Phrase-based Multimedia Information Extraction

摘要

StreamSage proposed to develop a prototype software system that would specifically deal with the two primary challenges of speech data on the performance of information extraction: degraded input data and the time-based nature of the content. In order to overcome these two challenges, this effort focused on two general areas: mitigating the degraded quality of speech data and improving entity identification. Technologies developed under this project for audio/video named entity identification and end user access to relevant information could have tremendous value for both military and commercial entities.