Rich Transcription of Multi-lingual Spoken Documents for Translation and Distillation

Principal Investigator
Mari Ostendorf, Katrin Kirchhoff

Sponsor(s)
SRI International (flow through from DARPA)

Award Period
09/01/2005 - 08/31/2010

Abstract
This project will develop speech recognition and annotation technology for spoken documents, such as news broadcasts, with the goal of enabling higher accuracy and more efficient information retrieval, extraction and summarization. In addition, the technology will be developed for multiple languages, including English, Mandarin, and Arabic, with modules designed to interact with machine translation for non-English languages. The goal is to produce English transcripts complete with punctuation and capitalization and enriched with information about speaker role, emphasis, etc.

Updates or corrections to this page should be sent to gheaton@u.washington.edu.