---
title: "Unsupervised Data Selection for TTS — Baali 2023"
description: "The paper asks whether usable Arabic text-to-speech can be built without a studio corpus by reusing broadcast news collected for ASR."
authors: ["Massa Baali", "Tomoki Hayashi", "Hamdy Mubarak", "Soumi Maiti", "Shinji Watanabe", "Wassim El-Hajj", "Ahmed Ali"]
year: 2023
venue: "arXiv:2301.09099"
source: "https://arxiv.org/abs/2301.09099"
doi: "10.48550/arxiv.2301.09099"
arxiv: "2301.09099"
licence: "arXiv — non-exclusive license to distribute"
licence_url: "http://arxiv.org/licenses/nonexclusive-distrib/1.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/baali-2023-unsupervised-data-selection-tts-using/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/baali-2023-unsupervised-data-selection-tts-using/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2023](https://md.research.elamur.ai/free/en/papers/2023/page/8/) / Unsupervised Data Selection for TTS: Using Arabic Broadcast News as a Case Study

# Unsupervised Data Selection for TTS: Using Arabic Broadcast News as a Case Study

**Authors:** [Massa Baali](https://md.research.elamur.ai/free/en/authors/massa-baali/), Tomoki Hayashi, [Hamdy Mubarak](https://md.research.elamur.ai/free/en/authors/hamdy-mubarak/), Soumi Maiti, [Shinji Watanabe](https://md.research.elamur.ai/free/en/authors/shinji-watanabe/), [Wassim El-Hajj](https://md.research.elamur.ai/free/en/authors/wassim-el-hajj/), [Ahmed Ali](https://md.research.elamur.ai/free/en/authors/ahmed-ali/) · **Year:** 2023 · **Venue:** arXiv:2301.09099 · **Licence:** [arXiv — non-exclusive license to distribute](http://arxiv.org/licenses/nonexclusive-distrib/1.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/2301.09099) · [PDF](https://arxiv.org/pdf/2301.09099) · [DOI 10.48550/arxiv.2301.09099](https://doi.org/10.48550/arxiv.2301.09099) · [arXiv 2301.09099](https://arxiv.org/abs/2301.09099) · [OpenAlex](https://openalex.org/W4317940367)

**Citations:** 4 citations (openalex-snapshot · 2026-10-14) · [5](#cited-by) in our library, of the 140884 papers whose bibliographies we read

## Summary
The paper asks whether usable Arabic text-to-speech can be built without a studio corpus by reusing broadcast news collected for ASR. The authors start from Aljazeera recordings in MGB-2/QASR and keep only two anchor speakers, one male and one female. They then filter for clean segments in two ways: a linguist sorted segments into six classes, and three automatic MOS predictors (MOSNet, wv-MOS, DNSMOS) scored each segment, with a threshold of 4. An Arabic ASR system repairs faulty transcripts, and automatic diacritization adds the missing vowels. For modelling, Tacotron2 and Transformer-TTS are pre-trained on English LJSpeech and fine-tuned on about one hour of Arabic. These autoregressive models then provide durations to a FastSpeech2-based Conformer. The final system reaches 3.9% CER, against 1.3% for real recordings, and scores MOS 4.4 for intelligibility and 4.2 for naturalness. Listeners often recognized the actual broadcaster. The study shows that a small, carefully filtered subset of found data can be enough to build TTS for a language with few resources.

## Key points
- **Source data:** MGB-2 Aljazeera broadcasts from 2004 to 2015, covering more than 4,000 episodes across 19 programs, with audio at 16 kHz. The speaker metadata was messy; Barack Obama alone appeared under 9 name variants. It was cleaned through repeated automatic parsing followed by manual checks.
- **Manual labelling:** Among the anchors' segments, the two biggest problems were bad recording quality (40%) and overlapping speech (33%). About one hour of clean audio per speaker survived.
- **Automatic selection:** DNSMOS and wv-MOS agreed closely with the human "good" labels. A TTS trained on the DNSMOS selection reached 4.0% CER, compared with 3.9% for the hand-picked selection.
- **Diacritization:** Adding vowels to the transcripts cut CER from 32.2% to 3.9%.
- **Training and release:** The AR models start from weights pre-trained on LJSpeech (24 h). The token embedding layer is not transferred because the symbol sets differ, and guided attention loss is added. Audio is generated with a Parallel WaveGAN vocoder. The recipe is published in ESPnet as `qasr_tts`.

