The SPEECHDAT(E) project: Creating speech databases for eastern European languages
Henk van den Heuvel, V. I. Galunov, Herbert S. Tropf · 1998
This paper gives some information regarding the SpeechDat(E) project for collecting telephone speech databases for teleservices in Eastern European countries. The project's position in the framework of other SpeechDat projects is explained; the organisation of the project is sketched; some details regarding contents and validation are presented; and finally, the current status of the project is briefly addressed. Introduction Recognition performance is the key to a successful teleservice. Satisfactory performance, however, is only achievable if realistic training data are available. The training data should comprise between several hundred and a few thousand speakers per gender depending on the number of speakers of the language in question. It should cover different dialects and accents and it should be representative of the telephone channel conditions likely to be encountered. Further, it is particularly important in Europe, that the service can be offered in different languages, ...