Mathias Müller, Dr.
Tel.: +41 44 635 72 46
I'm a postdoctoral researcher interested in machine translation, unreasonably effective neural networks, phonetics and acoustics, markup technologies. Also, I find teaching super exciting!
What I am currently doing
I am working with Sarah Ebling, Annette Rios and others on automatic sign language translation. I am part of a large EU project called EASIER.
Currently I am most interested in the following:
- Scientific integrity, reproducibility, evaluation protocols, signifiance testing
- Translating sign languages
- Teaching NMT
- Novel decoding algorithms for NMT
- Robustness of NMT systems
Planning a programming project or thesis related to the topics mentioned above? Come talk to me, I'd love to hear about it.
- 03/2022: Our paper on sign language translation for African sign languages was accepted to AfricaNLP 2023!
- 02/2022: I obtained a Bridge Postdoc grant funded by the Digital Society Initiative (DSI) to work on sign language machine translation for 4 more years!
- 02/2022: Our paper on multimodal, multi-task sign language translation was accepted to ICLR 2023.
- 02/2022: Our paper on machine translation using Signwriting was accepted to EACL 2023.
- 12/2022: We have a new preprint article: Considerations for meaningful sign language machine translation based on glosses
- 12/2022: The WMT-SLT22 shared task is now concluded and I presented the outcome at WMT/EMNLP in Abu Dhabi. Findings paper
- 04/2022: I gave an invited lecture in Lena Jäger's machine learning course, on Reproducibility in Natural Language Processing.
- 01/2022: The WMT-SLT22 shared task on sign language translation will be funded by the European Association for Machine Translation (EAMT) and the Microsoft Accessibility division. Thank you to the sponsors!
- 12/2021: Together with others I am launching an exciting new WMT shared task, on sign language translation (WMT-SLT22).
- 11/2021: Invited talk at the NLP Meetup Zurich, topic: A Casual Q&A Session on Machine Translation.
- 10/2021: Invited lecture at the University of Berne, Switzerland, on recurrent neural networks. Recording.
- 10/2021: Invited talk in the SIGTYP lecture series, Exploring a Sampling-based Alternative to Beam Search. More info and recording.
- 09/2021: I finished my PhD! Read the thesis here. I can recommend the background chapters, especially the introduction to NMT.