Audimus.media (seminar)

From HLT@INESC-ID

Hugo Meinedo

Date

  • 14:30, January 26, 2007
  • 3rd floor meeting room

Speaker

Abstract

The presentation will review briefly our current Broadcast News automatic speech recognition system, Audimus.media and will focus in recent and future enhancements:

  • Unsupervised annotation of training data
  • Gender dependent acoustic models
  • Speaker dependent acoustic models