Competence Network for Advanced Speech Technologies


The competence network COAST is concerned with research and development in the field of speech recognition and speech interpretation with large lexicons for professional applications. The main research problems are: 

  • Development, improvement, and refinement of algorithms applied for speech recognition from the fields of statistics, acoustics, and signal processing, and their application specific parameterization.
  • Application of new techniques of semantic interpretation using artificial intelligence to improve speech recognition results and their usability.
  • Application-specific improvement and optimization of speech recognition, i.e., analysis of the question as to how speech recognition can optimally support concrete applications. The main focus here is on professional transcription of documents, messages, and meetings, as well as media mining for broadcast data.
  • Testing of new speech recognition applications.

The establishment of COAST is a significant contribution to the strengthening of Austria‘s position in research and industrial applications of speech technology. Bringing together different disciplines – speech recognition, signal processing, and semantic interpretation – allows to address limitations of traditional statistical speech recognition; limitations which are currently hindering a broad, comprehensive application of speech recognition in audio/media mining, information input, and professional document creation. COAST is organized as an association (“Verein”) whose members are the industrial and scientific partners, i.e., Philips Speech Recognition Systems Vienna, Sail Labs Technologies Vienna, SPSC Lab, the Austrian Research Institute for Artificial Intelligence OFAI, NOA Audio Solutions Vienna, FH Technikum Vienna, Zydacron Graz, PCS Villach, and ebit Linz. The SPSC Lab has taken the scientific leadership in the COAST association. In 2009, Philips Speech Recognition Systems was acquired by the globally acting Nuance Communications USA, but still the research program will be completed as originally planned.

Funding Program: 
K-net industrial cooperative research networks FFG and BMWFJ, ZIT Vienna and SFG
2006 - 2010
Research Area: