Fourth ISCA ITRW on Speech Synthesis

August 29 - September 1, 2001
Perthshire, Scotland

Pmtools: A Pronunciation Modeling Toolkit

Richard Sproat

AT&T Labs - Research, Florham Park, NJ, USA

This paper reports on a pronunciation modeling toolkit - pmtools tools - that allows one to train a weighted fniute-state transducer using a Classification and Regression Tree (CART) training paradigm. Tools are provided to automatically align a pronunciation dictionary consisting of a set of words and their pronunciations, train a set of CART trees on the aligned dictionary and compile those trees out into a special class of weighted finite-state transducer. Most of the complexity - aligning the data, labeling the data with features and training the trees - is hidden from the user.

While some new techniques, e.g. in automatic alignment, are introduced here, the main focus of this work is to provide a toolkit to ease the development of pronunciation models using fairly standard techniques. By the time of the workshop, pmtools will be available free for non-commercial use.

Full Paper

Bibliographic reference.  Sproat, Richard (2001): "Pmtools: A pronunciation modeling toolkit", In SSW4-2001, paper 104.