MOOC Concept Extraction from Chinese Texts: A Rule and Graph Propagation Based Method

In recent years, massive open online courses (MOOCs) have provided learners with a wealth of information resources. Extracting concepts from MOOCs has become an important research topic. However, in recent Chinese concept extraction methods based on graph propagation, the features of concepts are not fully taken into account during the generation of candidate concepts. Consequently, this results in the inclusion of excessive non-concept words or phrases during the subsequent ranking of candidate concepts, thus affecting the accuracy of concept extraction. In this paper, we propose a concept extraction method based on rules and graph propagation algorithm (CE-RGPA), which can extract concepts from Chinese texts in MOOCs. Specifically, we construct one language rule and nine contextual rules to extract candidate concepts by analyzing caption texts. Then, the graph propagation algorithm is adopted to rank the candidate concepts based on confidence scores. Finally, the candidate concepts with high confidence scores are selected as the concepts. We conduct experiments on real-world data and validate the effectiveness of the proposed method.