CN-122023072-A - Multi-dimensional guided teaching assistance method and system based on retrieval enhancement

CN122023072ACN 122023072 ACN122023072 ACN 122023072ACN-122023072-A

Abstract

The invention discloses a multi-dimensional guided teaching assistance method and system based on retrieval enhancement, wherein the method comprises the following steps: according to the method, firstly, through heuristic dialogue, student questions, answers, actual operation feedback and historical data are collected, cognitive weak points, learning requirements and problem suitability are identified through a multidimensional analysis model, and a personalized chemical emotion image is constructed. And customizing personalized content of the adaptive teaching scene by depending on the large model and the image, embedding a multi-source knowledge resource library tracing identifier, and ensuring traceability and accuracy of the content through cross verification. Through the visual carrier, multiple rounds of interaction are carried out by using a step-type question and scene example, dynamic feedback is collected, and teaching strategies are optimized. And finally, quantitatively evaluating the three-dimensional degree of the effectiveness of the guiding interaction according to the learning condition positioning accuracy, the content generation suitability, and giving out the comprehensive score of the learning result of the student, thereby improving the teaching quality and the learning efficiency.

Inventors

SHI ZHENHAO
XU WEI
ZHANG YAN
LI HU

Assignees

大唐互联科技(武汉)有限公司

Dates

Publication Date: 20260512
Application Date: 20251226

Claims (10)

