CN-121981090-A - Composition intelligent correction and group management method and system based on large language model

CN121981090ACN 121981090 ACN121981090 ACN 121981090ACN-121981090-A

Abstract

The invention relates to the technical field of natural language processing and intelligent education, and discloses a composition intelligent correction and group management method and system based on a large language model, wherein the composition intelligent correction and group management method based on the large language model comprises the steps of carrying out multidimensional semantic deep analysis on an original composition text by adopting the large language model; extracting features and evaluating the capability of the comprehensive semantic analysis result; performing cluster analysis and similarity calculation on the personalized student portraits to generate an optimized learning group configuration scheme; the method comprises the steps of carrying out strategy matching processing to generate a differential correction strategy, carrying out comprehensive processing on the differential correction strategy and comprehensive semantic analysis results to generate a detailed correction feedback report, carrying out collaborative scheme design, organizing a group collaborative learning activity scheme, carrying out data statistics and effect analysis to generate a system performance evaluation report and an optimization improvement scheme, and solving the problems that the traditional composition correction efficiency is low and the personalized guidance and group management functions are lacked.

Inventors

ZHU XIWEI
YAN CHAO
Jiang Runnan
ZHAO PENGFEI
ZHU XILIN
HUANG LILI
FENG CHENG

Assignees

贵州中科恒运软件科技有限公司
贵州电子信息职业技术学院

Dates

Publication Date: 20260505
Application Date: 20260408

Claims (10)

