Personalized intelligent evaluation system and method based on multi-role intelligent agent

Abstract

The invention provides a personalized intelligent evaluation system and method based on a multi-role intelligent agent, and belongs to the technical field of intelligent evaluation in the education field, and the system comprises a data input module which is used for collecting student homework data, personal learning information and evaluation standards; the personal learning information base module is used for storing personal learning information of students and comprises a learning condition analysis unit, a learning style and preference unit, a learning demand and target unit and a teacher marking information unit; the evaluation standard library module is used for storing evaluation standards and evaluation cases preset by teachers; by combining the pre-training language model, sentiment analysis, supervised learning and deep learning algorithms, the homework content of students can be deeply understood and analyzed, grammar errors, logic problems and innovation points can be automatically identified, the language style of teachers can be understood through sentiment color analysis, and personalized modification suggestions can be provided for the students.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent evaluation in the field of education, and specifically relates to a personalized intelligent evaluation system and method based on multi-role agents, providing an automated and personalized solution for evaluating and providing feedback on subjective creative assignments of college students, such as planning proposals, creative writing, speech drafts, etc. Background Art

[0002] In the field of education, especially in college teaching, evaluating students' assignments is an important part of teaching activities. Common online learning platforms have made significant progress in handling the grading of objective questions, enabling batch and accurate grading. However, for subjective creative assignments, such as planning proposals, creative writing, speech drafts, etc., it mainly relies on manual grading by teachers. This method has the following deficiencies:

[0003] 1. Low efficiency: Teachers need to spend a large amount of time reviewing students' subjective creative assignments one by one, which is not only time-consuming and laborious but also difficult to handle a large number of assignments in a short time, resulting in low grading efficiency.

[0004] 2. Insufficient personalized feedback: Due to the huge workload, teachers often have difficulty providing personalized and accurate feedback to each student, which limits the quality improvement and personalized development of students in subjective creative assignments.

[0005] 3. Inconsistent evaluation quality: Manual grading is affected by teachers' personal experience and emotions, resulting in possible inconsistencies in evaluation criteria and difficulty in ensuring the objectivity and fairness of evaluation.

[0006] 4. Insufficient cultivation of innovation ability: Traditional evaluation methods lack the cultivation of students' innovation ability and higher-order thinking, and cannot effectively stimulate students' internal drive for optimizing assignments, creativity, and critical thinking.

[0007] To address the above problems, although there have been some attempts to use artificial intelligence technology to assist in evaluating subjective creative assignments, these attempts are often limited to a single evaluation model and lack flexibility and personalization. In addition, these systems usually cannot handle teachers' personalized evaluation needs and students' emotional needs well, nor can they effectively combine with students' personal learning information and specific needs. Summary of the Invention

[0008] The present invention provides a personalized intelligent evaluation system and method based on multi-role agents, which can achieve technical implementation in aspects such as understanding and analyzing students' homework, analyzing teachers' language styles and emotions, training evaluation models, and identifying grammar errors, logical problems, and innovation points in homework, so as to improve the efficiency and quality of subjective creative homework evaluation. This system can not only provide students with instant personalized evaluations and optimization suggestions, but also provide teachers with auxiliary tools for batch personalized marking of subjective creative homework, while cultivating students' innovation ability and higher-order thinking. Through the comprehensive application of natural language processing technology, machine learning technology, and database technology, this system can achieve in-depth understanding and analysis of students' homework content, generate personalized marking opinions and modification suggestions, thereby significantly improving the efficiency and quality of evaluation.

[0009] To solve the above technical problems, the present invention adopts the following technical solutions:

[0010] A personalized intelligent evaluation system based on multi-role agents, comprising:

[0011] A data input module, which is used to collect students' homework data, personal learning information, and evaluation criteria;

[0012] A personal learning information library module, which is used to store students' personal learning information, including a learning situation analysis unit, a learning style and preference unit, a learning need and goal unit, and a teacher annotation information unit;

[0013] An evaluation criteria library module, which is used to store the evaluation criteria and evaluation cases preset by teachers;

[0014] An evaluation processing agent module, which includes a student self-evaluation agent and a teacher evaluation agent. The evaluation processing agent module is used to call personal learning information and the evaluation criteria preset by teachers to form personalized evaluation indicators, and then generate marking opinions and modification suggestions; and

[0015] A data output module, which is used to display the marking opinions and modification suggestions to students and teachers.

[0016] Preferably, the personalized intelligent evaluation system based on multi-role agents further includes a data feedback module for collecting feedback information from students and teachers. The data feedback module includes a student self-evaluation function terminal and a teacher batch marking terminal.

[0017] Preferably, the learning situation analysis unit includes a prerequisite course sub-unit, a learning background sub-unit, a knowledge reserve sub-unit, and a skill mastery sub-unit.

[0018] Preferably, the learning style and preference unit includes a cognitive style subunit, a learning method preference subunit, an example preference subunit, an interest area subunit, a social media behavior subunit, a feedback preference subunit, and a mental state and emotional change subunit.

[0019] Preferably, the learning need and goal unit includes a career planning subunit, an ability development goal subunit, a weak point of knowledge architecture subunit, and a personal development plan subunit.

[0020] Preferably, the teacher annotation information unit includes a classroom performance subunit, a personality trait subunit, and an emotional color subunit.

[0021] Preferably, the evaluation criterion library module includes a number of assignment units, and each assignment unit includes an evaluation case subunit and an evaluation criterion subunit. Among them, the evaluation criterion subunit includes evaluation indicators and evaluation weights.

[0022] According to the foregoing usage method of the personalized intelligent evaluation system based on multi-role agents, it includes the following steps:

[0023] S0: Start;

[0024] S1: Submit an assignment. Whether the student self-evaluates. If so, go to S2; if not, go to S3;

[0025] S2: The student self-evaluation agent calls the evaluation criteria and personal learning information to generate modification suggestions, and determines whether the student asks questions about the modification suggestions. If so, repeat S2; if not, go to S1;

[0026] S3: The teacher evaluation agent generates review opinions and determines whether the teacher modifies the review opinions. If so, go to S4; if not, go to S5;

[0027] S4: The teacher marks the modification requirements and gives feedback, and then enters S3 again;

[0028] S5: Release the review opinions, and then enter S6;

[0029] S6: End.

