NAIST-SIC-Aligned-ST

This repository stores the Aligned-ST data we used in our paper (https://aclanthology.org/2023.iwslt-1.34/) based on NAIST-SIC-Aligned (https://arxiv.org/abs/2304.11766), the first large scale English-Japanese Interpretation Data.

From NAIST-SIC-Aligned, we selected INTRA, AUTO-DEV and AUTO-TEST as train, dev and test data, respectively. For all the SI sets, we aligned the English text segments with the corresponding audio tracks in MuST-C v2 (https://aclanthology.org/N19-1202/) using an English forced-aligner Gentle (https://github.com/lowerquality/gentle). Here, we excluded segments not aligned with the source speech from the aligned dataset.

Data size (in # of segments):

• Train: 65,083 (*1)
• Dev: 165
• Test: 511

*1: The size difference in "Train" between Table 1 in the paper (https://aclanthology.org/2023.iwslt-1.34/) is due to the setting mismatch and the deficiencies in data preprocessing.

Terms of Use

USAGE: NAIST-SIC-Aligned-ST may be used for research purposes. Commercial use is prohibited.

REDISTRIBUTION: The interpretation transcripts in the corpus shall not be redistributed. It is permissible, however, to cite examples from the corpus for the purpose of presenting research results.

CITATION: All reports/publications using NAIST-SIC-Aligned-ST must acknowledge its use. We recommend that this is done through a citation to the paper describing the corpus and a link to the corpus page:

Yuka Ko, Ryo Fukuda, Yuta Nishikawa, Yasumasa Kano, Katsuhito Sudoh, Satoshi Nakamura.
Tagged End-to-End Simultaneous Speech Translation Training Using Simultaneous Interpretation Data
Proceedings of the 20th International Conference on Spoken Language Translation (IWSLT 2023), pp. 363--375, 2023

NAIST-SIC-Aligned-ST: https://dsc-nlp.naist.jp/data/NAIST-SIC/Aligned-ST/

WARRANTY: The corpus is provided WITH NO WARRANTY.

---

Download

You can download NAIST-SIC-Aligned-ST (4,853,143 bytes, ZIP file) upon the agreement of the Terms of Use above.

I acknowledge my understanding of the Terms of Use above and accept them. All other members of my organization that will use the data agree to abide by these guidelines. I am qualified to make this commitment.

---

Related Paper

Yuka Ko, Ryo Fukuda, Yuta Nishikawa, Yasumasa Kano, Katsuhito Sudoh, Satoshi Nakamura.
Tagged End-to-End Simultaneous Speech Translation Training Using Simultaneous Interpretation Data.
Proceedings of the 20th International Conference on Spoken Language Translation, pp. 363--375, 2023

sudoh [0x40] is.naist.jp