Human-Centered Data Science
We are a research group founded by Prof. Lisa Beinborn. We work on natural language processing (NLP) with a human-centered perspective and are affiliated with the Computer Science Institute and the Campus Institute Data Science at the University of Göttingen.
Team
Our team is growing. Come join us! |
 |
 |
Prof. Lisa Beinborn Group Head |
Victor Zimmermann PhD Student |
Jana Hackethal Team Assistant |
 |
 |
 |
| Miyu Oba Guest Researcher |
Jonathan Kamp Associated PhD |
Jenia Kim Associated PhD |
News
- Lisa Beinborn was invited to give the ceremonial lecture at the university's new years reception. She spoke about "Sprache modellieren — Das Zusammenspiel von Informatik, Linguistik und Psychologie"
- Register for our HumanCLAIM workhop taking place in Göttingen on March 26th/27th.
- Lisa Beinborn has been awarded an Impulsprofessur grant to work towards poly-vocal language models that can account for cross-lingual and individual differences.
- Our group has been very active at EMNLP 2024 in Miami:
- We explored the frequency bias of current language models and propose a new approach called Syntactic Smoothing that reduces both frequency bias and anisotropy of the representational space.
- Our group won both of the outstanding paper awards at the ConLL BabyLM Challenge. Lisa Beinborn cooperated with researchers from Cambridge to analyze the capabilities of a language model that learns from speech-like input represented as phonemes. Our guest researcher Miyu Oba worked on variation sets. Together with a Groningen-Tokyo-Nara combo she systematically used sets of varied sentences expressing a similar intent during language model pre-training.
- Jenia Kim presented her ideas for adaptive simplification of municipal texts at the TSAR workshop.
- Miyu Oba showed that language models still have a hard time inducing grammatical knowledge from indirect evidence.
- We explored the frequency bias of current language models and propose a new approach called Syntactic Smoothing that reduces both frequency bias and anisotropy of the representational space.
- Lisa Beinborn and Nora Hollenstein wrote a book on Cognitive Plausibility in Natural Language Processing.
Research
We develop computational models that account for human variation, uncertainty, and cognitive complexity and are currently most interested in the following topics:
- Multilingual NLP
- Cognitively plausible representation learning
- Interpretability and bias of machine learning models
- Educational language technology
- Modelling eye-tracking data of language processing
Teaching
We are always looking for motivated students to work with us. If you like our courses and want to dig deeper, reach out to claplab@uni-goettingen.de. Have a look at our thesis topics or pitch your own idea based on a recent paper or a shared task.