assessment information; Information processing module: receives analysis requirements submitted by users and uses natural language processing models to extract keywords from classified information based on analysis requirements; Creating a decision tree module: Creating decision trees for the classified information respectively, using a natural language processing model to set the discrimination criteria for the keywords, and adding the keywords as child nodes to the decision trees according to the discrimination criteria; Adjustment module: calculate the information entropy of the decision tree. If the information entropy is higher than the first threshold, execute the decision tree creation module again. If the information entropy is lower than the first threshold, calculate the probability of the category corresponding to each sub-node in the decision tree, compare the category corresponding to the maximum probability and obtain the decision result. Calculate the correlation between the decision result corresponding to each decision tree and the decision results corresponding to other decision trees. If the correlations are all greater than a second threshold, use the decision result as the information analysis result. Otherwise, execute the decision tree creation module again. Output plan module: After processing the information analysis results using the natural language processing model, output the intervention plan.

Images