## Interesting ideas
- **Name the failure types first, then validate the filter:** The authors defined explicit classes of bad data on a sample labelled by an expert, and only then checked an automatic filter against that sample. The filter's quality becomes measurable instead of assumed. — For Mova (learn): Classify silver-tree failures (e.g. wrong attachment, tokenization error, garbage input) on a hand-checked subset. Then measure how well witness-agreement filtering recovers the trees that humans judged good.
- **A small clean set can beat a large noisy one:** A cheap quality score with a fixed threshold produced a training subset that matched expert selection. This suggests aggressive filtering often pays off more than collecting more volume. — For Mova (learn): Train parsers only on silver sentences above a strict agreement threshold, and compare against the full silver pool to show how much noise costs.
- **Repair good samples instead of discarding them:** Segments with clean audio but faulty transcripts were corrected with a second model (ASR) rather than thrown away. This keeps scarce data that would otherwise be lost. — For Mova (learn): When witness parsers disagree on only one arc, apply self-induced repair rules to fix that arc. This keeps the rest of the sentence's tree rather than dropping it.
- **Restore missing input information before modelling:** Adding diacritics, which the text simply lacked, made a larger difference than any architecture choice. — For Mova (tokens): Normalize information that is often missing from input before tagging, such as dropped apostrophes, casing, and contractions. This keeps downstream errors from being blamed on the parser.
- **Use a slow, robust model to train a fast one:** The autoregressive model, which learns alignment reliably, provided durations to the faster non-autoregressive model, which could not learn alignment well from one hour of data. — For Mova (tree): The GPU transformer parser gparse could provide arc or label targets for training the faster perceptron parser on unlabelled text.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

Fields: [Linguistics & languages](https://md.research.elamur.ai/free/en/shelves/linguistics/)

## Cited by

Cited by

5 papers of our library cite it (of the 140884 whose bibliographies we read); 2 of them have a page here, 3 not yet.

- [FOOCTTS: Generating Arabic Speech with Acoustic Environment for Football Commentator](https://md.research.elamur.ai/free/en/papers/baali-2023-fooctts-generating-arabic-speech-acoustic/)
- [ClArTTS: An Open-Source Classical Arabic Text-to-Speech Corpus](https://md.research.elamur.ai/free/en/papers/kulkarni-2023-clartts-open-source-classical-arabic/)

