This work was funded by LRE Project 62-058 `Speech recogniser Quality Assessment for Linguistic Engineering' (SQALE) and by the Linguistic Data Consortium. Additionally, Tony Robinson was supported by an EPSRC Advanced Research Fellowship.
Many thanks are due to the people who helped us collect the WSJCAM0 Speech Database. Firstly, to Doug Paul, without whose text utilities, scripts (ftp site gov.nist. ncsl.jaguar) and his order in the WSJ data we wouldn't have been able to record this database. Secondly to Jonathan Foote, whose data collection software we used and for setting up the recording equipment. Thirdly, we owe a lot to all at NIST for providing the SPHERE utilities and the WSJ0 texts. Finally, a big thank you goes to all the people in the SVR lab, who were patient throughout this resource intensive project.