Video Based Transcript Summarizer for Online Courses using Natural Language Processing

Abstract

Online education has become an effective way to deliver quality education to students. They have become more popular because of their high graphical and pictorial content, delivered by experts in the subjects and convenient for learning at anytime and anywhere. But sometimes, students may not be able to go through the course content due to shortage of time. Video transcript summarizer has got a lot scope in this situation. It highlights the important topics from the video. The idea of summarizing the videos can be extended to online courses videos. This will help students save a lot of time as they can understand the gist of the class within less time without actually watching the video and by just going through the summary. Our system focuses on the development of a module using Natural Language Processing with python to summarize an online class video. The methodology adopted in this project uses Natural Language Processing (NLP) algorithms such as Term Frequency-Inverse Document Frequency (TF-IDF) and Gensim to obtain the summary of video of online course. The model takes URL of a video from user as input. We have implemented summarization process with the help of two algorithms. TF-IDF is an information retrieval algorithm which uses frequency of a term and its inverse document frequency. Gensim is a NLP package that deals with topic modeling. The model also gives the flexibility to the user to decide on as to what percentage of summary is needed compared to the original lecture. The summarization technique is a subjective process. We have incorporated two prominent methods. One is cosine similarity and the other one is ROUGE score. The former does not require human generated summary for reference, whereas latter requires it. The efficiency obtained using Cosine similarity is greater than 90% in both the cases: TF-IDF and Gensim. The efficiency obtained in case of ROUGE score is in between 40-50%.