## Cite

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@misc{baali2023unsupervised,
  title = {{Unsupervised Data Selection for TTS: Using Arabic Broadcast News as a Case Study}},
  author = {Baali, Massa and Hayashi, Tomoki and Mubarak, Hamdy and Maiti, Soumi and Watanabe, Shinji and El-Hajj, Wassim and Ali, Ahmed},
  year = {2023},
  doi = {10.48550/arxiv.2301.09099},
  eprint = {2301.09099},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2301.09099}
}
```

- [← previous Unravelling Indirect Answers to Wh-Questions: Corpus Construction, Analysis, and Generation](https://md.research.elamur.ai/free/en/papers/yusupujiang-2023-unravelling-indirect-answers-wh-questions/)
- [next → Using LSTM and GRU With a New Dataset for Named Entity Recognition in the Arabic Language](https://md.research.elamur.ai/free/en/papers/shaker-2023-using-lstm-gru-new-dataset/)
- [↑ back to papers of 2023](https://md.research.elamur.ai/free/en/papers/2023/page/8/)

## References

31 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. H. Zen, V. Dang, R. Clark, Y. Zhang, Ron J Weiss, Y. Jia, Z. Chen, and Y. Wu, “LibriTTS: A corpus derived from Librispeech for text-to-speech,” Proc. Interspeech 2019. [find](https://scholar.google.com/scholar?q=H%2E+Zen%2C+V%2E+Dang%2C+R%2E+Clark%2C+Y%2E+Zhang%2C+Ron+J+Weiss%2C+Y%2E+Jia%2C+Z%2E+Chen%2C+and+Y%2E+Wu%2C+%E2%80%9CLibriTTS%3A+A+corpus+derived+from+Librispeech+for+text%2Dto%2Dspeech%2C%E2%80%9D+Proc%2E+Interspeech+2019%2E) [our card](https://md.research.elamur.ai/free/en/papers/zen-2019-libritts-corpus-derived-librispeech-text/)
2. Y. Ren, Y. Ruan, X. Tan, T. Qin, S. Zhao, Z. Zhao, and T. Liu, “Fastspeech: Fast, robust and controllable text to speech,” Advances in neural information processing systems, 2019. [find](https://scholar.google.com/scholar?q=Y%2E+Ren%2C+Y%2E+Ruan%2C+X%2E+Tan%2C+T%2E+Qin%2C+S%2E+Zhao%2C+Z%2E+Zhao%2C+and+T%2E+Liu%2C+%E2%80%9CFastspeech%3A+Fast%2C+robust+and+controllable+text+to+speech%2C%E2%80%9D+Advances+in+neural+information+processing+systems%2C+2019%2E)
3. H. Tachibana, K. Uenoyama, and S. Aihara, “Efficiently trainable text-to-speech system based on deep convolutional networks with guided attention,” in ICASSP, 2018. [find](https://scholar.google.com/scholar?q=H%2E+Tachibana%2C+K%2E+Uenoyama%2C+and+S%2E+Aihara%2C+%E2%80%9CEfficiently+trainable+text%2Dto%2Dspeech+system+based+on+deep+convolutional+networks+with+guided+attention%2C%E2%80%9D+in+ICASSP%2C+2018%2E)
4. Je. Donahue, S. Dieleman, M. Bińkowski, E. Elsen, and K. Simonyan, “End-to-end adversarial text-to-speech,” International Conference on Learning Representations, 2020. [find](https://scholar.google.com/scholar?q=Je%2E+Donahue%2C+S%2E+Dieleman%2C+M%2E+Bin%CC%81kowski%2C+E%2E+Elsen%2C+and+K%2E+Simonyan%2C+%E2%80%9CEnd%2Dto%2Dend+adversarial+text%2Dto%2Dspeech%2C%E2%80%9D+International+Conference+on+Learning+Representations%2C+2020%2E)