[0030] Preferably, the foregoing usage method further includes the step of constructing a student self-evaluation agent, including the following steps:

[0031] S20: Start;

[0032] S21: Identify the assignment format and clean the data, and enter S22;

[0033] S22: Call the preset evaluation criteria and the student's personal learning information to form personalized evaluation indicators, and enter S23;

[0034] S23: Call the large model and the preset case library, generate personalized evaluation content and output it, then enter S24;

[0035] S24: Whether the student has follow-up information about the evaluation content. If so, enter S22; if not, enter S25;

[0036] S25: End.

[0037] Preferably, the usage method further includes the step of constructing a teacher evaluation agent, including the following steps:

[0038] S30: Start;

[0039] S31: Identify the homework format and clean the data, then enter S32;

[0040] S32: Call the preset evaluation criteria and the student's personal learning information to form personalized evaluation indicators, then enter S33;

[0041] S33: Call the large model and the preset case library to generate personalized evaluation content, then enter S34;

[0042] S34: Determine whether teacher's modification opinions are received. If so, enter S32; if not, output the personalized evaluation content and enter S35;

[0043] S35: Determine whether there are unmarked homework. If so, enter S31; if not, enter S36;

[0044] S36: Batch release the personalized evaluation content, then enter S37;

[0045] S37: End.

[0046] From the above technical solutions, the present invention has the following beneficial effects:

[0047] 1. In the present invention, by combining the pre-trained language model, sentiment analysis, supervised learning and deep learning algorithms, the present invention can deeply understand and analyze the student's homework content, automatically identify grammar errors, logical problems and innovation points, and understand the teacher's language style through sentiment color analysis, so as to provide personalized modification suggestions for students. At the same time, the system can continuously optimize the evaluation model through the supervised learning algorithm to improve the evaluation accuracy and personalization degree. The system aims to improve the efficiency and quality of homework evaluation in college teaching and promote the development of education informatization and intelligence.

[0048] 2. In the present invention, especially in terms of the technical implementation of understanding and analyzing students' homework, analyzing teachers' language styles and emotions, training evaluation models, and identifying grammar errors, logical problems, and innovation points in homework, the efficiency and quality of subjective creative homework evaluation can be improved. This system can not only provide students with immediate personalized evaluations and optimization suggestions, but also provide teachers with an auxiliary tool for batch personalized marking of subjective creative homework, while cultivating students' innovative abilities and higher-order thinking. Through the comprehensive application of natural language processing technology, machine learning technology, and database technology, this system can achieve in-depth understanding and analysis of students' homework content, generate personalized marking opinions and modification suggestions, thereby significantly improving the efficiency and quality of evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is the framework flowchart of the present invention;

[0050] Figure 2 is the flowchart for constructing the student self-evaluation agent;

[0051] Figure 3 is the flowchart for constructing the teacher evaluation agent;

[0052] Figure 4 is the block diagram of the personal learning information library module;

[0053] Figure 5 is the block diagram of the evaluation criteria library module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0054] The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0055] To achieve the above object, the embodiments of the present invention adopt the following technical solutions: Refer to Figure 1 、 Figure 4 、 Figure 5 , the personalized intelligent evaluation system based on multi-role agents includes a data input module, a personal learning information library module, an evaluation criteria library module, an evaluation processing agent module, and a data output module:

[0056] Among them, the data input module is used to collect students' homework data, personal learning information, and evaluation criteria, and the key parameters include the integrity, accuracy, and timeliness of the data;

[0057] The personal learning information repository module is used to store students' personal learning information, including the learning situation analysis unit, learning style and preference unit, learning needs and goals unit, and teacher annotation information unit. It should be noted that the learning style and preference, as well as the learning needs and goals in this personal learning information repository module, can be updated in real time by students according to actual needs. At the same time, the teacher annotation information can be updated in real time according to the actual needs of teachers, and the learning situation analysis can be updated regularly according to the needs of teachers and students. The key parameters are the detail level and update frequency of the information;

[0058] The evaluation standard library module is used to store the evaluation standards and evaluation cases preset by teachers, including key evaluation indicators and expected language styles. The key parameters are the coverage range and specificity of the evaluation indicators;

[0059] The evaluation processing agent module includes a student self-evaluation agent and a teacher evaluation agent. The evaluation processing agent module is used to call personal learning information and the evaluation standards preset by teachers to form personalized evaluation indicators, and then generate review opinions and modification suggestions. The key parameters include the accuracy of the algorithm, response time, and degree of personalization;

[0060] The data output module is used to display review opinions and modification suggestions to students and teachers. The key parameter is the clarity and understandability of the output.

[0061] As a preferred technical solution of this embodiment, the personalized intelligent evaluation system based on multi-role agents further includes a data feedback module for collecting feedback information from students and teachers. The data feedback module includes a student self-evaluation function terminal and a teacher batch review terminal. Among them, the student self-evaluation function terminal collects the interactive opinions of students and review results and feeds them back into the personalized evaluation indicators to update and optimize the personalized evaluation indicators. The teacher batch review terminal collects the interactive opinions of teachers on the effect of batch review and feeds them back into the personalized evaluation indicators to update and optimize the personalized evaluation indicators.

[0062] Furthermore, the learning situation analysis unit includes a prerequisite course sub-unit, a learning background sub-unit, a knowledge reserve sub-unit, and a skill mastery sub-unit. Among them, the prerequisite course sub-unit stores the prerequisite course information of students, the learning background sub-unit stores the learning background materials of students, the knowledge reserve sub-unit stores the knowledge reserve of students, and the skill mastery sub-unit stores the skill mastery situation of students.

[0063] Further, the learning style and preference unit includes a cognitive style subunit, a learning method preference subunit, an example preference subunit, an interest area subunit, a social media behavior subunit, a feedback preference subunit, and a mental state and emotion change subunit. Among them, the cognitive style subunit stores the cognitive style of the student, the learning method preference subunit stores the relevant learning methods and preferences of the student, the example preference subunit stores the example preferences of the student, the interest area subunit stores the interest areas of the student, the social media behavior subunit stores the social media situation of the student, the feedback preference subunit stores the learning feedback of the student, and the mental state and emotion change subunit stores the mental state and emotion changes of the student.

[0064] Further, the learning need and goal unit includes a career planning subunit, an ability development goal subunit, a knowledge structure weakness subunit, and a personal development plan subunit. Among them, the career planning subunit stores the career planning of the student, the ability development goal subunit stores the ability development goals of the student, the knowledge structure weakness subunit stores the knowledge structure weaknesses in the student's learning process, and the personal development plan subunit stores the personal development plan of the student.

[0065] Further, the teacher annotation information unit includes a classroom performance subunit, a personality trait subunit, and an emotional color subunit. The classroom performance subunit stores the classroom performance information of the student annotated by the teacher, the personality trait subunit stores the information about the student's personality traits, and the emotional color subunit stores the emotional color information of the teacher.

[0066] Further, the evaluation standard library module includes a number of assignment units, and each assignment unit includes an evaluation case subunit and an evaluation standard subunit. Among them, the evaluation standard subunit includes evaluation indicators and evaluation weights.

[0067] When in use, data collection, data processing, data output, and data feedback are carried out respectively. Among them, data collection includes:

[0068] Collecting students' assignment data through the data input module, including planning schemes, creative writing, speech drafts, etc.;

[0069] At the same time, collecting students' personal learning information by using the students' personal learning information library module, including learning preferences, specific learning needs, etc.;

[0070] Collecting evaluation criteria, language styles, and emotional colors through the evaluation standard library module.

[0071] Data processing includes:

[0072] Call the student's personal learning information database and evaluation indicators, adjust the guidance content, list cases and presentation styles in combination with the student's learning preferences and specific needs, use natural language processing technology to understand and analyze the student's homework content, and use machine learning technology to train the evaluation model to improve the evaluation accuracy.

[0073] The data output includes:

[0074] The student self-evaluation agent outputs evaluation opinions, modification suggestions and optimization guidance; the teacher evaluation agent outputs batch marking opinions and guidance suggestions with emotional colors.

[0075] The data feedback includes:

[0076] The student self-evaluation function terminal collects the interactive opinions of the student and the marking results and feeds them back into the personalized evaluation indicators to update and optimize the personalized evaluation indicators; the teacher batch marking terminal collects the interactive opinions of the teacher on the batch marking effect and feeds them back into the personalized evaluation indicators to update and optimize the personalized evaluation indicators.

[0077] This embodiment involves four key technical fields, namely: understanding and analyzing the student's homework content, analyzing the teacher's language style and emotion, training the evaluation model with supervised learning algorithms, and identifying grammar errors, logical problems and innovation points in the homework with deep learning algorithms.

[0078] In terms of understanding and analyzing the student's homework content, the key technology lies in how to accurately understand the student's homework content and automatically extract relevant information for intelligent evaluation according to the specific requirements and knowledge points of the homework. This embodiment realizes the in-depth understanding of the homework content by combining the pre-trained language model and natural language processing (NLP) technology, and uses the following technologies:

[0079] Homework understanding based on pre-trained language models such as BERT: By fine-tuning the pre-trained language model, the model can deeply understand the text structure, semantics and context in the student's homework, automatically extract key knowledge points, and identify whether the student has fully expressed the core idea of a specific topic or problem.

[0080] Topic model and knowledge point extraction: Through topic models (such as LDA) or autoencoders (AE) in deep learning, the system can extract the main topics in the student's homework, judge whether the student has covered the core knowledge points in the relevant field, and perform a relevance score.

[0081] Deep semantic analysis: Use deep neural networks (DNN) to perform semantic-level understanding, analyze the logical structure and argumentation process in the student's homework, and identify whether there is insufficient information or unclear expression.

[0082] In the aspect of teacher language style and sentiment analysis, in this embodiment, the teacher's evaluation language style and sentiment analysis are crucial for the intelligent evaluation process. Sentiment analysis not only helps the system understand the teacher's evaluation tendency and sentiment color, but also enables the system to provide targeted feedback to students. This technical module includes the following content:

[0083] Understanding of sentiment color and teacher language style: Through sentiment analysis algorithms, the system can identify the sentiment color (such as positive, negative, neutral, etc.) and language style (such as rigorous, encouraging, critical, etc.) in the teacher's evaluation, and generate feedback suggestions suitable for the student's situation.

[0084] Deep learning-based sentiment analysis: Use deep learning models such as LSTM (Long Short-Term Memory Network) and BERT for sentiment analysis, understand the subtle sentiment fluctuations in the teacher's language, and identify the teacher's language style through sentiment classification algorithms. This process helps the system automatically identify the sentiment tone of the teacher's evaluation, so as to provide sentiment-matched feedback to students.

[0085] Analysis of teacher feedback sentiment tendency and feedback optimization: The system can analyze the teacher's feedback sentiment tendency, combine the student's personalized needs and emotional state, and provide feedback that meets the student's learning emotional needs, increasing the acceptance of the evaluation and the teaching effect.

[0086] In terms of training the evaluation model with supervised learning algorithms, in order to ensure the accuracy and effectiveness of the intelligent evaluation system, supervised learning algorithms are used to train the evaluation model. In this process, the model will learn how to accurately score students' homework based on a large amount of labeled data and optimize itself. The specific content includes:

[0087] Supervised learning algorithms and data annotation: The system collects and annotates a large number of student homework samples. The annotated data includes information such as grammar correctness, logical rigor, and creative performance. Through these annotated data, the system can learn how to score students' homework according to each evaluation dimension and provide feedback to students.