1. A multi-dimensional guided teaching assistance method based on search enhancement, comprising: S1, acquiring questioning contents, answering results, real operation feedback and historical learning data input by students in a teaching scene based on a heuristic dialogue interaction mode, identifying the cognitive weak points, learning requirements and suitability of the questions and teaching subjects of the students through a multi-dimensional analysis model, constructing personalized chemical emotion images, and providing a targeting basis for subsequent teaching guidance; S2, customizing teaching related contents which adapt to different teaching scenes and meet personalized requirements based on the content generation capacity of the large model and combined with the constructed personalized chemistry condition file, embedding a source tracing identifier of a specific position where the name and core information of the related resource source library are located when the teaching related contents are generated, and ensuring traceability and accuracy of the generated contents through binding of the source tracing identifier and the generated contents and cross verification of multi-source knowledge; S3, outputting customized teaching contents to students through a visual interaction carrier, carrying out multi-round teaching interaction through guide logic of step-type question and scene example, simultaneously collecting dynamic feedback data of the students in the interaction process in real time, and dynamically optimizing a teaching strategy, a content difficulty gradient and a guide mode based on the data to adapt individual cognition difference of the students and real-time change of teaching scenes; s4, based on dynamic feedback data and personalized chemical emotion images in the teaching interaction process, carrying out quantitative evaluation on implementation effects of the current teaching process by adopting standardized evaluation indexes from three core dimensions of learning emotion positioning accuracy, content generation suitability and guiding interaction effectiveness, synthesizing calculation results of all evaluation indexes, and finally giving out comprehensive scores of learning results of the students.
2. The multi-dimensional guided teaching assistance method based on search enhancement according to claim 1, wherein the heuristic dialogue interaction method in S1 specifically comprises: Aiming at the information fed back by the students, a progressive guiding speaking operation is generated around the current teaching theme and the cognitive state of the students and is output to the students, so that the students are guided to deepen thinking and expression of knowledge gradually; In the interaction process, the follow-up feedback information of the student is continuously received, the guiding direction and the questioning logic are dynamically adjusted, the dialogue is ensured to focus on the teaching core target all the time, meanwhile, various expression data and behavior feedback data of the student in the interaction process are completely collected, and comprehensive support is provided for multidimensional analysis of the learning condition.
3. The multi-dimensional guided teaching assistance method based on search enhancement according to claim 1, wherein the multi-dimensional analysis model in S1 includes specific analysis modes including logic coherence analysis, term accuracy determination, query direction disassembly and language expression recognition, which are specifically respectively: The logic consistency analysis is used for carrying out semantic disassembly and logic carding on question text and answer expressions of students, judging the integrity of a student thinking chain by analyzing the relevance and connection rationality of core views and deduction links, identifying logic faults and associated deviations existing in the deduction process, and positioning weak attributes of the students on a logic deduction level; The term accuracy judgment is used for comparing subject related terms, concepts and rules related to student expressions with authoritative knowledge standards, analyzing whether the use of the subject related terms, concepts and rules meets the specification requirements, identifying understanding deviation and scene misuse in term application, and determining specific blind zone types of students in knowledge mastering level; The query direction disassembly is used for extracting core appeal information in student questions, combining teaching subjects and preset teaching targets, judging type attributes of the questions and suitability of the questions with teaching contents, and determining deviation characteristics and range attributes of the questions deviating from the teaching subjects or exceeding the preset teaching range; The language expression recognition is used for comprehensively analyzing vocabulary characteristics, sentence pattern structures and behavior feedback data in the interaction process in the student expression to judge the mastering proficiency and cognitive confidence of the students on related knowledge and assist in verifying the understanding state of the students in the corresponding knowledge field.
4. The multi-dimensional guided learning assistance method based on search enhancement according to claim 1, wherein the multi-source knowledge resource library in S2 includes a course standard library, a teaching material resource library, an experiment database, and a name school high-quality course library, which correspond to the following steps: The course standard library is used for storing official teaching instruction files of each school segment and each subject, and the file content clearly comprises teaching targets, knowledge capacity requirements, teaching content ranges and evaluation standards, so that compliance basis is provided for teaching content generation and knowledge accuracy verification; The teaching material resource library is used for recording the complete text content of each version of subject teaching material, the structured dismantling result of knowledge points, the detailed analysis of example questions and the exercise questions after class, completing the standardized storage and accurate adjustment of the teaching material knowledge and adapting to the basic teaching content generation requirement; the experiment database is used for integrating the standard operation flow, the core operation key points, the standard data parameters, the experimental phenomenon analysis and the clear result conclusion corresponding to each subject experiment, comprises the core parameters and the logic rules required by the construction of the virtual experiment scene, and supports the accurate generation of the experimental teaching content; the high-quality course library for the famous universities is used for converging high-quality teaching course resources of the famous universities, wherein the resource types comprise classroom teaching video clips, standardized teaching courseware, heavy difficulty special analysis, thematic lecture data and typical teaching cases, and high-quality and diversified teaching resource support is provided for personalized teaching.
