Research On Knowledge Point Annotation Fusion Strategy Based On Collective Intelligence

In recent years,with the vigorous development of educational informatization,people’s educational needs are developing from standardized teaching to personalized learning,and personalized learning has become a research hotspot.Through online learning platform,it can help learners to realize personalized learning initially.However,marking the knowledge points of test questions provides support for learners to generate learning paths,personalize diagnosis reports and recommend personalized learning resources.Therefore,it is necessary to analyze the knowledge relevance between knowledge points and test resources,and mark the test resources with knowledge points,so as to protect the individualized learning of learners.Although the accuracy of traditional expert labeling is guaranteed,the implementation efficiency is low;Automatic association labeling based on semantic analysis and machine learning is efficient but not accurate.However,intelligent tagging is relatively efficient and convenient,but there are differences in attributes such as users’knowledge,ability and work attitude,which leads to some errors in the results.Therefore,how to combine the advantages of automatic tagging and intelligent tagging to realize the relationship between test resources and knowledge points is the focus of this study.In this paper,a knowledge point labeling fusion strategy based on collective intelligence is proposed,and the specific work is as follows:(1)automatically labeling the knowledge points of test text resources into text classification problems,analyzing the characteristics of physical texts and different deep learning models,and selecting Bi-LSTM network for processing.before processing,physical text test questions need to be stop words-removed and segmented,and finally,they are classified with Softmax to realize automatic labeling of test text resources.(2)Combing the whole process of collective intelligence labeling,designing a reasonable task allocation and reward mechanism for collective intelligence labeling,and promoting the good operation of the platform.This paper analyzes the influence of users’confidence on the results of intelligence fusion,improves the definition of confidence,and adds confidence threshold to filter users based on weighted voting method.Finally,the improved algorithm is used to calculate the intelligence tags,and obtains the associated labels of intelligence features.(3)The text feature base classifier is obtained by automatically labeling the test text.The public intelligence label is calculated by the user’s confidence,and the public intelligence-based classifier is obtained.This paper puts forward a fusion strategy of public intelligence labeling results based on combined multi-classifiers,and uses weighted voting method to process,so as to realize the final labeling of test resources and knowledge points.Based on the above strategy,the knowledge point annotation fusion module based on public intelligence is designed and implemented.Firstly,the overall planning of the module is completed;From the functional design,it mainly includes automatic labeling module,association labeling module of intelligent features,intelligent labeling result fusion module combined with multiple classifiers and user module,which improves the logical relationship between modules.Finally,the function and performance of the module are tested to ensure the stability of the system.