DARPA EARS visit, CUED September 2003

Agenda

Monday 29th September

9:00 Arrive + timetable for visit.
9:15 Introduction to CUED EARS Project + Overview of Progress P.C. Woodland
10:00 Overview of CTS English systems from 2002 and 2003 P.C. Woodland
11:00 break
11:15 Work on Broadcast News for Eval03 & post-eval Experiments D.Y. Kim
11:50 Design of Fast LVCSR Systems G. Evermann
12:20 Experiments with lightly supervised discriminative training on TDT data H.Y. Chan
13:00 Lunch break (LR4)
14:00 CU-HTK Mandarin CTS System M.J.F. Gales
14:30 Progress in Metadata Extraction at Cambridge University S.E. Tranter
15:00 Diarisation Research at Cambridge University S.E. Tranter
15:30 Slash Unit Research at CUED M. Tomalin
16:00 break / EARS board meeting
16:40 Single Pronunciation Dictionaries - Construction and Performance T. Hain
17:15 Development of HTK G. Evermann
17:40 Discussion of contract status/issues

Tuesday 30th September

Talks by PhD Students
9:00 Automatic Model Complexity Control Using Marginalized Discriminative Growth Functions X. Liu
9:30 Discriminative Speaker Adaptive Training L.Wang
9:50 Precision and Covariance Matrix Modelling K.C. Sim
10:10 Acoustic Factorization: Theory and Initial Experiments K. Yu
10:30 Continuous Space Language Modelling D. Mrva
11:00 break
11:15 Tour of lab
11:30 Experiments with Quick Transcriptions & Fisher Data P.C. Woodland
11:45 General Discussion
12:15 Future Plans & wrap-up
13:00 Lunch (LR6)

Detailed Project info

Consult a snapshot (September 2003) of the website of the CUED EARS HTK Rich Audio Transcription for a high-level overview, list of personnel and copies of all papers/presentations produced in the last year.

The online website, with updated information and new publications can be reached via this link: CUED HTK Rich Audio Transcription project

Web Links

CUED Machine Intelligence Laboratory

CUED HTK Rich Audio Transcription project

HTK3 webpage


Gunnar Evermann
Last modified: Sun Sep 28 17:54:52 BST 2003