Adopting Argumentation Mining for Claim Extraction from TED Talks

Abstract

Engagement is critical for academic learning. It's commonly believed that motivating students to learn is crucial in education. We think that by providing students some interesting content based on what they are learning is a good idea. Since TED Talks share attractive new ideas, we are planning to motivate students by recommending TED Talks relevant to their learning content. Also, we found it's important to have some ``teasing texts'', which are used to convince students to watch TED Talks we recommended. to get these texts, we are going to adopt an argumentation mining technique called ``Claim Extraction'' on TED Talk subtitles.Claim extraction uses classifiers trained on a dataset to extract claim sentences from the given texts. And these claim sentences can be used as the ``teasing texts''. Due to the fact that there isn't any TED Talk based corpus and building one is extremely expensive, we have to train classifiers on the existing Wikipedia dataset. It means we have to deal with the cross-domain learning problem. This thesis will introduce our approach of building a TED Talk claim extraction system. This system will use classifiers trained on existing corpus and can extract claim sentences from TED Talk subtitles. Also, this thesis proposes using claims extracted from TED Talk subtitles can promote students to watch the recommended TED Talks.