K.U.Leuven
  Search for Staff Students Organisational chart Search matrix Keywords

Seminar 2006-11-17

FLaVoR Workshop: On Flexible Architectures For Large Vocabulary Recognition

November 17, 2006
Faculty Club, Leuven, Belgium

Abstract

As closing of the FLAVOR project, a 1 day workshop on flexible architectures for (very) large vocabulary recognition is being organized in Leuven on November 17, 2006.

During the workshop the main results of the project will be presented, as well as a number of invited talks from other institutions who have been working on issues that are similar or related too the issues tackled in the FLAVOR project.

The major goal of the FLAVOR project was to move away from the classical all-in-one HMM large vocabulary decoding strategy in order to allow the integration of more sophisticated linguistic knowledge sources than the usual fixed lexicon + N-gram language models - the state-of-the-art approach for about 20 years.

In order to achieve these goals, 2 issues needed to be tackled: (i) the design of a layered architecture without the loss of accuracy compared to the traditional approach, (ii) the definition and incorporation of adequate linguistic knowledge sources for morphologically (moderately) productive languages such as Dutch.

The morning session of the workshop focuses on architecture and software issues. The afternoon session focuses on the linguistic resources and their integration into the ASR system.

Program

9:45 Welcome & coffee
10:15 Opening and introduction to the FLaVoR project
D. Van Compernolle, K.U.Leuven
10:45The SPEERAL decoder (invited talk) [PPT]
P. Nocera, U. Avignon
11:15 Break
11:30Phone decoding for large vocabulary recognition
K. Demuynck, K.U.Leuven
12:00FST tools in ASR (invited talk)
B. Hoffmeister, RWTH Aachen
12:30 Lunch
14:00A memory-based learning-plus-inference approach to morphological segmentation (invited talk) [PPT]
A. van den Bosch, U.Tilburg
14:30Morpho-syntactic language models for ASR
D.H. Van Uytsel, K.U.Leuven
Morpho-syntactic analysis and language modeling using machine learning techniques [PPT]
G. De Pauw, U. Antwerpen
15:15 Coffee break
15:30 Pronunciation modeling
S. Gillis, U. Antwerpen
16:00FLaVoR: System, issues, results & demos
K. Demuynck, K.U.Leuven
16:30Conclusions & discusssion
16:45 Cocktail

Travel information

If you come by car: driving directions can be found at the Faculty Club's website.

If you arrive by train at Leuven railway station, you can take bus 2 in the direction of "Heverlee, Campus" (about every 7 minutes); get off at stop "St.-Kwintenskerk" (13 minutes later). The Faculty Club, situated in the Groot Begijnhof, is a 5 minutes walk down from there: descend from the Sint-Kwintensberg leaving the church at your lefthand side, cross Sint-Beggaberg and Schapenstraat, and enter the Groot Begijnhof. There you will find arrows to the Faculty Club and to the "Huis van Chièvres", the venue for the main program.

Registration has been closed.

The deadline, November 13, was passed.

A few late registrations may possibly be arranged; you can try calling +32-16-321859 or +32-16-321055.

K.U.Leuven - Claim Copyright © Katholieke Universiteit Leuven | Comments on the content:
Production: | Most recent update: May 22, 2012 | Disclaimer
URL: