Phonological Features for 0-shot Multilingual Speech Synthesis

TL;DR

This paper demonstrates that using a small set of phonological features derived from IPA enables multilingual TTS models to perform code-switching and generate intelligible speech in unseen languages at test time.

Contribution

The introduction of phonological features allows monolingual TTS models to handle code-switching and synthesize speech in unseen languages without retraining.

Findings

Enables code-switching in unseen languages

Produces intelligible, natural-sounding speech

Approximates sounds not seen during training

Abstract

Code-switching---the intra-utterance use of multiple languages---is prevalent across the world. Within text-to-speech (TTS), multilingual models have been found to enable code-switching. By modifying the linguistic input to sequence-to-sequence TTS, we show that code-switching is possible for languages unseen during training, even within monolingual models. We use a small set of phonological features derived from the International Phonetic Alphabet (IPA), such as vowel height and frontness, consonant place and manner. This allows the model topology to stay unchanged for different languages, and enables new, previously unseen feature combinations to be interpreted by the model. We show that this allows us to generate intelligible, code-switched speech in a new language at test time, including the approximation of sounds never seen in training.