Phonological Features for 0-shot Multilingual Speech Synthesis

By Marlene Staib, Tian Huey Teh, Alexandra Torresquintero, Devang S. Ram Mohan, Lorenzo Foglianti, Raphael Lenain, Jiameng Gao. 2020

38 citations recorded by the source corpus.

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.

Subfields: Artificial Intelligence, Experimental and Cognitive Psychology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Tao Qin
2. Tie-Yan Liu
3. Xu Tan
4. Wei-Ning Hsu
5. Lei He
6. Frank K. Soong
7. Yossi Adi
8. Michael Auli

In Experimental and Cognitive Psychology

1. Tao Qin
2. Frank K. Soong
3. Tie-Yan Liu
4. Xu Tan
5. Lei He

Citations in the corpus

• A Survey on Neural Speech Synthesis (2021, background): “Some works [438, 37, 206, 442, 107, 313, 219, 446, 106] propose to transfer voices across languages, e.”
• Scaling Speech Technology to 1, 000+ Languages (2023, background): “…work explored various approaches including byte encoding to unify text representations which was evaluated on English, Spanish, and Chinese [Li et al., 2019] Further studies explored other input representations, including phonemes [Zhang et al., 2019] and phonological features [Staib et al., 2020].”
• Multilingual Byte2Speech Text-To-Speech Models Are Few-shot Spoken Language Learners (2021, background): “Besides, transfer from other languages could be obtained by designing a uni-fied set of phonemes (Cai et al., 2020), often derived from IPA (Chen et al., 2019; Demirsahin et al., 2018), or creating feature vectors by phonology (Staib et al., 2020).”