The Kestrel TTS text normalization system

By Peter Ebden, Richard Sproat. 2014

90 citations recorded by the source corpus.

Subfields: Artificial Intelligence

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. Yu Zhang
2. Tao Qin
3. Yonghui Wu
4. Ye Jia
5. Tie-Yan Liu
6. Yoshua Bengio
7. Quoc V. Le
8. Xu Tan

Citations in the corpus

• LibriTTS: A Corpus Derived from LibriSpeech for Text-to-Speech (2019, methodology): “, abbreviations, numbers and currency expressions) and semiotic classes [36] (text tokens and token sequences that represent particular entities that are semantically constrained, such as measure phrases, addresses and dates) in the sentences are detected and normalized by a weighted finite state tr”
• Sample Efficient Adaptive Text-to-Speech (2018, methodology): “Each step is explained as follows: Sequence of phonemes The sequence of phonemes were predicted by the text-analysis front-end pipeline (Ebden and Sproat, 2015) using a tokenizer, POS tagger, dependency parser, text normalizer, lexicon, and grapheme-to-phoneme rules.”
• CVSS Corpus and Massively Multilingual Speech-to-Speech Translation (2022, positive): “We normalize the translation text from CoVoST 2 using a proprietary weighted finite state transducer (WFST)- based text normalizer (Ebden and Sproat, 2015).”
• Neural Models of Text Normalization for Speech Applications (2019, methodology): “The Bell Labs multilingual TTS system (Sproat 1996, 1997) introduced the use of weighted finite-state transducers for text normalization, and this approach is still in use in deployed systems, such as Google’s Kestrel text-normalization system (Ebden and Sproat 2014).”
• Improving the Prosody of RNN-Based English Text-To-Speech Synthesis by Incorporating a BERT Model (2020, methodology): “The preprocessor is a text-normalization front-end [23, 24], and is not updated as part of the CHiVE-BERT training procedure.”
• RNN Approaches to Text Normalization: A Challenge (2016, methodology): “As described in (Ebden and Sproat, 2014), Kestrel’s verbalizations are produced by first tok-enizing the input and classifying the tokens, and then verbalizing each token according to its semi-otic class.”