Automatic Lecture Video Content Summarizationwith Attention-based Recurrent Neural Network

摘要

This paper propose an automatic summarization method for a lecture video transcript that uses attention based Recurrent Neural Network (RNN) to capture the content of the lecture. Our fully data-driven model also utilize segmentation to split the input video transcript in order to increase topic coherency in each segment. We also use linguistic-based feature to help the model identify important word and key topic in the segment which improves the quality of the summary. Our model shows a significant improvement in the term of ROUGE score compared to the baseline models.