Retrieval of Keyphrase Automatically from Video Lectures using Semi Supervised Machine Learning Algorithm

A few advanced lecture browsers synchronize text with lecture video and also allow search within the transcript. However, these systems are of little use if for instance, a student wants to quickly scan the contents of a particular lecture among a series of lectures. So this project proposes a system that can automatically generate and display section-wise annotations using lecture transcripts. This approach uses a simpler keyphrase-based annotation technique, which functionally strikes a middle ground between detailed annotation and basic video tagging. It uses a supervised machine learning algorithm, based on a Naive-Bayes classifier to extract relevant keyphrases. The goal of keyphrase extraction is to generate an optimal set of phrases appearing in the lecture, which best summarizes its content. The project has shown that a combination of automatic keyphrase extraction and segmentation enhances the functionality of a lecture browser system.