Phrase break prediction with bidirectional encoder representations in Japanese text-to-speech synthesis

1 code implementation • 26 Apr 2021 • Kosuke Futamata, Byeongseon Park, Ryuichi Yamamoto, Kentaro Tachibana

We propose a novel phrase break prediction method that combines implicit features extracted from a pre-trained large language model, a. k. a BERT, and explicit features extracted from BiLSTM with linguistic features.

Language Modelling Large Language Model +3