Institut

Studium

Forschung


 

 

  AUTOMATIC SEGMENTATION
OF
SPOKEN
LANGUAGE
  activities
experiments  
  past
report no.1  (german) deals with the basics of the project and the synchronization of the project partners. 
report no.2 (german) deals with the design of the evaluation lexicon with is first labelled automatically and then corrected by hand.
report no.3  (german) documents the basic evaluation of the Aligner without any adaptions and optimizations to the specific speech material.
report no.4, part 1  (german) shows the first results on using whole word models for non-speech specific phenomena.

present

present experiments concerning the adaptation of the Aligner to discourse specific phenomena like turntaking signals and hesitations as well as non-speech specific phenomena like noise, laughter, music, etc (klick here for first results)
report no.4, part 2  (german) shows the final results on using whole word models for non-speech specific phenomena.
report no.5 (german) shows the final results on using whole word models for hesitation and reception specific phenomena.

future

future experiments will be optimization of the text preprocessing and treatment of parts of speech signals in which more than one person are talking at the same time. 
meetings  
  internal meetings
IDS-Kolloquium am 16. Juli 1998
Text-Ton-Alignment von Korpora gesprochener Sprache,  Praesentation des Rahmenprogramms  (pdf-Format) und Projekts Alignment (Folien, postscript-, pdf-Format)

others 

Sprache und Gehirn
Gesellschaft fuer Angewandte Linguistik, 29 Jahrestagung in Dresden, 24.09. - 26.09.98 Abstract in html
Sprache und neue Medien
35. Jahrestagung des Instituts für Deutsche Sprache Mannheim, 16.-18. März 1999 
Zusammenfassung der Praesentation in html , ausführliche Beschreibung  des Projekts in pdf (2.7MB) & html und das Poster des Stuttgarter Aligners (pdf (12,8MB)) 

 



 
HOME RESEARCH PROGRAMME STAFF LABS & EQUIPMENT ACTIVITIES