Multi-crowd adaptive training evaluation system

Abstract

The invention, which relates to the technical field of artificial intelligence and skill training examination, discloses a multi-crowd adaptive training evaluation system comprising the following components: a user data management module, an RAG module, an answer processing module, a score analysis module, an intensive training module and a question bank updating module. According to the method, comprehensive and accurate assessment of the user ability is realized by constructing a multi-crowd adaptive training assessment system, the system constructs the user ability portrait by collecting historical data and real-time performance of the user in multiple dimensions and utilizing advanced data processing and analysis technologies, and dynamically adjusts the training content and difficulty, so that the user ability assessment efficiency is improved. According to the method and the system, the ability level and the learning requirement of different users are adapted, the personalized training mode not only improves the participation degree and the learning efficiency of the users, but also helps the users to quickly and accurately identify the ability weakness of the users, and powerful support is provided for subsequent improvement.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence and skill training and examination, and in particular to a multi-group adaptive training and evaluation system. Background Art

[0002] With the rapid development of artificial intelligence technology, especially the breakthrough progress of big model technology in recent years, and the significant improvement of natural language processing capabilities, these cutting-edge technologies have deeply penetrated into every corner of society, including education, entertainment, customer service and many other key areas. In the field of education, the concept of personalized teaching is increasingly valued. It emphasizes tailoring teaching content and methods according to the individual differences and ability levels of each student in order to achieve more efficient and targeted learning results. In this context, natural language processing technology based on big models provides strong technical support for personalized teaching.

[0003] Traditional training and examination systems mostly use preset question banks and test papers. This method lacks flexibility and is difficult to dynamically adjust according to individual differences of users. Therefore, traditional systems often cannot meet users' needs for personalized training. In addition, traditional systems are often based on static rules in question generation, which makes it difficult to dynamically adjust according to the actual ability level and weak links of the candidates, resulting in poor training targeting and effectiveness. This not only limits the speed at which candidates improve their abilities, but also increases training costs and time costs.

[0004] Therefore, it is necessary to develop a multi-group adaptive training evaluation system. The application of this system will greatly improve the training efficiency and quality, reduce the training cost, and has broad application prospects. Summary of the invention

[0005] The purpose of the present invention is to make up for the shortcomings of the existing technology and provide a multi-group adaptive training evaluation system. The system includes user data management, risk assessment, early warning, intervention measure recommendation and closed-loop management modules. By collecting user data, a capability portrait is constructed, and the RAG module is used to generate questions adapted to user capabilities. The answer processing module provides real-time interactive feedback. The scoring analysis module evaluates the user's ability scores in five dimensions. The intensive training module generates personalized wrong question reinforcement questions for weak links. The question bank update module supports question generation and knowledge base optimization.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a multi-group adaptive training evaluation system, the system comprising: a user data management module, a RAG module, a question processing module, a scoring analysis module, an intensive training module and a question bank update module;

[0007] The user data management module collects historical answer records of users in the system, information on various training courses participated in, business operation data in actual work scenarios, and self-evaluation and learning needs information actively submitted by users from the internal system of the enterprise and the self-input of users. After cleaning, preprocessing and standardization, the comprehensive ability level of users is evaluated using the comprehensive ability evaluation formula of users, and a user ability portrait is constructed;

[0008] The RAG module includes a knowledge base retrieval submodule and an agent submodule. The knowledge base retrieval submodule uses semantic retrieval technology to query keywords and similar questions. The agent submodule includes multiple question-generating agents. The question-generating agents generate questions based on the provided keywords, calculate the difficulty value of the questions to determine the difficulty of the questions, and select questions that are suitable for the user's ability in combination with the user's ability level.