5. Y. Wang, R. Skerry-Ryan, D. Stanton, Y. Wu, R. Weiss, N. Jaitly, Z. Yang, Y. Xiao, Z. Chen, S. Bengio, et al., “Tacotron: Towards end-to-end speech synthesis,” Proc. Interspeech 2017. [find](https://scholar.google.com/scholar?q=Y%2E+Wang%2C+R%2E+Skerry%2DRyan%2C+D%2E+Stanton%2C+Y%2E+Wu%2C+R%2E+Weiss%2C+N%2E+Jaitly%2C+Z%2E+Yang%2C+Y%2E+Xiao%2C+Z%2E+Chen%2C+S%2E+Bengio%2C+et+al%2E%2C+%E2%80%9CTacotron%3A+Towards+end%2Dto%2Dend+speech+synthesis%2C%E2%80%9D+Proc%2E+Interspeech+2017%2E)
6. Y. Ren, C. Hu, T. Qin, S. Zhao, Z. Zhao, and T. Liu, “Fastspeech 2: Fast and high-quality end-to-end text-to-speech,” International Conference on Learning Representations, 2020. [find](https://scholar.google.com/scholar?q=Y%2E+Ren%2C+C%2E+Hu%2C+T%2E+Qin%2C+S%2E+Zhao%2C+Z%2E+Zhao%2C+and+T%2E+Liu%2C+%E2%80%9CFastspeech+2%3A+Fast+and+high%2Dquality+end%2Dto%2Dend+text%2Dto%2Dspeech%2C%E2%80%9D+International+Conference+on+Learning+Representations%2C+2020%2E)
7. R. Sonobe, S. Takamichi, and H. Saruwatari, “JSUT corpus: free large-scale japanese speech corpus for end-to-end speech synthesis,” arXiv:1711.00354, 2017. [↗](https://arxiv.org/abs/1711.00354)
8. T. Hayashi, R. Yamamoto, K. Inoue, T. Yoshimura, S. Watanabe, T. Toda, K. Takeda, Y. Zhang, and X. Tan, “Espnet-tts: Unified, reproducible, and integratable open source end-to-end text-to-speech toolkit,” in ICASSP, 2020. [find](https://scholar.google.com/scholar?q=T%2E+Hayashi%2C+R%2E+Yamamoto%2C+K%2E+Inoue%2C+T%2E+Yoshimura%2C+S%2E+Watanabe%2C+T%2E+Toda%2C+K%2E+Takeda%2C+Y%2E+Zhang%2C+and+X%2E+Tan%2C+%E2%80%9CEspnet%2Dtts%3A+Unified%2C+reproducible%2C+and+integratable+open+source+end%2Dto%2Dend+text%2Dto%2Dspeech+toolk)
9. T. Tu, Y. Chen, C. Yeh, and H. Lee, “End-to-end text-to-speech for low-resource languages by cross-lingual transfer learning,” Proc. Interspeech 2019. [find](https://scholar.google.com/scholar?q=T%2E+Tu%2C+Y%2E+Chen%2C+C%2E+Yeh%2C+and+H%2E+Lee%2C+%E2%80%9CEnd%2Dto%2Dend+text%2Dto%2Dspeech+for+low%2Dresource+languages+by+cross%2Dlingual+transfer+learning%2C%E2%80%9D+Proc%2E+Interspeech+2019%2E)
10. Y. Chung, Y. Wang, W. Hsu, Y. Zhang, and R. Skerry-Ryan, “Semi-supervised training for improving data efficiency in end-to-end speech synthesis,” in ICASSP, 2019. [find](https://scholar.google.com/scholar?q=Y%2E+Chung%2C+Y%2E+Wang%2C+W%2E+Hsu%2C+Y%2E+Zhang%2C+and+R%2E+Skerry%2DRyan%2C+%E2%80%9CSemi%2Dsupervised+training+for+improving+data+efficiency+in+end%2Dto%2Dend+speech+synthesis%2C%E2%80%9D+in+ICASSP%2C+2019%2E)
11. J. Xu, X. Tan, Y. Ren, T. Qin, J. Li, S. Zhao, and T. Y. Liu, “Lrspeech: Extremely low-resource speech synthesis and recognition,” in Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2020. [find](https://scholar.google.com/scholar?q=J%2E+Xu%2C+X%2E+Tan%2C+Y%2E+Ren%2C+T%2E+Qin%2C+J%2E+Li%2C+S%2E+Zhao%2C+and+T%2E+Y%2E+Liu%2C+%E2%80%9CLrspeech%3A+Extremely+low%2Dresource+speech+synthesis+and+recognition%2C%E2%80%9D+in+Proceedings+of+the+26th+ACM+SIGKDD+International+Conference+on+Kno)
