ISF Gastvortrag von Michael R. Lomnitz
19. September 2019
Seminar Room, Wohllebengasse 12-14 / Erdgeschoß
Improving speech technology with the open source VOiCES dataset
"Speech recognition technology is becoming a larger part of our daily lives and is poised to be the primary means of interaction with technology in the future, but before it becomes so, there are a number of things that need to be addressed. Machine learning systems for speech recognition are trained on recorded speech data, but few existing datasets reflect the complexity of the environmental settings. We present the VOiCES data-set, a freely available open source corpus designed to improve machine learning in speech applications, such as Automatic Speech Recognition (ASR) and Speaker Recognition (SR). We present results from the VOiCES at a Distance challenge at Interspeech 2019."