5. The multi-dimensional guided teaching assistance method based on search enhancement of claim 1, wherein the generating method of the traceability identifier in S2 is as follows: Firstly, constructing a tracing identifier basic feature set, wherein the mathematical expression of the tracing identifier basic feature set is S base ＝{I lib ,P store ,K module , S base is the tracing identifier basic feature set, three types of core feature parameters including a resource source, a storage position and a knowledge module are integrated, I lib is resource source library identification information, P store is a specific position feature value of the core information, K module is a knowledge module feature value corresponding to teaching content, and the knowledge module feature value is extracted by a large model based on a cognitive weak point in a personalized chemistry figure; secondly, carrying out multidimensional information fusion coding based on the basic feature set to generate a unique traceability identification, wherein the mathematical expression is as follows: The method comprises the steps of obtaining a unique tracing identifier, wherein the ID trace is a unique tracing identifier which is finally generated, the H (-) is a preset hash coding function, and an input and output rule of the H (-) is adapted to a storage coding specification of a multi-source knowledge resource base; the method is used for carrying out layer-by-layer information fusion on the I lib ,P store ,K module in the basic feature set S base for bitwise exclusive OR operation; Thirdly, carrying out knowledge matching verification on the traceability identification and the teaching content to ensure suitability of the traceability information and the personalized teaching content, wherein the mathematical expression is V match ＝sign(|f(ID trace )-K module |, V match is a knowledge matching verification result, f (-) is a traceability identification feature extraction function and is used for reversely analyzing a knowledge module feature value from ID trace , sign (-) is a sign function, i|is absolute value operation, and when V match =0, the analyzed knowledge module feature value is consistent with the original K module ; finally, binding validity verification of the tracing identification and the teaching content is carried out, reliability of a tracing link is guaranteed, a mathematical expression of the tracing link is V bind ＝card(M(ID trace ,C teach ), V bind is a binding validity verification result, M (-) is a mapping relation set of the tracing identification and the teaching content, ID trace and a large model are stored to generate a corresponding relation of the teaching content C teach , card (-) is set base operation and is used for counting the number of elements in the mapping relation set, and when V bind =1, the tracing identification and the teaching content are bound in a one-to-one correspondence mode.
6. The multi-dimensional guided teaching assistance method based on search enhancement according to claim 1, wherein the dynamic feedback data in S3 includes answer response duration, operation path selection, and query focusing dimension, which are respectively: the answering response time length refers to a continuous time interval from the beginning of answering to the submitting of complete answers after the students receive answering tasks in the customized teaching contents, and is used for reflecting thinking time consumption and proficiency of the students on corresponding knowledge points; The operation path selection refers to the step sequence, the function button click sequence and the parameter setting logic which are selected when the students perform the actual operation in the visual interactive carrier, and is used for embodying understanding and executing thought of the students to the actual operation flow; The query focusing dimension refers to a knowledge module, a problem type and a research depth pointed by students when the students give follow-up questions in multiple rounds of teaching interaction, and is used for accurately capturing newly-increased cognitive demands and understanding stuck points of the students in the learning process.
7. The multi-dimensional guided teaching assistance method based on search enhancement according to claim 1, wherein the specific process of dynamically optimizing teaching strategies, content difficulty gradients and guiding modes based on the data in S3 is as follows: Firstly, constructing a dynamic feedback data feature set and completing quantization characterization, wherein the mathematical expression is S feedback ＝{T resp ,L op ,D query , S feedback is the dynamic feedback data feature set, three core features of answer response time length, operation path selection and query focusing dimension are integrated, T resp is an answer response time length quantization value, L op is an operation path matching degree quantization value, and D query is a query focusing adaptation degree quantization value; Secondly, calculating a comprehensive value of teaching fitness, and judging the matching degree of the current teaching setting and the learning state of the student, wherein the mathematical expression is F adapt ＝min{T resp /T std ,L op ,D query , F adapt is the comprehensive value of the teaching fitness, the range of the value is 0 to 1, the closer the value is 1, the more the current teaching strategy, the content difficulty gradient and the guiding mode are matched with the learning state of the student, T std is the standard answering time length of the corresponding knowledge point, T resp /T std is the answering response time length adapting ratio, min { and } is a minimum function, and the weakest adapting link in the current teaching setting is positioned by taking the minimum value of three types of characteristic parameters, so that a targeting basis is provided for targeted optimization; Thirdly, positioning an optimized direction based on the comprehensive value of the teaching fitness, wherein the mathematical expression is O dir ＝argmin{T resp /T std ,L op ,D query , O dir is an optimized direction identifier, indexes corresponding to three types of dynamic feedback characteristic parameters, argmin { and } is a minimum value index function, and the argmin { and } is used for determining a core characteristic parameter which causes insufficient teaching fitness, namely an optimized core direction; Finally, carrying out optimization effect verification, wherein the mathematical expression is V opt ＝|F′ adapt -F adapt I, and V opt is an optimization effect difference value and is used for representing the improvement amplitude of the teaching fitness after optimization; F' adapt is an optimized teaching adaptation degree comprehensive value, and is obtained by collecting dynamic feedback data of teaching interaction of the next round after optimization according to the teaching adaptation degree comprehensive value calculation formula; And when V opt ≥Δ std , judging that the optimization is effective, wherein delta std is a preset optimization effect threshold value, determining the analysis result of the multi-dimensional analysis model on the historical optimization data, and if V opt <Δ std , repeating the optimization process based on the new dynamic feedback data until the optimization effect reaches a preset standard.
