Unsupervised, language-independent grapheme-to-phoneme conversion by latent analogy

Abstract

Automatic, data-driven grapheme-to-phoneme conversion is a challenging but often necessary task. The top-down strategy implicitly followed by traditional inductive learning techniques tends to dismiss relevant contexts when they have been seen too infrequently in the training data. The bottom-up philosophy inherent in pronunciation by analogy allows for a markedly better handling of unusual patterns, but also relies heavily on individual, language-dependent alignments between letters and phonemes. To avoid such supervision, this paper proposes an alternative solution, dubbed pronunciation by latent analogy, which adopts a more global definition of analogous events. For each out-of-vocabulary word, a neighborhood of globally relevant pronunciations is constructed through an appropriate data-driven mapping of its graphemic form. Phoneme transcription then proceeds via locally optimal sequence alignment and maximum likelihood position scoring. This method was successfully applied to the speech synthesis of proper names with a large diversity of origin.