NLP Driven Adaptive Video Playback Rate Adjustment

Abstract

This paper suggests the first method to adjust video playback rate actively according to the importance of the scenes using Natural Language Processing (NLP). For this purpose, the relationship between scene information and required playback time has been formalized. We propose a data set that consists of several YouTube videos from different categories. Our learning powered model mainly relies on some statistical extractive summarization techniques and various additional features such as context-specific important words. Furthermore, we have incorporated our model with a success-proven summarization model and studied its impact on the results. Lastly, we show subjective ratings for the proposed approach, which drastically reduces the required watching time without loss of any information.