8. The multi-dimensional guided teaching assistance method based on search enhancement according to claim 1, wherein the specific process of quantitatively evaluating the implementation effect of the current teaching process by using the standardized evaluation index in S4 is as follows: firstly, carrying out learning condition positioning accuracy index measurement, taking a cognitive weak point identified in a personalized chemical condition image as a reference, extracting a set of answering error points, real operation error points and doubtful focusing points in dynamic feedback data, and constructing a positioning fitness calculation model, wherein the expression is as follows: Wherein A loc is a learning condition positioning accuracy value, the value range is 0-1;S error is a problem set for representing a student knowledge short board in dynamic feedback data, S weak is a cognitive weak point set identified in a personalized chemical condition image, card (·) is set cardinal number operation for counting the element number in the set, S error ∩S weak is the intersection of two sets, namely the overlapping part of a student actual problem and an image positioning weak point, and the closer the A loc value is to 1, the higher the learning condition positioning accuracy is; The method comprises the steps of carrying out content generation suitability index measurement and calculation, taking a knowledge module characteristic value corresponding to teaching content and a student cognitive level characteristic value as core parameters, and constructing a suitability verification model, wherein an expression is A cont ＝sign(|K cont -K stu |), wherein A cont is a content generation suitability value, a value of 0 or 1;K cont is a knowledge module characteristic value of the teaching content, the knowledge module characteristic value is determined by a difficulty level of corresponding resources in a multi-source knowledge resource base, K stu is a student cognitive level characteristic value and is generated by analyzing a personality chemistry image, sign (-) is a sign function, i|is an absolute value operation, when I K cont -K stu |=0, A cont =1 indicates that the content is completely matched with the student cognitive level, and otherwise, the content suitability is 0, and the content is insufficient; Finally, carrying out guide interaction effectiveness index measurement and calculation, and constructing an effectiveness calculation model according to the change of answering response time of students before and after teaching interaction, wherein the expression is A inter ＝max{0,(T pre -T post )/T pre , A inter is a guide interaction effectiveness value, the value range is 0-1;T pre is the answering response time of first-round interaction, T post is the answering response time of last-round interaction, max { and } is a maximum function for limiting the index value to be nonnegative, and the closer the A inter value is to 1, the more remarkable the effect of the guide interaction on improving the answering efficiency of the students is shown; After three dimension index measurement is completed, substituting A loc 、A cont 、A inter into a set operation model S eval ＝{A loc ,A cont ,A inter , obtaining a final evaluation result by calculating the arithmetic mean value of set elements, and integrating various indexes to complete comprehensive scoring of the student learning result.
9. A multi-dimensional guided instruction assistance system based on search enhancement, comprising: The learning condition image construction unit is used for collecting question contents, answer results, actual operation feedback and historical learning data input by students in a teaching scene based on a heuristic dialogue interaction mode, identifying the cognitive weak points, learning requirements and suitability of the questions and teaching subjects of the students through a multi-dimensional analysis model, constructing personalized chemical condition images and providing a targeting basis for subsequent teaching guidance; The teaching content generation traceability unit is used for customizing and adapting to teaching related contents of different teaching scenes and meeting personalized requirements based on the content generation capacity of the large model and combining with the constructed personalized chemistry condition file, embedding the traceability identification of the specific position of the name and the core information of the related resource source library when generating the teaching related content, and ensuring traceability and accuracy of the generated content through binding of the traceability identification and the generated content and cross-validation of multi-source knowledge; The teaching strategy dynamic optimization unit is used for outputting customized teaching contents to students through the visual interaction carrier, carrying out multi-round teaching interaction through the guiding logic of the step-type question and the scene example, collecting dynamic feedback data of the students in the interaction process in real time, and dynamically optimizing the teaching strategy, the content difficulty gradient and the guiding mode based on the data to adapt individual cognition difference of the students and real-time change of the teaching scene; The teaching effect comprehensive scoring unit is used for quantitatively evaluating the implementation effect of the current teaching process by adopting standardized evaluation indexes based on dynamic feedback data and personalized chemical emotion images in the teaching interaction process from three core dimensions of learning emotion positioning accuracy, content generation suitability and guiding interaction effectiveness, synthesizing calculation results of all evaluation indexes, and finally giving out comprehensive scoring of the learning results of the students.
10. A computer-readable storage medium, on which a computer program is stored, characterized in that the computer program is executed by a processor for a search-enhancement-based multi-dimensional guided teaching assistance method according to any one of claims 1-8.