[0009] The answer processing module provides users with personalized test interface layout and display settings, provides different display effects for different types of questions, provides real-time feedback function, provides instant feedback on user operations during the user's answering process, records the user's operation details during the answering process, device information and network environment information during the answering process, and stores all information;

[0010] The scoring analysis module: obtains the output data of the question setting agent and the user's answer data through the scoring agent, analyzes the user's answer in combination with the preset scoring criteria, generates the user's five-dimensional ability scores in the five dimensions of communication skills, problem-solving ability, business knowledge, risk prevention, and upgrade awareness, and provides a detailed analysis report;

[0011] The intensive training module: constructs a user ability knowledge graph based on the user's five-dimensional ability score and analysis report, analyzes the user's weak links in the knowledge system, predicts the user's improvement potential and difficulty in different knowledge points, integrates the user's answer data, business operation data, and training and learning data for comprehensive analysis, analyzes the causes of the wrong questions that the user encounters during the answering process, determines the user's ability areas that need priority intensive training, generates personalized wrong question reinforcement questions, and adjusts the difficulty of subsequent questions based on the user's real-time performance;

[0012] The question bank update module provides an administrator interface for importing text files in various formats, and performs format conversion and preliminary processing on the imported text. The question-generating agent generates questions based on text analysis combined with the RAG module. The administrator reviews the generated questions. Unqualified questions are regenerated by the agent. Passed questions and related knowledge are stored in the knowledge base. The knowledge base is regularly optimized and maintained, outdated and repeated content is deleted, and a knowledge association mechanism is established to facilitate users to retrieve related knowledge.

[0013] Furthermore, the user data management module uses the user ability comprehensive evaluation formula to evaluate the user's comprehensive ability level in multiple aspects. Suppose the user's performance data in n different tasks and scenarios is x i (i=1,2,…,n), the weight corresponding to each task and scenario is w i , and each task and scenario has m evaluation indicators, and the indicator value is y ij (j=1,2,…,m), the indicator weight is v ij , calculate the ability score for each task and scenario: Calculate the comprehensive evaluation score of user capabilities:

[0014] Furthermore, the RAG module calculates the difficulty value of the question through the question difficulty prediction formula to determine the question difficulty. Assuming that the question contains k knowledge points and the difficulty coefficient of knowledge point j is d j , the association complexity between knowledge points is c ij , the user's mastery of knowledge point j is m j The formula is:

[0015] Furthermore, the Agent and corresponding functions involved in the RAG module:

[0016] Multiple-choice question agent: responsible for generating multiple-choice questions, providing answer analysis and scoring. When generating questions, it will refer to the user's historical scores and knowledge weaknesses;

[0017] Multiple-choice question agent: responsible for generating multiple-choice questions. When generating questions, it considers multiple correct answers and provides detailed scoring rules when scoring.

[0018] True or False Question Agent: Responsible for generating true or false questions and providing true or false judgments and explanations;

[0019] Question-and-Answer Agent: It handles the generation and scoring of open-ended questions. It can generate open-ended questions and provide scoring and feedback based on the user's answers.

[0020] Furthermore, the five capability dimensions covered in the scoring analysis module are:

[0021] Communication skills: assess the user's ability to express, understand and communicate, the user's logic, clarity and interactivity;

[0022] Problem-solving ability: Tests the user's ability to analyze, propose solutions and make decisions when faced with complex problems, emphasizing practicality and innovation;

[0023] Business knowledge: assess the user's mastery of professional knowledge, including basic theories, application skills and the latest trends;

[0024] Risk prevention: examines the user's ability to identify, assess and respond to risks, and their judgment in uncertain environments;

[0025] Upgrade awareness: Assess users' initiative and awareness in career development, skill improvement and learning new knowledge, and their ability to accept and apply new technologies and methods.

[0026] Furthermore, the intensive training module evaluates the user's improvement potential P on a specific knowledge point through the user knowledge point improvement potential evaluation formula i , let the average score of the user in the past n exercises involving knowledge point i be S i , the score rate in the most recent exercise involving knowledge point i is L i , let T be the average learning time of the user in the past m times of learning knowledge point i i , the learning time of the most recent learning knowledge point i is R i , let the number of associated knowledge points of knowledge point i in the entire knowledge system be C i , the formula is: Among them, α, β, and γ are weight coefficients.

