The architecture and the implementation of a finite state pronunciation lexicon for Turkish

This paper describes the architecture and the implementation of a full-scale pronunciation lexicon for Turkish using finite state technology. The system produces at its output, a parallel representation of the pronunciation and the morphological analysis of the word form so that further disambiguation processes can be used to disambiguate pronunciation. The pronunciation representation is based on the SAMPA standard and also encodes the position of the primary stress. The computation of the position of the primary stress depends on an interplay of any exceptional stress in root words and stress properties of certain morphemes, and requires that a full morphological analysis be done. The system has been implemented using XRCE Finite State Toolkit.