Creating a mexican Spanish version of the CSLU toolkit

The CSLU Toolkit is designed to facilitate the rapid development of spoken dialogue systems for a wide variety of applications, as well as to provide a framework for conducting research in the underlying speech technologies. This paper describes the creation of a Mexican Spanish version of the CSLU Toolkit (both synthesis and recognition) undertaken at the Universidad de las Américas in Puebla, México. Based on the Festival Speech Synthesis System of the University of Edinburgh, we have developed a complete concatenative text-to-speech system for Mexican Spanish, which is currently incorporated into the toolkit and includes both a male and female voice. In the area of recognition, we have created a set of task-specific Spanish recognizers for continuous digits, spelled words, and yes/no phrases, as well as a "general-purpose" phonetic recognizer suitable for arbitrary sub-domains. Using the Rapid Application Developer (RAD) component of the CSLU Toolkit, it is now possible to quickly prototype spoken dialogue systems in Spanish. The Spanish components of the CSLU Toolkit are freely available for noncommercial use from the following web page: http://info.pue.udlap.mx/~sistemas/tlatoa.