[0027] Furthermore, the intensive training module calculates the difficulty D of the knowledge point through the knowledge point difficulty calculation formula. i , let the level of knowledge point i in the knowledge base be H i The error rate of questions related to this knowledge point among all questions is E i , let the content length of knowledge point i be L eni , the average content length of all knowledge points is The formula is: Among them, ω, φ, and θ are weight coefficients.

[0028] Furthermore, the intensive training module determines the ability areas that need the most priority intensive training according to the ability gap, ability importance and improvement potential through the training demand priority formula. Suppose the gap between the user's current ability score and the target ability score on the ability dimension l is g l , the importance coefficient of capability dimension l in the overall capability system is i l , the user's potential coefficient for improvement in this capability dimension is p l , the formula is: TRP l =g l ×i l ×p l .

[0029] Furthermore, the difficulty of subsequent questions in the training module is adjusted according to the user's real-time performance:

[0030] Difficulty increase rules: If you answer ≥3 questions correctly in a row, a higher difficulty question will be automatically generated;

[0031] Rules for reducing difficulty: If you answer ≥ 3 questions incorrectly in a row, basic questions will be pushed to make up for the difference.

[0032] Compared with the prior art, this multi-person adaptive training evaluation system has the following beneficial effects:

[0033] 1. The present invention realizes a comprehensive and accurate assessment of user capabilities by constructing a multi-group adaptive training assessment system. The system collects historical data and real-time performance of users in multiple dimensions, uses advanced data processing and analysis technologies, constructs user capability portraits, and dynamically adjusts training content and difficulty to adapt to the ability levels and learning needs of different users. This personalized training method not only improves user participation and learning efficiency, but also helps users quickly and accurately identify their own ability shortcomings, providing strong support for subsequent improvement.

[0034] 2. Through the synergistic effect of the RAG module and multiple question-setting agents, the present invention can automatically generate questions that match the user's ability level, and adjust the difficulty of subsequent questions in real time according to the user's answers. At the same time, the intensive training module can also build a user ability knowledge map based on the user's five-dimensional ability score and feedback report, analyze the user's weak links in the knowledge system, and generate personalized reinforcement questions for wrong questions. This intelligent intervention and intensive training mechanism not only helps users consolidate what they have learned, but also effectively improves users' performance in key ability dimensions such as communication skills, problem-solving ability, business knowledge, risk prevention and upgrade awareness.

[0035] Other advantages, objectives and features of the present invention will be set forth in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0037] Figure 1 A flowchart of a multi-group adaptive training evaluation system;

[0038] Figure 2 It is a schematic diagram of the data fragment of the question bank;

[0039] Figure 3 This is a schematic diagram of a historical five-dimensional scoring data segment;

[0040] Figure 4 This is a schematic diagram of a user's answer data segment;

[0041] Figure 5 To personalize the test flow chart;

[0042] Figure 6 Generate flowcharts for knowledge-based questions. DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] Embodiment 1:

[0045] Implementation of multi-group adaptive training and assessment system in enterprise employee training and competency assessment:

[0046] Embodiment Background: A large financial enterprise has diversified businesses and many employee positions. In order to improve the professional quality of employees, the enterprise wants to establish an adaptive training and capability assessment system to adapt to the changing needs of the market and business.

[0047] Example process:

[0048] User data collection and management, obtain data from multiple systems within the enterprise and employees' self-input, including information on training courses, business operations, self-evaluation and learning needs. After cleaning, preprocessing and standardization, the comprehensive ability of employees is evaluated using the user ability comprehensive evaluation formula. Suppose the performance data of employees in n different tasks or scenarios is x i (i=1,2,…,n), the weight corresponding to each task or field is w i , and each task or scenario has m evaluation indicators, and the indicator value is y ij (j=1,2,…,m), the indicator weight is v ij , then the ability score for each task or scenario is: The comprehensive evaluation scores of user capabilities are as follows: Take employee Xiao Li as an example. Assume that his ability in three task scenarios is evaluated: customer communication (weight w1), business operation (weight w2), and risk assessment (weight w3). In the customer communication scenario, there is language expression clarity (indicator value y 11 , weight v 11 ), customer feedback satisfaction (index value y 12 , weight v 12 ) evaluation indicators; transaction processing efficiency (indicator value y 21 , weight v 21 ), business operation accuracy (index value y 22 , weight v 22 ); Risk identification accuracy in risk assessment scenarios (index value y 31 , weight v 31 ), risk response timeliness (index value y 32 , weight v 32 ), calculate Xiao Li’s ability score in each scenario, such as the score in the customer communication scenario: S1 = ∑ i＝1 v 1j y 1j , (here m = 2, that is, S1 = v 11 y 11 +v 12 y 12 ); the scores in the business operation scenario are: Risk assessment scenario Then calculate Xiao Li's comprehensive user ability evaluation score as: Build a user capability profile.

