Applying Transducers to Spoken Language Processing for Portuguese
Isabel M. Trancoso · 2003
This paper has two different goals. The primary aim is to illustrate the advantages of weighted finite state transducers for spoken language processing, namely in terms of their capacity to efficiently integrate different types of knowledge sources. We have chosen three areas to emphasize several aspects of the application of transducers: large vocabulary continuous speech recognition, automatic alignment and grapheme-to-phone conversion. The secondary goal is to simultaneously present the state of the art in these areas for European Portuguese.