[0088] Model training and optimization: Common supervised learning algorithms (such as Support Vector Machine SVM, Random Forest, etc.) are used to train the evaluation model. During the model training process, by continuously adjusting parameters and training strategies, the evaluation accuracy and generalization ability of the model are improved.

[0089] Multi-task learning and model fusion: In order to make the evaluation model have stronger capabilities, this embodiment adopts multi-task learning methods to jointly train multiple tasks such as grammar checking, logical analysis, and creative evaluation, improving the overall performance of the model. At the same time, model fusion technologies such as ensemble learning are used to ensure the stability and high accuracy of the evaluation results.

[0090] In the aspect of identifying grammar errors, logical problems, and innovative points in deep learning algorithm recognition tasks, the identification of grammar errors, logical problems, and innovative points is one of the core contents of student assignment evaluation. Deep learning algorithms, especially natural language processing technologies based on neural networks, can efficiently identify these problems and provide feedback. The specific implementation is as follows:

[0091] Identification of grammar errors:

[0092] Grammar error detection based on the Transformer model: Using Transformer architecture models such as BERT and T5 for grammar error detection and correction, it can identify tense errors, subject-verb agreement errors, punctuation errors, etc. in grammar and provide corresponding modification suggestions.

[0093] Context-aware grammar analysis: Through the context-aware model, ensure that grammar error detection is not limited to a single sentence, but can also perform long-distance dependency analysis across sentences to avoid misjudgment.

[0094] Identification of logical problems:

[0095] Natural Language Inference (NLI) task model: Based on the Natural Language Inference (NLI) model, the system can judge whether the logical relationships in the assignment are reasonable and identify illogical parts, such as reasoning errors and contradictory viewpoints.

[0096] Dependency analysis: By analyzing the syntactic and semantic relationships between sentences in the text, identify logical loopholes in students' assignments to ensure the rationality of the reasoning chain in the assignment.

[0097] Identification of innovative points:

[0098] Text generation model and similarity analysis: Using text generation models such as GPT, by comparing with a large number of literature or academic articles, identify novel viewpoints and innovative content in students' assignments.

[0099] Deep semantic matching and innovation evaluation: The system evaluates the innovation in students' assignments through deep semantic matching technology, identifies creative solutions or unique thinking patterns, and gives personalized feedback.

[0100] Model training and optimization:

[0101] Multi-task learning and transfer learning: Train grammar error identification, logical problem analysis, and innovation point evaluation tasks through multi-task learning methods to achieve joint optimization of multiple tasks. At the same time, through transfer learning methods, ensure that the model can quickly adapt to specific tasks in the education field.

[0102] It should be noted that in terms of the complexity of the technical framework and data requirements, although this system involves multiple data processing and model training processes, in practical applications, its complexity and data requirements are significantly different from those of traditional large-scale data processing systems. Specifically, the application scenarios of this system are mainly for the personalized assignment evaluation of college teachers and students. In particular, teachers can customize and apply it according to the specific requirements of the courses they teach.

[0103] Data processing and computing resource requirements: When teachers use this system, it generally only involves several classes or a small range of student groups of the courses they teach. Therefore, the amount of data that the system needs to process is relatively small. Teachers only need to provide assignment demonstration templates, evaluation criteria, and the basic information of students, rather than a large amount of data. During the daily use process by teachers, there is no need for a large amount of labeled data and computing resources, which greatly reduces the implementation complexity of the system.

[0104] Personalized learning information library and real-time feedback: The personalized learning needs of students can be achieved by updating the personal learning information library in real time, and this process is adjusted based on students' self-evaluation and teachers' feedback. The learning data and evaluation criteria of each student are closely related to the specific courses they are in. Therefore, the system does not need to process all the data and can accurately match the specific class and course requirements.

[0105] Teacher-led evaluation preset: Since teachers can flexibly set and adjust evaluation criteria according to the content and objectives of their teaching, it ensures that the system only provides feedback and evaluation based on a small number of criteria preset by teachers. This method avoids relying on too much labeled data and also reduces the computational pressure for model training and optimization.

[0106] In summary, the technical solution of this embodiment does not require the use of a large amount of data and high computing resources in each link. Instead, in practical applications, the goals of homework correction and evaluation are efficiently achieved through teachers' personalized settings and students' real-time feedback. The complexity and data requirements of the system are mainly limited to the scope of specific courses and classes, ensuring its efficiency and practicality.

[0107] Referring to Figure 1 , this embodiment also provides a usage method of the personalized intelligent evaluation system based on the aforementioned multi-role agents, including the following steps:

[0108] S0: Start.

[0109] Specifically, this step is the starting point of the process.

[0110] S1: Submit an assignment. Whether the student self-evaluates. If so, go to S2; if not, go to S3.

[0111] Specifically, students submit their assignments. Before submitting the assignments to the teacher for grading, students can choose whether to use the student self-assessment agent to self-assess and optimize their assignments. If they choose self-assessment, they enter S2; if they do not choose self-assessment, they enter S3.

[0112] S2: The student self-assessment agent calls the evaluation criteria and personal learning information to generate modification suggestions, and determines whether the student asks questions about the modification suggestions. If so, repeat S2; if not, enter S1.

[0113] Specifically, the student self-assessment agent calls the evaluation criteria and the student's personal learning information to generate modification suggestions. Subsequently, the student can choose whether to ask questions about the modification suggestions. If the student chooses to ask questions, repeat S2; if the student does not choose to ask questions, the student optimizes the assignment according to the agent's grading opinions and modification suggestions, submits it again, and enters S1, and then enters S3.

[0114] S3: The teacher evaluation agent generates grading opinions and determines whether the teacher modifies the grading opinions. If so, enter S4; if not, enter S5.

[0115] Specifically, the teacher evaluation agent reviews the grading results and decides whether to be satisfied. If satisfied, the grading opinions are not modified, and the final grading opinions are feedback to the student, and then enter S5; if not satisfied, the grading opinions are modified, opinions are put forward and reprocessed, and enter S4.

