5th International Conference on Spoken Language Processing
This paper describes an undergraduate course in speech recognition, based on the CSLU Toolkit, which was taught at the Universidad de las Americas in Puebla, Mexico. Throughout the course, laboratory assignments based on the toolkit guided students through the process of creating a recognizer, while in-class lectures consistently referred to the architecture of the toolkit as a concrete example of an existing system. The class was organized so that lectures and laboratory assignments followed the steps taken in the creation of a new recognizer. The students first recorded and labeled their own corpus, then proceeded to design and train neural network based recognizers, before finally testing for performance and creating sample applications. As a final project, students performed simple, well-defined experiments using the recognizers they had constructed. The CSLU Toolkit is freely available for non-commercial use from http://cslu.cse.ogi.edu/. In the future, similar courses based on the toolkit could be created and shared by many researchers in the speech community via the world-wide web.
Bibliographic reference. Serridge, Ben (1998): "An undergraduate course on speech recognition based on the CSLU toolkit", In ICSLP-1998, paper 0925.