Developing a collaborative MOOC learning environment utilizing video sharing with discussion summarization as added-value

Abstract

With the fast-growing Massive Open Online Courses (MOOC) community and the increase in the number of Learning Management Systems (LMSs) available online, the amount of shared information is massive. Current LMS - in particular MOOC providers - offer many advanced content delivery techniques: interactive video, active retrieval practices, and quizzes to enhance the pedagogical process. The main knowledge creation assets within MOOCs are encapsulated in other tools such as discussion forums, blogs, and wikis. Although these tools exist as separate entities within the platform, they still follow traditional techniques. We believe these tools need to be fully integrated to the main content and encourage spontaneous collaboration. From my experience with some MOOCs, the amount of collaboration and information-sharing is still overwhelming due to the massive number of participants and the limited range of collaborative tools. However, most of the shared information could be redundant or irrelevant. This information must be processed in order to provide the most concise knowledge. Therefore, we need to summarize this information from the discussions, blogs, and wikis and include the most relevant data in the course content. This thesis addresses this shortcoming by suggesting a new system with two primary components to accomplish this task. In the first component, we link the discussion tools to the main course content. Then, in the second component, we apply Natural Language Processing (NLP) techniques to present a summary of all shared content. We use techniques such as Term Frequency - Inverse Document Frequency (TF-IDF), stemming algorithm, Vector Space Model (VSM), and Cosine Similarity to rank the sentences. We then tune the TF-IDF values and boost the sentence ranks using the main content by delegating the first component’s features. The next step involves choosing the most relevant sentence to build our summary. Finally, we evaluate our result using Recall-Oriented Understudy for Gisting Evaluation (ROUGE) system, which compares our automated summary to human extracted summaries. These results demonstrate that we can achieve high improvement summar y compared to the baseline and other similar techniques.