[0049] The RAG module sets questions and determines the difficulty. The knowledge base retrieval submodule uses semantic retrieval technology to search for relevant keywords and similar questions based on Xiao Li's ability profile and learning needs. The multiple-choice question agent in the Agent submodule generates multiple-choice questions for Xiao Li. Suppose the questions contain k knowledge points and the difficulty coefficient of knowledge point j is d. j , the association complexity between knowledge points is c ij (i≠j), Xiao Li’s mastery of knowledge point j is m j , then the difficulty of the question is: Generate questions that are moderately difficult and challenging.

[0050] Question answering processing: Xiao Li logs in to answer questions. The system provides a personalized interface with different display effects for different questions. The system provides real-time feedback during answering, records and stores operation details, equipment and network information. For example, if Xiao Li chooses the wrong answer to the risk assessment multiple-choice question, the system will immediately prompt the correct answer and analysis, and record the answering time and the number of times the answer is modified.

[0051] Scoring analysis: The scoring agent obtains the data of the question-setting agent and Xiao Li's answers, and scores them according to the preset scoring standards. It evaluates them from five dimensions: communication skills, problem-solving ability, business knowledge, risk prevention, and upgrade awareness. If Xiao Li answers this risk assessment multiple-choice question incorrectly, he will get a certain score after deducting points according to the scoring rules. The system analyzes his performance in the risk prevention dimension and generates a five-dimensional ability score and a detailed feedback report.

[0052] Strengthen training, build a user ability knowledge map based on Xiao Li's score and report, analyze weak links, and evaluate Xiao Li's improvement potential P in specific knowledge points through the user knowledge point improvement potential evaluation formula i , let Xiao Li’s average score in the past n exercises or exams involving knowledge point i be S i , the score rate in the most recent exercise or exam involving knowledge point i is L i , the average learning time of the past m learning times of knowledge point i is T i , the learning time of the most recent learning knowledge point i is R i , the number of related knowledge points of knowledge point i in the entire knowledge system is C i , assuming weight coefficients α, β, γ, the formula is Calculation knowledge point difficulty D i , let the level of knowledge point i in the knowledge base be H i The error rate of questions related to this knowledge point among all questions is E i , the content length of knowledge point i is L eni , the average content length of all knowledge points is Assume weight coefficients ω, φ, θ, and the formula is: According to the training demand priority formula, the priority training ability areas are determined. Let the difference between Xiao Li’s current ability score and the target ability score on ability dimension l be g. l , the importance coefficient of capability dimension l in the overall capability system is i l , Xiao Li’s potential coefficient for improvement in this ability dimension is p l , the formula is: TRP l =g l ×i l ×p l After confirmation, the system generates personalized reinforcement questions for wrong questions. The difficulty of the questions is adjusted according to Xiao Li’s real-time performance. For example, if he answers ≥3 questions correctly in a row, more difficult questions will be generated according to the difficulty increase rules. If he answers ≥3 questions incorrectly in a row, basic questions will be pushed to make up for the difference according to the difficulty reduction rules.

[0053] The question bank is updated, and the administrator imports the text file of the financial industry. After the system processes it, the question agent generates questions, and the administrator reviews them. Qualified questions are stored in the knowledge base, and regular optimization and maintenance are carried out to delete outdated and repeated content and establish a knowledge association mechanism. Through this embodiment, the system of the present invention plays a full role in enterprise employee training and capacity assessment, improving employee capabilities and enterprise competitiveness.