[0116] S4: The teacher marks the modification requirements and gives feedback, and then enters S3 again.

[0117] Specifically, when the teacher needs to modify the grading opinions, mark the modification requirements and feedback them to the teacher evaluation agent, and then enter S3 again.

[0118] S5: Release the grading opinions, and then enter S6.

[0119] Specifically, feedback the final grading opinions to the student and then enter the end process.

[0120] S6: End.

[0121] Specifically, this step is the end point of the process.

[0122] It should be noted that when the student asks questions about the modification suggestions, the system will collect the student's question information. At the same time, when the teacher modifies the grading opinions, the system will collect the teacher's modification requirements. In this way, the system can feedback the collected question information and modification requirements to the corresponding agents to update the personalized evaluation indicators.

[0123] Refer to Figure 2 , furthermore, the described usage method further includes the step of constructing a student self-assessment agent, including the following steps:

[0124] S20: Start.

[0125] Specifically, this step is the starting point of the process.

[0126] S21: Identify the assignment format and clean the data, then proceed to S22.

[0127] Specifically, identify the format of the assignment and clean the data to ensure the accuracy and availability of the data, and then proceed to S22.

[0128] S22: Call the preset evaluation criteria and the student's personal learning information to form personalized evaluation indicators, then proceed to S23.

[0129] Specifically, call the preset evaluation criteria to evaluate the assignment, and call the student's personal learning information for personalized evaluation. Then, form personalized evaluation indicators based on the student's personal learning information and proceed to S23. It should be noted that the student's personal learning information can be updated in real time.

[0130] S23: Call the large model and the preset case library to generate personalized evaluation content and output it, then proceed to S24.

[0131] Specifically, call the large model to assist in the evaluation process, and refer to the preset case library to enrich the evaluation content. Generate personalized evaluation content for the student according to the above steps and output it, and then proceed to S24.

[0132] S24: Does the student have follow-up information about the evaluation content? If yes, proceed to S22; if no, proceed to S25.

[0133] Specifically, the student can ask questions or make further inquiries about the evaluation content. If the student has follow-up information, integrate the follow-up information into the personalized evaluation indicators to update and generate personalized evaluation indicators, then proceed to S22 to continue providing personalized evaluation content, and then proceed to S25. If the student has no follow-up information, the process ends and proceeds to S25.

[0134] S25: End.

[0135] Specifically, this step is the end point of the process.

[0136] Refer to Figure 3 , in some other embodiments, the usage method further includes the step of constructing a teacher evaluation intelligent agent, including the following steps:

[0137] S30: Start.

[0138] Specifically, this step is the starting point of the process.

[0139] S31: Identify the homework format and clean the data, then proceed to S32.

[0140] Specifically, identify the format of the homework and clean the data to ensure the accuracy and availability of the data, and then proceed to S32.

[0141] S32: Call the preset evaluation criteria and the student's personal learning information to form personalized evaluation indicators, then proceed to S33.

[0142] Specifically, call the preset evaluation criteria to evaluate the homework. At the same time, obtain the student's personal learning information for personalized evaluation, and then proceed to S33. It should be noted that the student's personal learning information can be updated in real time.

[0143] S33: Call the large model and the preset case library to generate personalized evaluation content, then proceed to S34.

[0144] Specifically, call the large model to assist in the evaluation process, and at the same time refer to the preset case library to enrich the evaluation content. Form personalized evaluation content according to the above steps, and then proceed to S34.

[0145] S34: Determine whether teacher's modification opinions are received. If so, proceed to S32; if not, output the personalized evaluation content and proceed to S35.

[0146] Specifically, check whether teacher's modification opinions on the evaluation content are received. If teacher's modification opinions are received, proceed to S32, adjust the evaluation content according to the teacher's modification opinions, and output the final personalized evaluation content; if teacher's modification opinions are not received, skip the modification step, output the personalized evaluation content, and proceed to S35.

[0147] S35: Determine whether there are unmarked homework. If so, proceed to S31; if not, proceed to S36.

[0148] Specifically, check whether there is still homework to be marked and whether the personalized evaluation content has been approved for release. If there is still homework to be marked, continue to mark the next new homework and proceed to S31. If there is no homework to be marked, proceed to S36.

[0149] S36: Batch release the personalized evaluation content and proceed to S37.

[0150] Specifically, batch release the approved personalized evaluation content to the students and proceed to S37.

[0151] S37: End.

[0152] Specifically, this step is the end of the process.

[0153] The following uses specific embodiments to verify the beneficial effects of the present invention:

[0154] Scenario setting: In the "Conference Planning and Organization" course at a certain university, students are required to submit a conference planning plan. The teacher hopes to improve the efficiency and quality of assignment evaluation through this system.

[0155] Step 1: Data input:

[0156] Students submit the conference planning plan through the personalized intelligent evaluation system based on multi-role agents.

[0157] Step 2: Data processing:

[0158] The evaluation processing agent module calls the personal learning information library module and the evaluation standard library module to generate personalized evaluation indicators:

[0159] (1). Call the personal learning information of the student, including previous application writing scores, preferred case types, teaching styles with high acceptance, and personal characteristics and applicable guiding emotional colors marked by the teacher through classroom observation;

[0160] (2). Call the preset evaluation standards, such as key evaluation indicators, weights, and corresponding evaluation cases for the structural standards of the conference name, planning structure standards, etc. in conference planning.

[0161] Using natural language processing technology, the agent analyzes the content of the student's planning plan and identifies key parts such as goal setting, innovation points, activity process, budget allocation, etc. in the plan.

[0162] The machine learning model generates personalized review opinions and modification suggestions according to the personalized evaluation indicators.

[0163] Step 3: Data output:

[0164] Student self-evaluation function terminal: The system shows the review opinions to the student, points out the innovative points and existing problems in the plan, and gives specific modification suggestions, such as specific operation ideas and cases for optimizing budget allocation or enhancing activity interactivity.

[0165] Teacher batch review terminal: The system provides batch processing review opinions for the teacher and adjusts the tone of the comments according to the emotional color set by the teacher to make the feedback more encouraging or challenging.

