Content-based retrieval systems for personalization of educational videos

Most of the traditional educational learning materials exist in the form of videos, audio, slides, and textbooks. In many cases, they are not catered for on-line distribution, presentation, and personalization. Since the amount of information from these materials can be enormous, annotating them manually can be laborious and is not feasible. An automatic way is thus needed to transform the raw text and video resources into high level representation before they are ready for distribution. In our project, we are pursuing a user centric approach to E-Learning in which the system performs semantic analysis of the lecture to create a space of possible presentations. In this paper, we present an approach where the slides content and video content is used to index lecture material into semantic concepts such as 'example', 'definition' etc. These concepts are determined using a rule-based approach along with using some contextual information. The experiments done over the 26 lecture videos of Algorithms course have shown encouraging results.