[0054] To sum up, the multi-group adaptive training evaluation system of the present invention has achieved remarkable results in enterprise applications. The user data management module provides accurate portraits and lays the foundation for personalized training. The RAG module sets questions scientifically to stimulate employees' challenges. The answer processing module provides real-time feedback to optimize the learning experience. The scoring analysis module conducts multi-dimensional evaluation to indicate the direction of improvement. The enhanced training module conducts in-depth analysis to achieve adaptive improvement. The question bank update module keeps pace with the times to ensure knowledge reserves. All modules work together to build an efficient closed-loop learning ecosystem, effectively improving employee quality and corporate competitiveness. It has broad promotion value and application prospects in many industries.

[0055] Embodiment 2:

[0056] Implementation of multi-group adaptive training evaluation system in online education platform:

[0057] Background of the implementation example: With the rapid development of online education, a large online education platform has a large number of students from different regions, different age levels and different learning backgrounds. The platform provides a rich variety of courses, covering multiple fields such as language learning, vocational skills training, and academic examination tutoring. In order to improve the teaching quality and meet the personalized learning needs of students, the platform decided to introduce the multi-group adaptive training evaluation system of the present invention to achieve accurate teaching and efficient learning.

[0058] Example process:

[0059] User data collection and management. The platform obtains information about the courses students have participated in from their learning history records, including course categories, learning duration, and course completion. It collects homework, test, and exam answer data from students' answer records. Students can also actively submit self-evaluations in their personal settings, such as learning goals, learning styles, and information on knowledge points where they feel they are weak, as well as feedback and suggestions on platform functions and course content (regarded as learning needs). The collected data undergoes strict data cleaning to remove erroneous data and duplicate records, and is standardized to make different types of data comparable.

[0060] The comprehensive ability evaluation formula of users is used to evaluate the comprehensive ability of students. Suppose the performance data of students in different course types (such as language courses and skill courses are regarded as different tasks or scenarios, set as n) is x i (i=1,2,…,n), the weight corresponding to each course type is w i, taking language courses as an example, there is vocabulary mastery (evaluation indicator 1, indicator value y 11 , with weight v 11 ), grammatical understanding (evaluation indicator 2, indicator value is y 12 , with weight v 12 ), oral expression fluency (evaluation indicator 3, indicator value is y 13 , with weight v 13 )m evaluation indicators, the ability score under the language course is: (Here m = 3, that is, S1 = v 11 y 11 +v 12 y 12 +v 13 y 13 ), and the ability scores S of other course types can be calculated in the same way i , and then the comprehensive evaluation score of the students' user capabilities is obtained as follows: Build a user capability profile of the students based on the assessment results to provide a basis for subsequent personalized learning.

[0061] The RAG module determines the questions and difficulty. The knowledge base retrieval submodule searches for relevant keywords and similar questions in the platform's massive question bank based on the student's ability profile and learning needs through semantic retrieval technology. For example, for students who have difficulty learning grammar, questions containing grammar knowledge points are retrieved. The multiple-choice question agent in the agent submodule generates multiple-choice questions for students. Assume that the generated questions involve k knowledge points: tense (knowledge point 1), clause (knowledge point 2), and subjunctive mood (knowledge point 3). The difficulty coefficient of knowledge point 1 is d1, the difficulty coefficient of knowledge point 2 is d2, and the difficulty coefficient of knowledge point 3 is d3; the association complexity between knowledge point 1 and knowledge point 2 is c 12 , the association complexity between knowledge point 1 and knowledge point 3 is c 13 , the association complexity between knowledge point 2 and knowledge point 3 is c 23 ; The student's mastery of knowledge point 1 is m1, the student's mastery of knowledge point 2 is m2, and the student's mastery of knowledge point 3 is m3. The difficulty of the question is calculated according to the question difficulty prediction formula, which is: Adapting the difficulty of generated questions to students’ current ability level will help them improve gradually.