[0166] Step 4: Data feedback:

[0167] The student modifies the planning plan according to the review opinions provided by the agent and resubmits it.

[0168] The teacher reviews the review results provided by the agent. If satisfied, it is directly feedback to the student; if not satisfied, the teacher puts forward modification opinions, and the agent adjusts the review opinions according to the teacher's feedback.

[0169] Step 5: Feedback Information Collection:

[0170] Feedback information from students and teachers is collected and used to update personalized evaluation indicators to optimize subsequent personalized evaluation content.

[0171] To implement a personalized and intelligent subjective creative assignment evaluation method and system, the present invention proposes a data foundation centered around the student's personal learning information database and the evaluation standard database, and designs a simple and efficient collection and update mechanism to ensure that data can be flexibly entered and updated in real time to meet the needs of personalized evaluation.

[0172] Among them, the specific implementation of the information collection and update mechanism is as follows:

[0173] 1. Construction and Update of the Student's Personal Learning Information Database

[0174] 1.1 Design of the Information Collection Interface

[0175] The present invention adopts a modular information collection interface, which divides the student's personal information into the following core modules:

[0176] Basic Information Module: Contains the learner's basic information;

[0177] Learning Style Module: Records personal learning preferences and methods;

[0178] Learning Goal Module: Sets short-term and long-term learning goals;

[0179] Personality Trait Module: Displays personality characteristics and emotional tendencies;

[0180] Classroom Performance Module: Records interaction participation and performance.

[0181] 1.2 Information Collection Methods

[0182] (1) Initial Information Collection

[0183] Adopt an intelligent questionnaire form, and design question types that combine multiple-choice questions, rating scale questions, and open-ended questions;

[0184] The questionnaire is filled out step by step to avoid filling in too much information at one time;

[0185] Set a progress saving function to support completion in multiple sessions;

[0186] The questionnaire results are automatically converted into structured data and stored in the personal information database.

[0187] (2) Information Update Mechanism

[0188] Provide a quick update entry, and students can modify their personal information at any time;

[0189] Set up a regular reminder function to recommend students to update their personal development status;

[0190] Support partial information update without having to fill in all the content again;

[0191] Record the history of information changes to form an individual growth trajectory;

[0192] Establishment and maintenance of an evaluation criteria library.

[0193] 2.1 Standard library architecture design

[0194] Store evaluation criteria by assignment type;

[0195] Support setting multi-dimensional evaluation indicators;

[0196] Allow setting weights for evaluation elements;

[0197] Provide a scoring criteria description template.

[0198] 2.2 Standard setting and update methods

[0199] (1) Initial standard setting

[0200] Support batch import of existing evaluation criteria;

[0201] Provide an evaluation criteria template library;

[0202] Set up a standard visual editing interface;

[0203] Support import and conversion of multiple file formats.

[0204] (2) Standard update mechanism

[0205] Real-time editing function to support adjusting evaluation criteria at any time; version control function to record the history of standard modifications;

[0206] Support the trial use of evaluation criteria and feedback on the effects;

[0207] Provide a channel for collecting suggestions for standard modifications.

[0208] 3. System interaction characteristics

[0209] 3.1 User-friendliness design adopts an intuitive graphical operation interface;

[0210] Set up an intelligent guidance function;

[0211] Provide operation tips and help documents;

[0212] Support multi-terminal access (PC, mobile).

[0213] 3.2 The data synchronization mechanism adopts real-time synchronization technology to ensure data timeliness; an automatic save function is set to prevent data loss;

[0214] A data backup and recovery function is provided;

[0215] Offline operation is supported, and automatic synchronization occurs after connecting to the network.

[0216] 4. Information security guarantee

[0217] A multi-level permission management mechanism is adopted;

[0218] Data is encrypted and stored;

[0219] An access control policy is set;

[0220] A data protection mechanism is established.

[0221] 5. System scalability

[0222] Interfaces for third-party systems are reserved;

[0223] Support for adding new evaluation dimensions is provided;

[0224] Customizable information collection modules are allowed;

[0225] A data export and analysis function is available.

[0226] Through the above design, the present invention realizes the simple collection and real-time update of students' personal information and evaluation criteria, providing a reliable data basis for subsequent intelligent evaluation. The modular design of the system and the friendly interaction interface ensure the convenience and operability of the data collection process, and at the same time, through multiple technical means, the timeliness and security of the data are guaranteed.

[0227] Implementation effect:

[0228] Through this embodiment, it can be seen how the intelligent evaluation system helps students self-evaluate and optimize their plans and teachers improve the evaluation efficiency and quality of meeting planning plans. Students receive immediate personalized feedback, while teachers improve their grading efficiency, and at the same time, the feedback mechanism of the system ensures the continuous optimization and personalization of the evaluation criteria.

[0229] The innovation points of the present invention compared with the prior art are analyzed as follows:

[0230] 1. Application of multi-role agents:

[0231] Utilize the collaborative work of multi-role agents, combined with technologies such as pre-trained language models, sentiment analysis, and deep learning algorithms, to achieve comprehensive intelligent analysis and evaluation of students' homework.

[0232] 2. Multi-dimensional in-depth understanding of students' homework:

[0233] In terms of understanding and analyzing the content of homework, a method combining deep semantic analysis and topic models is adopted to achieve multi-dimensional and in-depth understanding of students' homework.

[0234] 3. Emotional color evaluation:

[0235] In terms of the teacher's language style and sentiment analysis, an emotional analysis algorithm is innovatively applied to identify the emotional tendency of the teacher's evaluation language, so as to achieve more personalized and accurate feedback for students.

[0236] 4. Self-optimization and update:

[0237] By combining multi-task learning and supervised learning algorithms, the evaluation model can not only accurately analyze grammar errors, logical problems, and innovation points, but also achieve self-optimization and update. For example, the learning styles and preferences, learning needs and goals in the information library can be updated in real time according to the actual needs of students; the teacher's marked information can be updated in real time according to the actual needs of teachers; the analysis of the learning situation can be updated regularly according to the needs of teachers and students, improving the flexibility and adaptability of the evaluation and the degree of personalized accuracy.

