Research On Course Recommendation Algorithm Based On Neural Attention Network And Dependency Embeddings

Massive Open Online Courses(MOOCs),offering millions of high-quality courses from prestigious universities and prominent experts,are picking up momentum in popularity.Although users who enroll on MOOCs have free access to abundant knowledge,they may easily get overwhelmed by information overload.Therefore,recommending technology is needed as a fundamental and wellaccepted effective solution.Differing from many other online recommendations,recommending courses to users on MOOCs still faces two challenges.First,users’ knowledge background differs,so does their purpose of learning.Second,online courses are not independent but intertwined with dependencies,which represents the learning sequence of knowledge points.This also means that online courses are not independent,but are related through their dependencies.In addition,most of the knowledge of online courses is in the course videos,so the basic and important step of MOOCs recommendation is to extract the concepts representing the knowledge points from the video captions.In this thesis,the paper first proposes a semi-supervised learning method based on conditional random field to extract concepts from course video captions.Then it preresent a course recommending algorithm based on neural attention network and dependency embeddings.Specifically,the main contributions of this thesis are briefly presented as follows.(1)The paper propose a semi-supervised method to extract concepts from MOOCs video captions.It optimizes feature selection,using a set of features from captions.At the same time,the semi supervised learning framework is improved,and the self-training is set appropriately.Experimental results on different datasets show that the algorithm can obtain performance close to that of supervised learning algorithm with a small amount of labeled data.(2)The paper propose a course recommendation method based on neural attention network and dependency embedding.This method first uses a set of semantic-independent features based on structural information and statistical information to extract concept-level dependencies in captions.Then it uses a specific set of rules to extract course-level dependencies.Finally,the method uses the class-level dependencies information to improve the calculation of the attention coefficient,and embeds it in the neural network based on the attention mechanism(neural attention network),so that it can effectively distinguish the contribution of different courses selected by users to the recommendation effect.Experimental results on real data sets show that the proposed MOOCs recommendation method can effectively improve the performance of the recommendation system.