Dominik Macháček
Main Research Interests
Neural Machine Translation
Multilingual Neural Machine Translation
Simultaneous Speech Translation
Projects
NEUREM3
Past:
GAUK Strojový překlad tlumočených projevů / Machine translation of interpreted speech
Curriculum Vitae
2022: Internship at National Institute of Information and Communications Technology (NICT), Kyoto, Japan. Slides.
2021: PhD. thesis proposal "Multi-Lingual Machine Translation". Text, slides.
2019-2023: PhD at ÚFAL
2018: Internship at Unbabel, Portugal
...
Selected Bibliography
- Google Scholar
- ORCID: 0000-0002-5530-1615
- Scopus ID: 57203994469
- Researcher ID: GPK-8318-2022
Dominik Macháček, Ondřej Bojar, Raj Dabre (2022): MT Metrics Correlate with Human Ratings of Simultaneous Speech Translation. Technical report. (pdf)
Dávid Javorský, Dominik Macháček, and Ondřej Bojar. 2022. Continuous Rating as Reliable Human Evaluation of Simultaneous Speech Translation. In Proceedings of the Seventh Conference on Machine Translation (WMT), pages 154–164, Abu Dhabi, United Arab Emirates (Hybrid). Association for Computational Linguistics. (pdf)
Lost in Interpreting: Speech Translation from Source or Interpreter?. In: Proceedings of INTERSPEECH 2021, pp. 2376-2380, ISCA, Baxas, France (pdf, bibtex)
ESIC 1.0 -- Europarl Simultaneous Interpreting Corpus
Kocmi, Macháček, Bojar, 2021: The Reality of Multi-Lingual Machine Translation
Summaries
2023
I presented my two most recent papers at ACL and IWSLT 2023 in Toronto. See the page that links all the materials I presented.
2020-2022
On 14.12. 2022, I visited prof. Nakamura's lab at NAIST (Nara Institute of Science and Technology, Japan) and gave a one-hour talk about my dissertation topic Simultaneous Multi-Source Speech Translation from the original and simultaneous interpreting. In this talk, I explain motivation, specification of my focus on this task, and explain the benefits and risks of interpreting in speech translation. I describe current state of the art from ELITR project, ESIC evaluation corpus, and preliminary results and next plans. In the second part of the talk, I introduce Continuous Rating as a reliable method of human evaluation of simultaneous speech translation, and its correlation to offline MT metrics.
This talk summarizes my PhD research in 2019-2022. The slides are available.
2017-2020
My master thesis Enriching Neural MT through Multi-Task Training, the follow-up paper and my experience from ELITR project is summarized in a book The Reality of Multi-Lingual Machine Translation, along with my colleagues' contributions. The pdf version and free copies are available.