Abstract

Language content is pervasive in many multimedia sources and typically conveys specific information about objects and events within the media itself (eg, the name of a person whose face is displayed on screen) and within the larger outside world (eg, an anchor speakers commentary on a developing news story). While it is easy for humans to extract and synthesize content from multiple modalities simultaneously, such multimodal synthesis is still beyond the reach of automated systems. A small but important step toward multimodal systems is the successful application of automatic speech recognition technology to language-rich video domains, such as real-time broadcast news monitoring, resulting in valuable analytic tools, such as real-time alerting and keyword search.
While speech is a rich source of language content, it is only one part of such content in broadcast video. Often, a significant amount of language …

Date

2012

Authors

Prem Natarajan, Ehry Macrostie, Rohit Prasad, Jonathan Watson

Journal

Multimedia Information Extraction: Advances in Video, Audio, and Imagery Analysis for Search, Data Mining, Surveillance, and Authoring

Pages

175-184

Publisher

John Wiley & Sons, Inc.