[0062] Question answering processing: students log in to the platform to study and answer questions. The system provides them with a personalized learning interface, such as adjusting the interface color and font size according to the students' visual preferences, and providing different display effects and operation methods for different types of questions (such as listening questions, reading questions, and writing questions). During the answering process, the system provides real-time feedback. For example, if a student chooses the wrong answer when answering a tense multiple-choice question, the system immediately gives an error prompt, explains the reason for the correct answer, and provides explanation links to related knowledge points. The system also records the student's operational details during the answering process, such as the time used to answer the question, whether there is any modification of the answer, and the device information used by the student (such as device type, operating system) and network environment information (such as network speed and stability), and stores this information for subsequent analysis of the student's learning behavior and habits.

[0063] Scoring analysis: The scoring agent obtains the output data of the multiple-choice agent and the student's answer data, and scores according to the preset scoring criteria. Assuming that the full score of this tense multiple-choice question is 5 points, the student answers incorrectly, and 2 points are deducted according to the scoring rules, and the actual score is 3 points. The system evaluates the student from five dimensions: communication skills (such as interaction in oral expression), problem-solving ability (such as the ability to analyze and solve problems in reading comprehension), business knowledge (such as mastery of course knowledge), risk prevention (such as the ability to avoid careless mistakes in exams, which can be compared to avoiding mistakes in learning), and upgrade awareness (such as actively seeking to increase the difficulty of learning and expand knowledge). In this question, the business knowledge dimension is mainly involved. According to the analysis of the student's answer, the system judges the student's mastery of tense knowledge points and gives corresponding scores on this dimension. Finally, the student's five-dimensional ability score on this question is generated, and a detailed feedback report is provided, pointing out the student's weaknesses in tense, such as confusion between the past tense and the present perfect tense, and providing improvement suggestions, such as targeted review of relevant grammar rules and doing more exercises.

[0064] Strengthen training, build a user ability knowledge map based on the trainees' five-dimensional ability scores and feedback reports, and analyze and find that the trainees are relatively weak in clause-related knowledge points, such as being unskilled in the use of relative words in attributive clauses. Use the user knowledge point improvement potential evaluation formula to evaluate the trainees' improvement potential P in this knowledge point i , let S be the average score of the students in the past n exercises or exams involving clause knowledge points i , the score rate in the most recent exercise or exam involving this knowledge point is L i , the average learning time of the past m times of learning this knowledge point is T i The most recent learning time of this knowledge point is R i , the number of related knowledge points of the clause knowledge point in the entire language knowledge system is C i, let the weight coefficients α, β, γ, the formula is: Difficulty of calculating clause knowledge points D i , let the knowledge point be at the level of

[0065] Grade H i (such as the levels in elementary grammar, intermediate grammar, and advanced grammar), the error rate of questions related to this knowledge point among all questions is E i , the length of the knowledge point content (such as the length of text or video required to explain the knowledge point) is L eni , the average content length of all knowledge points is Assume weight coefficients ω, φ, θ, and the formula is: According to the training demand priority formula, determine the ability areas that students need to strengthen training most. Suppose the gap between the student's current ability score and the target ability score on the ability dimension l (such as the grammar knowledge sub-dimension under the business knowledge dimension) is g. l , the importance coefficient of this capability dimension in the overall capability system is i l , the student's potential coefficient for improvement in this ability dimension is p l , then TRP l =g l ×i l ×p l After confirmation, the system generates personalized reinforcement questions for wrong questions, such as special exercises for relative words in attributive clauses, including different types of sentence completion and error correction questions. During the students' intensive training, the difficulty of subsequent questions is adjusted according to their real-time performance. If students answer ≥3 questions correctly in a row, the system will automatically generate questions of higher difficulty, such as complex sentence analysis questions involving the mixed use of multiple clauses. If students answer ≥3 questions incorrectly in a row, basic questions will be pushed to make up for the gap, such as simple single-sentence clause recognition and relative word selection questions.

