Heterogeneous Audio Visual Internet Collection (HAVIC)
LDC built a large corpus of multi-modal data to support research in a variety of areas including spoken term detection and video event detection. The HAVIC (Heterogeneous Audio Visual Internet Collection) Corpus consists of thousands of hours of “real world” video data collected from the internet. The corpus especially targeted user-generated video content as opposed to professionally-produced or commercial video content. A significant portion of the collected video data was multi-modal; that is, it contained content beyond the video stream itself, like audio or text embedded in the video.
The HAVIC Corpus was used as part of the TRECVid Multimedia Event Detection (MED) technology evaluation. The goal of MED was to assemble core detection technologies into a system that could quickly and accurately search a multimedia collection for user-defined events. An event for MED was "an activity-centered happening that involves people engaged in process-driven actions with other people and/or objects at a specific place and time". A set of target events were defined in advance and a significant portion of the HAVIC Corpus is comprised of videos that illustrate these events. The corpus also includes negative examples (i.e. videos that are related to the target events but fail to satisfy the event definition) and off-topic videos that are completely unrelated to the target events. Each video added to the corpus was also labeled (annotated) with a set of judgments describing its event properties and other salient features.
Stephanie Strassel, Amanda Morris, Jonathan Fiscus, Christopher Caruso, Haejoong Lee, Paul Over, James Fiumara, Barbara Shaw, Brian Antonishek, Martial Michel
Creating HAVIC: Heterogeneous Audio Visual Internet Collection
LREC 2012: 8th International Conference on Language Resources and Evaluation, Istanbul, May 21-27
Available: Paper in PDF, Poster in PDF