12. H. Zhang and Y. Lin, “Unsupervised learning for sequence-to-sequence text-to-speech for low-resource languages,” Proc. Interspeech 2020. [find](https://scholar.google.com/scholar?q=H%2E+Zhang+and+Y%2E+Lin%2C+%E2%80%9CUnsupervised+learning+for+sequence%2Dto%2Dsequence+text%2Dto%2Dspeech+for+low%2Dresource+languages%2C%E2%80%9D+Proc%2E+Interspeech+2020%2E)
13. P. Gallegos, J. Williams, J. Rownicka, and S. King, “An unsupervised method to select a speaker subset from large multi-speaker speech synthesis datasets,” Interspeech, 2020. [find](https://scholar.google.com/scholar?q=P%2E+Gallegos%2C+J%2E+Williams%2C+J%2E+Rownicka%2C+and+S%2E+King%2C+%E2%80%9CAn+unsupervised+method+to+select+a+speaker+subset+from+large+multi%2Dspeaker+speech+synthesis+datasets%2C%E2%80%9D+Interspeech%2C+2020%2E)
14. H. Mubarak, A. Hussein, S. Chowdhury, and A. Ali, “QASR: QCRI Aljazeera speech resource–a large scale annotated Arabic speech corpus,” ACL, 2021. [find](https://scholar.google.com/scholar?q=H%2E+Mubarak%2C+A%2E+Hussein%2C+S%2E+Chowdhury%2C+and+A%2E+Ali%2C+%E2%80%9CQASR%3A+QCRI+Aljazeera+speech+resource%E2%80%93a+large+scale+annotated+Arabic+speech+corpus%2C%E2%80%9D+ACL%2C+2021%2E)
15. P. Bell, M. Gales, T. Hain, J. Kilgour, P. Lanchantin, X. Liu, A. McParland, S. Renals, Os. Saz, M. Wester, et al., “The MGB challenge: Evaluating multi-genre Broadcast media recognition,” in ASRU, 2015. [find](https://scholar.google.com/scholar?q=P%2E+Bell%2C+M%2E+Gales%2C+T%2E+Hain%2C+J%2E+Kilgour%2C+P%2E+Lanchantin%2C+X%2E+Liu%2C+A%2E+McParland%2C+S%2E+Renals%2C+Os%2E+Saz%2C+M%2E+Wester%2C+et+al%2E%2C+%E2%80%9CThe+MGB+challenge%3A+Evaluating+multi%2Dgenre+Broadcast+media+recognition%2C%E2%80%9D+in+ASRU%2C+20)
16. A. Ali, P. Bell, J. Glass, Y. Messaoui, H. Mubarak, S. Renals, and Y. Zhang, “The MGB-2 challenge: Arabic multi-dialect broadcast media recognition,” in SLT, 2016. [find](https://scholar.google.com/scholar?q=A%2E+Ali%2C+P%2E+Bell%2C+J%2E+Glass%2C+Y%2E+Messaoui%2C+H%2E+Mubarak%2C+S%2E+Renals%2C+and+Y%2E+Zhang%2C+%E2%80%9CThe+MGB%2D2+challenge%3A+Arabic+multi%2Ddialect+broadcast+media+recognition%2C%E2%80%9D+in+SLT%2C+2016%2E) [our card](https://md.research.elamur.ai/free/en/papers/ali-2016-mgb-2-challenge-arabic-multi/)
17. C. K. Reddy, V. Gopal, and R. Cutler, “Dnsmos: A non-intrusive perceptual objective speech quality metric to evaluate noise suppressors,” in ICASSP 2021. [find](https://scholar.google.com/scholar?q=C%2E+K%2E+Reddy%2C+V%2E+Gopal%2C+and+R%2E+Cutler%2C+%E2%80%9CDnsmos%3A+A+non%2Dintrusive+perceptual+objective+speech+quality+metric+to+evaluate+noise+suppressors%2C%E2%80%9D+in+ICASSP+2021%2E)