Description

Multi-dimensional guided teaching assistance method and system based on retrieval enhancement Technical Field The invention belongs to the technical field of intelligent education, and particularly relates to a multi-dimensional guided teaching assistance method and system based on retrieval enhancement. Background In the current teaching scene, the implementation of personalized teaching faces a plurality of reality barriers, and the traditional teaching mode is difficult to adapt to individual cognitive difference and various learning demands of students. Most teaching processes still use unified content output as a core, and a teacher is difficult to comprehensively capture knowledge mastering blind areas, thinking logic short boards and personalized learning demands of each student, so that teaching guidance is lack of pertinence, part of students are difficult to keep up with teaching rhythm because of inaccurate recognition of cognitive weak points, and the other part of students cannot be effectively promoted because of unmatched content difficulties. Meanwhile, obvious shortboards exist for generating and verifying teaching contents. The existing teaching content depends on fixed teaching materials or general resources, lacks of deep adaptation with specific students, is difficult to check in an effective mode in authority and accuracy of the content source, has the problems of nonstandard knowledge expression, low logic precision and the like, influences the teaching quality and also negatively influences the construction of a student knowledge system. In addition, interaction and feedback mechanisms in the teaching process are imperfect, mostly unidirectional content is infused, dynamic guiding interaction is lacked, dynamic feedback data such as a real-time learning state, operation deviation and new questions of students are difficult to collect and utilize in time, and the teaching strategy adjustment is lagged behind the change of the learning state of the students, so that an effective teaching closed loop cannot be formed. The evaluation mode of the teaching effect also has the problems of singleization and subjectivity, and depends on the result data such as examination results, and the like, lacks quantitative evaluation on core links such as learning condition positioning accuracy, content suitability, guidance interaction effectiveness, and the like, and is difficult to comprehensively and objectively reflect the actual effect of the teaching process, so that the teaching improvement lacks scientific basis. These problems not only restrict the improvement of teaching quality, but also influence the development of learning efficiency and comprehensive ability of students, so that a teaching auxiliary scheme capable of accurately grasping learning condition, generating reliable adaptation content, dynamically optimizing teaching process and scientifically evaluating effect is needed to solve the limitation of traditional teaching and promote the efficient landing of personalized teaching. Disclosure of Invention The invention aims to solve the problems of inaccurate positioning of the learning condition, insufficient content adaptability, feedback optimization lag and evaluation unilateral in the traditional teaching. The personalized chemical emotion image is built through heuristic interaction, traceable adaptive content is generated by means of a multi-source knowledge resource base, a teaching strategy is dynamically optimized, a multi-dimensional quantitative evaluation effect is achieved, an accurate, reliable and dynamic auxiliary scheme is provided for personalized teaching, and teaching quality and learning efficiency are improved. In view of the above-mentioned drawbacks or improvements of the prior art, as a first aspect of the present invention, the present invention provides a multi-dimensional guided teaching assistance method based on search enhancement, comprising: S1, acquiring questioning contents, answering results, real operation feedback and historical learning data input by students in a teaching scene based on a heuristic dialogue interaction mode, identifying the cognitive weak points, learning requirements and suitability of the questions and teaching subjects of the students through a multi-dimensional analysis model, constructing personalized chemical emotion images, and providing a targeting basis for subsequent teaching guidance; S2, customizing teaching related contents which adapt to different teaching scenes and meet personalized requirements based on the content generation capacity of the large model and combined with the constructed personalized chemistry condition file, embedding a source tracing identifier of a specific position where the name and core information of the related resource source library are located when the teaching related contents are generated, and ensuring traceability and accuracy of the generated contents through binding of the sour