Analysis And Application Research Of Massive Educational Video Materials Based On Cognitive Styles

With the continued deep integration of information technology and the education industry,online education is rapidly entering the public eye and becoming a rising star in the education sector.Compared to traditional classroom learning,online education retains a large number of educational videos,so that learners learn to break the time and space constraints,video learning not only has a greater impact on the learning effect of learners,but also through the learning process data left online after the learners watch the video,accurate discovery of problems in the learning process,and to provide them with guidance and advice,to achieve tailor-made teaching.However,it is a pressing problem for learners to quickly and accurately grasp the core content of educational videos and to find the most desired explanatory videos.Therefore,this thesis takes the educational video keyframe extraction technique as the research object and combines it with cognitive style theory to investigate the following aspects:1)To address the problem that it is difficult to extract keyframes from existing video keyframe extraction methods for educational videos with too single scenes,this thesis proposes an automatic keyframe extraction model for educational videos based on semantic analysis(KFEVSA).Since there is a lot of textual information in educational videos,this model integrates the TF-IDF algorithm in semantic text similarity calculation and the perceptual hashing algorithm to process text frames and text-free frames in educational videos separately.The model makes up for the shortcomings of traditional algorithms that only consider the underlying features of images to extract keyframes,and can greatly improve the accuracy of keyframe extraction in educational videos.In addition,when determining text similarity,the idea of using dynamic thresholds instead of static thresholds is proposed to reduce keyframe redundancy.2)Proposing an automatic classification model for educational videos based on cognitive styles EV-AC(Educational Videos in Automatic Classification),to improve the efficiency of learners in finding appropriate educational resources.The model combines cognitive style theory,text feature extraction,image feature extraction and video audio feature extraction for educational video keyframes,and uses the K-Nearest Neighbor(KNN)classification algorithm to classify video materials according to their cognitive style types.Using the EV-AC model,the process of classifying video material can be greatly simplified,which is important for reducing video retrieval time and improving learning outcomes for learners.3)An algorithm ASLM-DAD(Analyzing Student Learning Mastery based on Dynamic Assessment Data)is proposed to determine the effectiveness of learners’ video learning.To address the limitations of existing indicators for judging learners’ mastery,which are single and inaccurate,the ASLM-DAD algorithm takes full account of learners’ dynamic assessment process data(e.g.answer correctness,number of answer revisions,number of correct answer choices as a percentage of revision paths,average effective time for answering questions,etc.).The algorithm is more accurate and gives a true picture of the learner’s detailed mastery of the knowledge,thus providing a reference for the lecturer to give clear instructional advice.4)Based on the above research,this thesis designs a prototype system for keyframe extraction,classification,evaluation and analysis of educational video combined with massive educational video and dynamic evaluation data.Experiments on educational video keyframe extraction,cognitive style-based educational video classification and dynamic assessment data analysis have also been conducted in the system,Experiments have shown that the system provides a good display of experimental results and supports learners to learn better.