Creating Video Sequence Alternatives Across MOOCs Based on Document Similarity

Creating Video Sequence Alternatives Across MOOCs Based on Document Similarity

Grooff, Alexander (TU Delft Electrical Engineering, Mathematics and Computer Science; TU Delft Web Information Systems)

Lofi, Christoph (mentor)
Houben, Geert-Jan (graduation committee)
Zuñiga Zamalloa, Marco (graduation committee)

Delft University of Technology

2018-06-14

Since the introduction of Massive Online Open Courses (MOOCs) in 2008, the number of MOOCs offered by universities has increased enormously. Over 700 universities offer a total of 5924 MOOCs. Each MOOC holds a sequence of 10 to 140 videos and are meant to help online learners understand a given topic. These videos are intended to be watched in a specific, sequential order. This order can be seen as a chain where video 1 provides enough prerequisite knowledge for video 2, video 2 for video 3 and so on. In theory that is fine, however, this order is not necessarily the only ideal order to watch videos in. The online learner might want to take shortcuts through the MOOC's videos, find additional information on the topic in another video or explore what other subjects are available to study. In other words, the online learner could benefit from improved navigational features to better explore the surrounding relevant videos. By branching out the viewing order towards videos in other MOOCs, we add alternative sequence orders as possible learning paths for the online learner. This can help online learners plan out an alternative sequence order based on their own interests and backgrounds. The chain-like sequence order as described above, where one video provides the prerequisite knowledge for its successor, is crucial for creating alternative sequence orders. When video 1 from MOOC A is found to be similar to video 5 from MOOC B, it can be assumed that video 1 provides the prerequisite knowledge for video 6 in MOOC B. Similar videos between the MOOCs are found by comparing their transcripts, titles and durations. Based on this similarity a new sequence order is set up from one video to its similar counterparts’ successor video. We structure these alternative sequence orders by representing the videos from MOOCs as nodes into a graph and connect them with directed edges that represent the sequence order between videos. Several methods for finding similarities have been compared to find the most accurate way of comparing transcriptions where the method doc2vec yields the most accurate similarity cases. Evaluation based on a user study shows that our method creates alternative sequence orders that have 58.7% positive user ratings. This can be compared to semi-randomly picked sequence orders and the original sequence order defined in the MOOC, which have 8.0% and 76% positive user ratings respectively. This shows that the alternative sequence orders are significantly better than the semi-randomly picked sequence orders, and comparable but not as good as the original sequence order defined in the MOOC.

MOOCs
Document similarity
doc2vec

http://resolver.tudelft.nl/uuid:17e94139-275a-4301-804e-714e256ebc8f

Student theses

master thesis

© 2018 Alexander Grooff