[0238] Compared with the existing technologies:

[0239] In the existing field of intelligent homework correction, the two most common learning platforms or intelligent tools are "Learning Pass" and "Grammarly". These tools have made certain progress in intelligent homework correction, but compared with the present invention, there are significant limitations, especially in dealing with subjective creative homework and personalized evaluation.

[0240] In terms of Learning Pass, its functions and features are as follows: Learning Pass is a comprehensive online learning platform that supports functions such as homework submission, correction, and examination. It can handle the automatic correction of objective question types, such as multiple-choice questions and fill-in-the-blank questions, but for subjective creative homework (such as planning schemes, creative writing, etc.), it still mainly relies on manual marking by teachers. Learning Pass provides some basic homework correction functions, such as grammar checking and word count, but lacks in-depth understanding of the homework content and personalized feedback.

[0241] The differences between this technical solution and Learning Pass are as follows:

[0242] Processing of subjective creative homework: Learning Pass cannot perform automated and intelligent correction of subjective creative homework, while this technical solution can deeply understand and analyze the content of students' homework, automatically identify grammar errors, logical problems, and innovation points, and provide personalized modification suggestions by combining pre-trained language models, sentiment analysis, deep learning and other technologies.

[0243] Personalized feedback: Learning Pass lacks a personalized feedback mechanism and cannot generate accurate evaluations based on students' individual learning needs and teachers' emotional inclinations. In contrast, this technical solution can achieve fully personalized homework evaluations through a personalized learning information database and the analysis of teachers' emotional colors.

[0244] Collaborative work of multi-role agents: Learning Pass does not introduce the concept of multi-role agents. In this technical solution, through the collaborative work of student self-evaluation agents and teacher evaluation agents, the flexibility and personalization of evaluations are significantly improved.

[0245] Regarding Grammarly, its functions and features are as follows: Grammarly is a tool focused on grammar checking and writing optimization, widely used in the field of English writing. It can detect grammar errors, spelling mistakes, punctuation errors, etc., and provide modification suggestions. Grammarly also offers some suggestions on writing styles, such as tone adjustment and sentence structure optimization.

[0246] The differences between this technical solution and Grammarly are as follows:

[0247] In-depth understanding of homework content: Grammarly mainly focuses on the optimization of grammar and writing styles and cannot deeply understand the homework content. Especially for subjective creative assignments (such as planning proposals, creative writing, etc.), it cannot identify logical problems, innovative points, etc. In contrast, this technical solution can conduct multi-dimensional in-depth understanding of the homework content through technologies such as deep semantic analysis and topic models.

[0248] Personalized evaluation mechanism: Grammarly lacks a personalized evaluation mechanism and cannot generate accurate evaluations based on students' individual learning needs and teachers' emotional inclinations. In this technical solution, through a personalized learning information database and the analysis of teachers' emotional colors, fully personalized homework evaluations can be achieved.

[0249] Collaborative work of multi-role agents: Grammarly does not introduce the concept of multi-role agents. In this technical solution, through the collaborative work of student self-evaluation agents and teacher evaluation agents, the flexibility and personalization of evaluations are significantly improved.

[0250] Innovation of this technical solution in the personalized evaluation mechanism:

[0251] Compared with the intelligent correction tools on the current market, this technical solution has significant innovations, especially in the implementation of the personalized evaluation mechanism:

[0252] (1). Preset evaluation criteria: This technical solution allows course teachers to preset personalized evaluation criteria according to specific course content and teaching objectives. Teachers can flexibly adjust evaluation dimensions and weights according to the actual teaching requirements of the course. These criteria are not limited to the completion of assignment content, but also include subjective evaluation factors such as thinking depth and innovation. Different from other platforms that can only rely on fixed-rule evaluation methods, this technical solution can transform teachers' personalized teaching objectives into evaluation criteria to achieve precise course matching.

[0253] (2). Combination of dynamic adjustment and students' needs: Students can update their personal information databases (such as interests, learning progress, knowledge weak points, etc.) in real time according to their personalized needs and feedback this information into the assignment evaluation. The system intelligently adjusts evaluation criteria and marking strategies based on this feedback, enabling the assignment evaluation of each student to closely match their learning characteristics and needs. This dynamic and personalized adjustment mechanism cannot be provided by existing intelligent assignment marking tools and can effectively improve the accuracy and pertinence of assignment evaluation.

[0254] (3). Combination of intelligence and teacher guidance: In existing tools, intelligent marking is usually automated and relies on preset fixed rules, while this technical solution emphasizes the guiding role of teachers more. Teachers can intervene in the intelligent marking system at any time to adjust evaluation criteria or provide personalized feedback to students. This hybrid mode that combines artificial intelligence and teacher judgment makes intelligent marking more flexible and reliable.

[0255] In summary, this technical solution innovatively solves the problem that traditional intelligent marking tools cannot deeply process subjective creative assignments by combining personalized needs, course objectives, and intelligent marking, and provides a personalized learning experience closely related to students' individual needs.

[0256] Expected effects:

[0257] 1. Improve evaluation efficiency:

[0258] Improve the efficiency and quality of teachers' marking of subjective creative assignments through automated batch processing of the evaluation process.

[0259] 2. Enhance evaluation quality:

[0260] Provide accurate and personalized instant feedback through artificial intelligence technology, improve students' willingness to optimize independently and the frequency of human-computer interaction, and enhance the quality of subjective creative assignments.

[0261] 3. Cultivate innovation ability:

[0262] Improve students' autonomous learning ability through personalized and precise guidance, and cultivate students' optimization awareness and innovative spirit.

[0263] Application prospects:

[0264] 1. Market application prospects

[0265] Application in the education industry:

[0266] The present invention can be widely applied to the fields of higher education and K-12 education to improve the efficiency and quality of teaching evaluation. With the advancement of education informatization, the demand for efficient and personalized teaching evaluation tools is increasing day by day.

[0267] Enterprise training and assessment:

[0268] In the field of enterprise training, this system can be used for employee skill assessment and training effect evaluation, improving the personalization and pertinence of training.

[0269] Online learning and distance education platforms:

[0270] With the popularization of online education, this system can be integrated into various online learning platforms to provide instant feedback and evaluation services for distance learning.

[0271] Intelligent education solutions:

[0272] This system can be used as part of intelligent education solutions, combined with technologies such as intelligent classrooms and virtual teaching assistants, to provide comprehensive educational technology support.

[0273] 2. Commercial value

[0274] Improve teaching efficiency:

[0275] By automating the evaluation process, this system significantly reduces the workload of teachers, improves teaching efficiency, and thus reduces educational costs.

[0276] Personalized education services:

[0277] The personalized evaluation and feedback provided by this system can enhance the learning experience of students and improve learning effects, which has important commercial value for educational institutions.

[0278] Technical innovation advantages:

[0279] Utilizing the latest artificial intelligence technology, this system has obvious technical advantages in the market and can attract more educational institutions and enterprise customers.

[0280] Growing market demand:

[0281] With the continuous expansion of the artificial intelligence market, it is expected that in the next few years, the artificial intelligence market will maintain a high-speed growth trend, providing a broad market space for this system.

[0284] 3. Potential market value

[0285] According to market research, the market size of the artificial intelligence industry is expected to further expand, bringing huge potential market value to this system.

[0286] In summary, the present invention not only has broad application prospects in the field of education, but also has significant commercial value. It can meet the market's demand for efficient and personalized teaching evaluation tools and is expected to achieve rapid growth driven by artificial intelligence technology.

[0287] The above-described embodiments are only descriptions of the preferred embodiments of the present invention and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A personalized intelligent evaluation system based on multi-role agents, characterized by: include: A data input module, which is used to collect student homework data, personal learning information and evaluation criteria; A personal learning information library module, which is used to store students' personal learning information, including a learning situation analysis unit, a learning style and preference unit, a learning need and goal unit, and a teacher marking information unit; An evaluation standard library module, which is used to store evaluation standards and evaluation cases preset by teachers; An evaluation processing agent module, which includes a student self-evaluation agent and a teacher evaluation agent. The evaluation processing agent module is used to call personal learning information and evaluation criteria preset by the teacher to form personalized evaluation indicators, and then generate review opinions and modification suggestions; as well as A data output module is used to display review comments and modification suggestions to students and teachers.

2. The personalized intelligent evaluation system based on multi-role agents according to claim 1 is characterized in that: The personalized intelligent evaluation system based on multi-role intelligent agents also includes a data feedback module for collecting feedback information from students and teachers. The data feedback module includes a student self-evaluation function terminal and a teacher batch review terminal.

3. The personalized intelligence evaluation system based on multi-role agents according to claim 1 is characterized in that: The learning situation analysis unit includes a prerequisite course sub-unit, a learning background sub-unit, a knowledge reserve sub-unit and a skill mastery sub-unit.

4. The personalized intelligence evaluation system based on multi-role agents according to claim 1 is characterized in that: The learning style and preference unit includes a cognitive style subunit, a learning method preference subunit, an example preference subunit, an interest area subunit, a social media behavior subunit, a feedback preference subunit, and a psychological state and emotional change subunit.

5. The personalized intelligent evaluation system based on multi-role agents according to claim 1 is characterized in that: The learning needs and goals unit includes a career planning sub-unit, a capability development goal sub-unit, a knowledge architecture weakness sub-unit and a personal development plan sub-unit.

6. The personalized intelligent evaluation system based on multi-role agents according to claim 1 is characterized in that: The teacher marking information unit includes a classroom performance sub-unit, a personality trait sub-unit and an emotional color sub-unit.

7. The personalized intelligence evaluation system based on multi-role agents according to claim 1 is characterized in that: The evaluation standard library module includes a plurality of operation units, each of which includes an evaluation case sub-unit and an evaluation standard sub-unit, wherein the evaluation standard sub-unit includes an evaluation index and an evaluation weight.

8. The method for using the personalized wisdom evaluation system based on multi-role agents according to any one of claims 1 to 7, characterized in that: The following steps are involved: S0: Start; S1: Submit the homework, whether the student self-evaluates, if yes, then go to S2, if no, then go to S3; S2: The student self-assessment agent calls the evaluation criteria and personal learning information to generate modification suggestions and determines whether the student asks questions about the modification suggestions. If so, S2 is repeated; if not, S1 is entered. S3: The teacher evaluation agent generates review comments and determines whether the teacher modifies the review comments. If so, it goes to S4, if not, it goes to S5; S4: The teacher marks the revision requirements and gives feedback, and then enters S3 again; S5: Publish review comments and then proceed to S6; S6: End.

9. The method of use according to claim 8, characterized in that: The method of use also includes the step of constructing a student self-assessment agent, including the following steps: S20: Start; S21: job format recognition and data cleaning, then enter S22; S22: Calling preset evaluation criteria and students’ personal learning information to form personalized evaluation indicators, and then entering S23; S23: Call the large model and preset case library to generate and output personalized evaluation content, and enter S24; S24: Does the student have any further questions about the evaluation content? If yes, go to S22; if no, go to S25; S25: End.

10. The method of use according to claim 9, characterized in that: The method of use also includes the step of constructing a teacher evaluation agent, including the following steps: S30: Start; S31: job format recognition and data cleaning, then enter S32; S32: Calling preset evaluation criteria and student personal learning information to form personalized evaluation indicators, and entering S33; S33: Call the large model and preset case library to generate personalized evaluation content and enter S34; S34: Determine whether the teacher's revision comments are received, if yes, proceed to S32, if no, output the personalized evaluation content and proceed to S35; S35: Determine whether there is any unchecked homework, if yes, go to S31, if no, go to S36; S36: Batch publish personalized evaluation content and enter S37; S37: End.

Images