18. C. Lo, S. Fu, W. Huang, X. Wang, J. Yamagishi, Y. Tsao, and H. Wang, “Mosnet: Deep learning based objective assessment for voice conversion,” in Interspeech 2019. [find](https://scholar.google.com/scholar?q=C%2E+Lo%2C+S%2E+Fu%2C+W%2E+Huang%2C+X%2E+Wang%2C+J%2E+Yamagishi%2C+Y%2E+Tsao%2C+and+H%2E+Wang%2C+%E2%80%9CMosnet%3A+Deep+learning+based+objective+assessment+for+voice+conversion%2C%E2%80%9D+in+Interspeech+2019%2E)
19. P. Andreev, A. Alanov, O. Ivanov, and D. Vetrov, “Hifi++: a unified framework for neural vocoding, bandwidth extension and speech enhancement,” arXiv:2203.13086, 2022. [↗](https://arxiv.org/abs/2203.13086)
20. B. Naderi and R. Cutler, “An open source implementation of itu-t recommendation p. 808 with validation,” Proc. Interspeech 2020. [find](https://scholar.google.com/scholar?q=B%2E+Naderi+and+R%2E+Cutler%2C+%E2%80%9CAn+open+source+implementation+of+itu%2Dt+recommendation+p%2E+808+with+validation%2C%E2%80%9D+Proc%2E+Interspeech+2020%2E)
21. C. K. Reddy, V. Gopal, R. Cutler, E. Beyrami, R. Cheng, H. Dubey, S. Matusevych, R. Aichner, A. Aazami, S. Braun, et al., “The interspeech 2020 deep noise suppression challenge: Datasets, subjective testing framework, and challenge results,” Proc. Interspeech 2020. [find](https://scholar.google.com/scholar?q=C%2E+K%2E+Reddy%2C+V%2E+Gopal%2C+R%2E+Cutler%2C+E%2E+Beyrami%2C+R%2E+Cheng%2C+H%2E+Dubey%2C+S%2E+Matusevych%2C+R%2E+Aichner%2C+A%2E+Aazami%2C+S%2E+Braun%2C+et+al%2E%2C+%E2%80%9CThe+interspeech+2020+deep+noise+suppression+challenge%3A+Datasets%2C+subjective+t)
22. A. Hussein, S. Watanabe, and A. Ali, “Arabic speech recognition by end-to-end, modular systems and human,” Computer Speech & Language, 2021. [find](https://scholar.google.com/scholar?q=A%2E+Hussein%2C+S%2E+Watanabe%2C+and+A%2E+Ali%2C+%E2%80%9CArabic+speech+recognition+by+end%2Dto%2Dend%2C+modular+systems+and+human%2C%E2%80%9D+Computer+Speech+%26+Language%2C+2021%2E) [our card](https://md.research.elamur.ai/free/en/papers/hussein-2021-arabic-speech-recognition-end-end/)
23. A. Abdelali, K. Darwish, N. Durrani, and H. Mubarak, “Farasa: A fast and furious segmenter for Arabic,” in The North American chapter of the association for computational linguistics: Demonstrations, 2016. [find](https://scholar.google.com/scholar?q=A%2E+Abdelali%2C+K%2E+Darwish%2C+N%2E+Durrani%2C+and+H%2E+Mubarak%2C+%E2%80%9CFarasa%3A+A+fast+and+furious+segmenter+for+Arabic%2C%E2%80%9D+in+The+North+American+chapter+of+the+association+for+computational+linguistics%3A+Demonstrations%2C+) [our card](https://md.research.elamur.ai/free/en/papers/aclbulk-2016-abdelali-farasa-fast-furious-segmenter-arabic/)
