PI: Thien Nguyen

We propose comprehensive resources and models for understanding automatically
transcribed videos. In particular, in this project, we pursue a deep learning model for identifying the
important points and questions mentioned in a video transcript. To achieve this objective, we employ two
specific deep learning models. First, we construct the first hierarchical model for keyword extraction from
video transcripts. In this model, the key phrases mentioned at the sentence level and paragraph-level are
extracted. Moreover, the model is trained to be aware of the paragraph level boundaries. Second, we
propose a novel model for identifying various types of questions mentioned in video transcript. In addition,
a joint and pipeline model for recognizing the answers to the questions from the transcript is also
presented. For each task, specific resources are constructed from the transcript of live streamed videos on
behance.net.