Automatic Learning of Word Transducers from Examples

Abstract

This paper describes the application of markovian teaming methods to the inference of word transducers. We show how the proposed method dispenses from the difficult design of hand-crafted rules, allows the use of weighed non deterministic transducers and is able to translate words by taking into account their whole rather than by making decisions locally. These arguments are illustrated on two examples: morphological analysis and grapheme-tophoneme transcription.