Abstract
Public speaking is an effective way to move, persuade, and inspire. While many guidelines have been presented to teach public speaking skills, they are often based on anecdotal evidence and not customizable. Exploring high-quality speeches such as TED Talks could provide insights to eliminate limitations in existing guidelines. This study aims to explore and identify narration strategies by conducting visual analysis into the textural and acoustic information in public speeches. We present SpeechLens, an interactive visual analytics system to explore large-scale speech dataset with multiple level-of-details. SpeechLens features a novel focus+context design to enable intuitive and smooth analysis. Case studies indicate the effectiveness and usefulness of our approach.