[0066] Question bank update: platform administrators import authoritative language learning materials, the latest test questions, and exercises written by excellent teachers in various formats of text files through the administrator interface. The system performs format conversion and preliminary processing on the imported text, such as converting the document format into a question bank data format recognizable by the system and extracting key information. The question-generating agent generates new questions based on text analysis combined with the RAG module, such as generating relevant questions based on new grammar rules or common test points. The administrator reviews the generated questions, and unqualified questions (such as wrong answers and unclear expressions) are regenerated by the agent. The reviewed questions and related knowledge are stored in the knowledge base, and the knowledge base is regularly optimized and maintained, deleting outdated questions (such as questions related to knowledge points in old textbooks that are no longer applicable) and duplicate content. At the same time, a knowledge association mechanism is established to facilitate students to retrieve relevant knowledge, such as establishing association links between grammar knowledge points and corresponding examples and exercises, so that students can learn and review more systematically.

[0067] To sum up, the present invention shows significant advantages in the online education platform scenario. It lays the foundation for personalized learning by comprehensively collecting and analyzing student data, accurately evaluating student abilities and constructing portraits. The RAG module generates adaptive questions based on the student's status, and the answer processing module provides high-quality experience and real-time feedback. The scoring analysis module provides targeted guidance through multi-dimensional evaluation. The intensive training module deeply analyzes student needs and achieves accurate reinforcement and dynamic difficulty adjustment. The question bank update module ensures that knowledge is up-to-date and rich. The modules work together to effectively improve student learning outcomes and platform teaching quality, providing an efficient and intelligent solution for online education, and has broad application prospects and promotion value.

[0068] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A multi-group adaptive training evaluation system, characterized in that: The system includes: user data management module, RAG module, answer processing module, score analysis module, intensive training module and question bank update module; The user data management module collects historical answer records of users in the system, information on various training courses participated in, business operation data in actual work scenarios, and self-evaluation and learning needs information actively submitted by users from the internal system of the enterprise and the self-input of users. After cleaning, preprocessing and standardization, the comprehensive ability level of users is evaluated using the comprehensive ability evaluation formula of users, and a user ability portrait is constructed; The RAG module includes a knowledge base retrieval submodule and an agent submodule. The knowledge base retrieval submodule uses semantic retrieval technology to query keywords and similar questions. The agent submodule includes multiple question-generating agents. The question-generating agents generate questions based on the provided keywords, calculate the difficulty value of the questions to determine the difficulty of the questions, and select questions that are suitable for the user's ability in combination with the user's ability level. The answer processing module provides users with personalized test interface layout and display settings, provides different display effects for different types of questions, provides real-time feedback function, provides instant feedback on user operations during the user's answering process, records the user's operation details during the answering process, device information and network environment information during the answering process, and stores all information; The scoring analysis module: obtains the output data of the question setting agent and the user's answer data through the scoring agent, analyzes the user's answer in combination with the preset scoring criteria, generates the user's five-dimensional ability scores in the five dimensions of communication skills, problem-solving ability, business knowledge, risk prevention, and upgrade awareness, and provides a detailed analysis report; The intensive training module: constructs a user ability knowledge graph based on the user's five-dimensional ability score and analysis report, analyzes the user's weak links in the knowledge system, predicts the user's improvement potential and difficulty in different knowledge points, integrates the user's answer data, business operation data, and training and learning data for comprehensive analysis, analyzes the causes of the wrong questions that the user encounters during the answering process, determines the user's ability areas that need priority intensive training, generates personalized wrong question reinforcement questions, and adjusts the difficulty of subsequent questions based on the user's real-time performance; The question bank update module provides an administrator interface for importing text files in various formats, and performs format conversion and preliminary processing on the imported text. The question-generating agent generates questions based on text analysis combined with the RAG module. The administrator reviews the generated questions. Unqualified questions are regenerated by the agent. Passed questions and related knowledge are stored in the knowledge base. The knowledge base is regularly optimized and maintained, outdated and repeated content is deleted, and a knowledge association mechanism is established to facilitate users to retrieve related knowledge.

