Automatically Visualizing Audio Travel Podcasts.

UIST '17: The 30th Annual ACM Symposium on User Interface Software and Technology Québec City QC Canada October, 2017(2017)

引用 5|浏览85
Audio Podcasts have gained popularity because they are a compelling form of storytelling and are easy to consume. However, they are not as easy to produce since resources are invested in the research, recording, and editing process and the average length of an episode is over an hour. Some audio podcasts could benefit from visuals to increase engagement and learning, but manually curating them can be arduous and time-consuming. We introduce a tool for automatically visualizing audio podcasts, currently focused on the genre of travelogues. Our system works by first time-aligning the transcript of a given podcast, using NLP techniques to extract entities and track how interesting or relevant they are throughout the podcast, and then retrieving visual data appropriately to describe them, either through transitions on a map or professionally taken photographs with captions. By automatically creating a visual narrative to accompany a podcast, we hope our tool can provide listeners with a better sense of the podcast's topic.
Podcasts, Travelogues, Visualization
AI 理解论文
Chat Paper