24. J. Shen, R. Pang, R. Weiss, M. Schuster, N. Jaitly, Z. Yang, Z. Chen, Y. Zhang, Y. Wang, R. Skerrv-Ryan, et al., “Natural TTS synthesis by conditioning wavenet on mel spectrogram predictions,” in ICASSP, 2018. [find](https://scholar.google.com/scholar?q=J%2E+Shen%2C+R%2E+Pang%2C+R%2E+Weiss%2C+M%2E+Schuster%2C+N%2E+Jaitly%2C+Z%2E+Yang%2C+Z%2E+Chen%2C+Y%2E+Zhang%2C+Y%2E+Wang%2C+R%2E+Skerrv%2DRyan%2C+et+al%2E%2C+%E2%80%9CNatural+TTS+synthesis+by+conditioning+wavenet+on+mel+spectrogram+predictions%2C%E2%80%9D+in+ICAS)
25. N. Li, S. Liu, Y. Liu, S. Zhao, and M. Liu, “Neural speech synthesis with transformer network,” in AAAI, 2019. [find](https://scholar.google.com/scholar?q=N%2E+Li%2C+S%2E+Liu%2C+Y%2E+Liu%2C+S%2E+Zhao%2C+and+M%2E+Liu%2C+%E2%80%9CNeural+speech+synthesis+with+transformer+network%2C%E2%80%9D+in+AAAI%2C+2019%2E)
26. J. Chorowski, D. Bahdanau, D. Serdyuk, K. Cho, and Y. Bengio, “Attention-based models for speech recognition,” NIPS, 2015. [find](https://scholar.google.com/scholar?q=J%2E+Chorowski%2C+D%2E+Bahdanau%2C+D%2E+Serdyuk%2C+K%2E+Cho%2C+and+Y%2E+Bengio%2C+%E2%80%9CAttention%2Dbased+models+for+speech+recognition%2C%E2%80%9D+NIPS%2C+2015%2E) [our card](https://md.research.elamur.ai/free/en/papers/chorowski-2015-attention-based-models-speech-recognition/)
27. K. Ito and L. Johnson, “The lj speech dataset,” [https://keithito.com/LJ-Speech-Dataset](https://keithito.com/LJ-Speech-Dataset), 2017. Online: [↗](https://keithito.com/LJ-Speech-Dataset)
28. P. Guo, F. Boyer, X. Chang, T. Hayashi, Y. Higuchi, H. Inaguma, N. Kamo, C. Li, D. Garcia-Romero, J. Shi, et al., “Recent developments on ESPnet toolkit boosted by conformer,” ICASSP, 2021. [find](https://scholar.google.com/scholar?q=P%2E+Guo%2C+F%2E+Boyer%2C+X%2E+Chang%2C+T%2E+Hayashi%2C+Y%2E+Higuchi%2C+H%2E+Inaguma%2C+N%2E+Kamo%2C+C%2E+Li%2C+D%2E+Garcia%2DRomero%2C+J%2E+Shi%2C+et+al%2E%2C+%E2%80%9CRecent+developments+on+ESPnet+toolkit+boosted+by+conformer%2C%E2%80%9D+ICASSP%2C+2021%2E)
29. A. Łańcucki, “Fastpitch: Parallel text-to-speech with pitch prediction,” ICASSP, 2021. [find](https://scholar.google.com/scholar?q=A%2E+%C5%81an%CC%81cucki%2C+%E2%80%9CFastpitch%3A+Parallel+text%2Dto%2Dspeech+with+pitch+prediction%2C%E2%80%9D+ICASSP%2C+2021%2E)
30. N. Perraudin, P. Balazs, and P. L. Søndergaard, “A fast griffin-lim algorithm,” in 2013 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics. [find](https://scholar.google.com/scholar?q=N%2E+Perraudin%2C+P%2E+Balazs%2C+and+P%2E+L%2E+S%C3%B8ndergaard%2C+%E2%80%9CA+fast+griffin%2Dlim+algorithm%2C%E2%80%9D+in+2013+IEEE+Workshop+on+Applications+of+Signal+Processing+to+Audio+and+Acoustics%2E)
31. R. Yamamoto, E. Song, and J. Kim, “Parallel wavegan: A fast waveform generation model based on generative adversarial networks with multi-resolution spectrogram,” in ICASSP, 2020. [find](https://scholar.google.com/scholar?q=R%2E+Yamamoto%2C+E%2E+Song%2C+and+J%2E+Kim%2C+%E2%80%9CParallel+wavegan%3A+A+fast+waveform+generation+model+based+on+generative+adversarial+networks+with+multi%2Dresolution+spectrogram%2C%E2%80%9D+in+ICASSP%2C+2020%2E)