2. A multi-group adaptive training evaluation system according to claim 1, characterized in that: The user data management module uses the user ability comprehensive evaluation formula to evaluate the user's comprehensive ability level in multiple aspects. Suppose the user's performance data in n different tasks and scenarios is x i (i=1,2,…,n), the weight corresponding to each task and scenario is w i , and each task and scenario has m evaluation indicators, and the indicator value is y ij (j=1,2,…,m), the indicator weight is v ij , calculate the ability score for each task and scenario: Calculate the comprehensive evaluation score of user capabilities:

3. A multi-group adaptive training evaluation system according to claim 1, characterized in that: The RAG module calculates the difficulty value of the question through the question difficulty prediction formula to determine the question difficulty. Assume that the question contains k knowledge points and the difficulty coefficient of knowledge point j is d j , the association complexity between knowledge points is c ij , the user's mastery of knowledge point j is m j The formula is:

4. A multi-group adaptive training evaluation system according to claim 1, characterized in that: Agents and corresponding functions involved in the RAG module: Multiple-choice question agent: responsible for generating multiple-choice questions, providing answer analysis and scoring. When generating questions, it will refer to the user's historical scores and knowledge weaknesses; Multiple-choice question agent: responsible for generating multiple-choice questions. When generating questions, it considers multiple correct answers and provides detailed scoring rules when scoring. True or False Question Agent: Responsible for generating true or false questions and providing true or false judgments and explanations; Question-and-Answer Agent: It handles the generation and scoring of open-ended questions. It can generate open-ended questions and provide scoring and feedback based on the user's answers.

5. The multi-group adaptive training evaluation system according to claim 1, characterized in that: The five competency dimensions covered in the scoring analysis module are: Communication skills: assess the user's ability to express, understand and communicate, the user's logic, clarity and interactivity; Problem-solving ability: Tests the user's ability to analyze, propose solutions and make decisions when faced with complex problems, emphasizing practicality and innovation; Business knowledge: assess the user's mastery of professional knowledge, including basic theories, application skills and the latest trends; Risk prevention: examines the user's ability to identify, assess and respond to risks, and their judgment in uncertain environments; Upgrade awareness: Assess users' initiative and awareness in career development, skill improvement and learning new knowledge, and their ability to accept and apply new technologies and methods.

6. A multi-group adaptive training evaluation system according to claim 1, characterized in that: The intensive training module evaluates the user's improvement potential P on a specific knowledge point through the user knowledge point improvement potential evaluation formula i , let the average score of the user in the past n exercises involving knowledge point i be S i , the score rate in the most recent exercise involving knowledge point i is L i , let T be the average learning time of the user in the past m times of learning knowledge point i i , the learning time of the most recent learning knowledge point i is R i , let the number of associated knowledge points of knowledge point i in the entire knowledge system be C i , the formula is: Among them, α, β, and γ are weight coefficients.

7. The multi-group adaptive training evaluation system according to claim 1, characterized in that: The intensive training module calculates the difficulty D of the knowledge point through the knowledge point difficulty calculation formula i , let the level of knowledge point i in the knowledge base be H i The error rate of questions related to this knowledge point among all questions is E i , let the content length of knowledge point i be The average content length of all knowledge points is The formula is: Among them, ω, φ, and θ are weight coefficients.

8. The multi-group adaptive training evaluation system according to claim 1, characterized in that: The intensive training module determines the ability areas that need the most priority intensive training according to the ability gap, ability importance and improvement potential through the training demand priority formula. Suppose the gap between the user's current ability score and the target ability score on the ability dimension l is g l , the importance coefficient of capability dimension l in the overall capability system is i l , the user's potential coefficient for improvement in this capability dimension is p l , the formula is: TRP l =g l ×i l ×p l .

9. The multi-group adaptive training evaluation system according to claim 1, characterized in that: The difficulty of subsequent questions in the training module is adjusted according to the user's real-time performance: Difficulty increase rules: If you answer ≥3 questions correctly in a row, a higher difficulty question will be automatically generated; Rules for reducing difficulty: If you answer ≥ 3 questions incorrectly in a row, basic questions will be pushed to make up for the difference.

Images