This paper gives an overview of two prototype systems for phoneme-based, large-vocabulary speech recognition. During the last decade, the performance of automatic systems for continuous speech recognition has been drastically improved. This progress has been mainly achieved by sophisticated statistical techniques. The prototype systems to be described are being developed within European research projects: in the ESPRIT project POLYGLOT, an isolated word recognition system is being developed for several European languages; in the SPICOS project, a continuous speech recognition system has been built that is being tested for several tasks, such as spoken database queries and the DARPA RM task.
Bibliographic reference. Ney, Hermann / Billi, Roberto (1991): "Prototype systems for large-vocabulary speech recognition: polyglot and spicos", In EUROSPEECH-1991, 193-200.