1. The composition intelligent correction and group management method based on the large language model is characterized by comprising the following steps of: acquiring an original composition text submitted by a student, and carrying out multidimensional semantic deep analysis on the original composition text to generate a comprehensive semantic analysis result; carrying out feature extraction and capability assessment on the comprehensive semantic analysis result by adopting a statistical analysis method to construct a personalized student painting; Performing cluster analysis and similarity calculation on personalized student portraits, and generating an optimized learning group configuration scheme through capability level matching, learning requirement analysis and complementarity evaluation; Performing strategy matching processing on the learning group configuration scheme and the personalized student portraits, and generating a differential correction strategy according to different capability levels and learning characteristics; The method comprises the steps of comprehensively processing a differential correction strategy and a comprehensive semantic analysis result through a large language model, and generating a detailed correction feedback report through problem identification, improvement suggestion and personalized guidance; based on the optimized learning group configuration scheme and the detailed correction feedback report, carrying out collaborative scheme design, and organizing a group collaborative learning activity scheme through task allocation, mutual evaluation mechanism and collaborative guidance; And carrying out data statistics and effect analysis on the execution result of the group collaborative learning activity scheme to generate a system performance evaluation report and an optimization improvement scheme.
2. The method for intelligent composition correction and group management based on large language model as claimed in claim 1, wherein the step of generating comprehensive semantic analysis results comprises: Generating structured text data based on the original composition text, comparing standard grammar rules with context semantic recognition vocabulary by adopting a large language model, and obtaining grammar accuracy scores due to incorrect collocation, syntax structure problems and punctuation marks; Based on grammar accuracy scoring and structured text data, calculating semantic similarity between adjacent sentences, topic consistency between paragraphs and full-text logic fluency by adopting a large language model to obtain semantic consistency scoring; Based on the semantic consistency score, adopting a logic reasoning algorithm viewpoint, supporting discussion data and a reasoning process, and evaluating the demonstration integrity, convincing power and logic consistency to obtain a logic tightness score; Based on the logical tightness score and the original text content, the innovation evaluation algorithm is adopted to analyze the expression innovation degree of the text, and the innovation expression score is obtained by identifying novel views, unique expression modes, creative metaphors and original insights.
3. The method for intelligent composition correction and group management based on large language model as claimed in claim 1, wherein the step of constructing personalized student portraits comprises: Based on student historical composition data, extracting writing style characteristics, error mode characteristics, theme preference characteristics and capability change trend characteristics by adopting a statistical analysis method to obtain a historical characteristic vector; based on the historical feature vector and the comprehensive semantic analysis result, adopting a capacity assessment model to process historical data through time weighted average and perform weighted fusion with the current performance to obtain a capacity level index; based on the capability level index and the historical feature vector, generating a capability level label, a learning type label, a writing preference label and a development potential label by adopting a cluster analysis algorithm to obtain the personalized chemoportraits.
4. The method for intelligent composition correction and group management based on large language model as claimed in claim 1, wherein the step of generating optimized learning group configuration scheme comprises: Based on all personalized student images, adopting a multi-objective optimization theory to construct an objective function comprehensively considering the similarity of member capacities in groups, the difference between groups and the balance of group scales, and obtaining an optimization objective; based on the teaching actual demand and the system resource limitation, a constraint optimization method is adopted to set group scale limitation, capability distribution requirement and teacher distribution constraint, so as to obtain a constraint condition set; Searching an optimal group division scheme by adopting a genetic algorithm through population evolution, selection, crossing and mutation operations based on an optimization target and a constraint condition set to obtain initial group configuration; Based on initial group configuration and learning progress feedback, a dynamic adjustment algorithm is adopted to monitor learning effect differences in the groups and trigger the reassignment of group members according to a preset threshold value, so that a dynamically optimized group configuration scheme is obtained.
5. The method for intelligent composition correction and group management based on large language model as claimed in claim 1, wherein the step of generating the differential correction policy comprises: based on group configuration and member personalized student images, calculating average capability level, capability distribution characteristics, learning preference distribution and development potential of each group by adopting a statistical analysis method to obtain group feature vectors; Based on the group feature vector and a predefined strategy template library, adopting a pattern matching algorithm to select a basic enhancement type, a capability enhancement type or an innovative expansion type correction strategy template to obtain an initial strategy scheme; based on an initial strategy scheme and specific student personal portraits, analyzing the difference degree of the average characteristics of the personal portraits and the groups by adopting a strategy adjustment algorithm, and adjusting the important attention field, the feedback detail degree and the direction to obtain a personalized correction strategy; Based on the personalized correction strategy and the teaching schedule, a timing sequence planning algorithm is adopted to arrange correction time, feedback frequency and follow-up plan, so as to obtain the differential correction strategy.
6. The method for intelligent composition correction and group management based on large language model as claimed in claim 1, wherein the step of generating detailed correction feedback report comprises: Based on the personalized correction strategy and the semantic analysis result, a problem identification algorithm is adopted to locate specific problems in the text and to carry out classification marking according to the severity, the type and the improvement difficulty, so as to obtain a problem list; Based on the problem list and the language generation model, adopting a layered generation strategy to construct personalized feedback content comprising overall evaluation, specific problem indication, improvement suggestion and positive excellent expression; Based on text feedback content, adopting a quality evaluation and optimization algorithm to perform logic consistency check, expression definition evaluation, instructional value verification and emotion tendency adjustment; Based on the optimized feedback content, a visual feedback display interface comprising a layout template, color labels, chart elements and interactive functions is generated by adopting a visual technology, and a detailed correction feedback report is obtained.
7. The large language model based composition intelligent altering and group management method of claim 1, wherein the organizing group collaborative learning activity scheme comprises: Based on group configuration and member correction results, determining a cooperative task type, matching task difficulty, designing task flow, formulating evaluation standards and configuring resource support by adopting a task design algorithm according to the capability level and learning target of the group members to obtain a cooperative task scheme; Based on a collaborative task scheme, establishing a peer evaluation mechanism in a group by adopting a mutual evaluation algorithm, and establishing a mutual evaluation rule, designing an evaluation dimension, implementing anonymous allocation, setting cross verification, integrating feedback information and monitoring mutual evaluation quality; based on the group characteristics and the learning requirements, recommending proper learning resources for each group by adopting a recommendation algorithm, and generating personalized recommendation, evaluating recommendation quality and dynamically updating resources by identifying the learning requirements, constructing a resource library and matching the resource characteristics; based on the development condition of the collaborative learning activity and the participation feedback of students, the effect evaluation method is adopted to analyze the achievement of collaborative learning, and the group collaborative learning activity scheme is obtained by collecting quantitative data and qualitative data, calculating quantitative indexes, performing qualitative analysis, comprehensive effect evaluation, comparing analysis and verification and generating improved suggestions.
8. The method for intelligent composition correction and group management based on large language model according to claim 1, wherein the step of generating system performance evaluation report and optimizing improvement scheme comprises: Based on learning effect data and system operation indexes, a multidimensional analysis method is adopted to evaluate teaching effects, and an effect analysis report is generated by integrating multichannel data, analyzing student capacity improvement, evaluating system use satisfaction, analyzing teacher work efficiency and evaluating teaching quality improvement; Based on the effect analysis report, a problem diagnosis algorithm is adopted to identify the defects and problems in the system operation, and a problem diagnosis result is obtained by identifying abnormal data, comparing performance benchmarks, analyzing user feedback, mining system logs, positioning root causes and sequencing problem priorities; based on the problem diagnosis result, adopting an optimization decision algorithm to formulate a system improvement scheme, and obtaining an optimization improvement scheme through improvement target setting, solution design, resource constraint evaluation, scheme priority ordering, risk evaluation analysis and implementation plan formulation; Based on the optimization improvement scheme, a continuous improvement method is adopted to establish a self-adaptive optimization mechanism of the system, and a system performance evaluation report is obtained through feedback loop establishment, iterative optimization flow, self-adaptive adjustment mechanism, version management control, effect tracking evaluation and knowledge accumulation precipitation.
9. The intelligent composition correction and group management method based on a large language model according to claim 1, wherein the comprehensive processing of the differential correction strategy and the comprehensive semantic analysis result through the large language model comprises the steps of performing domain adaptability fine tuning training on a large-scale Chinese composition corpus through a pre-training language model based on a Transformer architecture by the large language model, wherein the method has grammar analysis, semantic understanding, logic reasoning and innovation evaluation capability; the capability assessment model adopts a multi-layer neural network structure, a quantitative assessment function of the writing capability of students is obtained through historical data training, and time sequence data can be processed and standardized capability indexes can be output; The genetic algorithm adopts a real number coding mode to represent a group division scheme, evaluates group configuration quality through a fitness function, and adopts tournament selection, single-point crossover and Gaussian variation operation to carry out population evolution.
10. The composition intelligent correction and group management system based on the large language model is characterized by being used for executing the composition intelligent correction and group management method based on the large language model as claimed in any one of claims 1-9, and comprising the following steps: the multi-dimensional semantic deep analysis module is used for acquiring an original composition text submitted by a student, carrying out multi-dimensional semantic deep analysis on the original composition text and generating a comprehensive semantic analysis result; the personalized student portrait construction module is used for carrying out feature extraction and capability assessment on the comprehensive semantic analysis result by adopting a statistical analysis method to construct a personalized student portrait; the intelligent group division module is used for carrying out cluster analysis and similarity calculation on personalized student portraits and generating an optimized learning group configuration scheme through capability level matching, learning requirement analysis and complementarity evaluation; the differential correction strategy generation module is used for carrying out strategy matching processing on the learning group configuration scheme and the personalized student portraits, and generating differential correction strategies according to different capability levels and learning characteristics; The intelligent correction feedback generation module is used for comprehensively processing the differential correction strategy and the comprehensive semantic analysis result through a large language model and generating a detailed correction feedback report through problem identification, improvement suggestion and personalized guidance; the group collaborative learning organization module is used for carrying out collaborative scheme design based on an optimized learning group configuration scheme and a detailed correction feedback report, and organizing a group collaborative learning activity scheme through task allocation, a mutual evaluation mechanism and collaborative guidance; And the teaching effect evaluation and optimization module is used for carrying out data statistics and effect analysis on the execution result of the group collaborative learning activity scheme to generate a system performance evaluation report and an optimization improvement scheme.

Description

Composition intelligent correction and group management method and system based on large language model Technical Field The invention relates to the technical field of natural language processing and intelligent education, in particular to a composition intelligent correction and group management method and system based on a large language model. Background With the deep development of education informatization, composition correction is an important link of Chinese teaching, and the traditional manual correction mode faces a plurality of challenges. In the middle school Chinese teaching practice, a teacher needs to process a large number of student compositions, the correction time of each composition needs 15-20 minutes on average, and the work load is large and the repeatability is high. Meanwhile, subjective differences exist in scoring standards of different teachers, and consistency and fairness of the evaluation results are difficult to ensure. The traditional correction mode also faces the problem of low feedback quality, and teachers can only give simple overall evaluation, so that detailed personalized guidance suggestions are difficult to provide for specific problems of each student. In recent years, with the rapid development of artificial intelligence technology, intelligent composition correction systems are gradually rising, and attempts are made to solve the problem of conventional correction by an automated technology. The existing intelligent correction system is mainly based on natural language processing technology, can identify grammar errors and structural problems to a certain extent, but has obvious defects in deep semantic understanding, innovative evaluation and personalized feedback. These systems often employ rule-driven or shallow machine learning methods, and it is difficult to understand the deep meaning and creative expression of the composition, resulting in a large gap between the evaluation result and the manual correction. Furthermore, the prior art has relatively weak research in group management and collaborative learning. Traditional teaching management mainly relies on experience and intuition of teachers to group students, and a scientific data support and an intelligent decision mechanism are lacked. The mutual evaluation and mutual assistance among students also lacks effective organization and guidance, and the educational value of the study of the peers is difficult to develop. Therefore, how to organically combine intelligent correction with group management to construct a complete intelligent teaching ecological system becomes an important subject of current education technology research. Disclosure of Invention The invention provides a composition intelligent correction and group management method and system based on a large language model, which solve the technical problems that detailed personalized guidance advice is difficult to provide for specific problems of each student and a scientific data support and intelligent decision mechanism is lacking in the related technology. The invention provides a composition intelligent correction and group management method based on a large language model, which comprises the following steps: acquiring an original composition text submitted by a student, and carrying out multidimensional semantic deep analysis on the original composition text to generate a comprehensive semantic analysis result; carrying out feature extraction and capability assessment on the comprehensive semantic analysis result by adopting a statistical analysis method to construct a personalized student painting; Performing cluster analysis and similarity calculation on personalized student portraits, and generating an optimized learning group configuration scheme through capability level matching, learning requirement analysis and complementarity evaluation; Performing strategy matching processing on the learning group configuration scheme and the personalized student portraits, and generating a differential correction strategy according to different capability levels and learning characteristics; The method comprises the steps of comprehensively processing a differential correction strategy and a comprehensive semantic analysis result through a large language model, and generating a detailed correction feedback report through problem identification, improvement suggestion and personalized guidance; based on the optimized learning group configuration scheme and the detailed correction feedback report, carrying out collaborative scheme design, and organizing a group collaborative learning activity scheme through task allocation, mutual evaluation mechanism and collaborative guidance; And carrying out data statistics and effect analysis on the execution result of the group collaborative learning activity scheme to generate a system performance evaluation report and an optimization improvement scheme. In a preferred embodiment, the step of generatin