The talk will give an overview of various approaches for supporting access to audio and audio-visual content. I will focus on two ways of supporting search, discovery and navigation: (a) search based on text queries, or (b) construction of links between semantically similar segments. The goal in both cases is to pinpoint relevant segments, and thus make navigation in audio-visual archives faster, more precise and more convenient for users. Both approaches are content-based, leveraging text mined from subtitles or automatic transcripts coupled with audio and visual features. The approaches described performed yielded the best in the 2014 MediaEval Search and Hyperlinking Task.
Petra Galuscakova is a postdoctoral researcher at the Institute for University of Maryland Advanced Computer Studies where she works on cross-language information retrieval. She recently completed her
Ph.D. in computational linguistics at Charles University in Prague. Her prior work has investigated methods for effective search and navigation in multimedia archives.