Multilingual language models
When we communicate in a foreign language, we often use knowledge from our mother tongue for better understanding. Computer models do not consider differences between languages, but still perform well in cross-language comprehension. Beinborn: "My research analyses whether multilingual models can reflect transfer effects and aims to develop models to support language learning."
Transfer effect
Beinborn: "For example, I am learning Dutch and I rely heavily on similarities with my mother tongue German. For me, it's easy to guess that bread in German is 'Brot', but this can be more difficult for someone with a background in Romance, Slavic or non-Indo-European languages. If we were able to model this process computationally, we could predict human behaviour and anticipate learning obstacles such as false friends - words that occur in two or more languages and are very similar and/or sound but mean something completely different. Multilingual models that are sensitive to these differences can help people with different language backgrounds understand each other better."
NWO Veni
Together with Vidi and Vici, Veni is part of the NWO Talent Programme (formerly: the Innovational Research Incentives Scheme). Veni is aimed at researchers who have recently obtained their PhD. Within the Talent Programme, researchers are free to submit their own subject for funding. In this way, NWO stimulates curiosity-driven and innovative research. NWO selects researchers on the basis of the quality of the researcher, the innovative character of the research, the expected scientific impact of the research proposal and possibilities for knowledge utilisation.
VU researcher Lisa Beinborn receives NWO Veni grant
11 April 2022
VU researcher Lisa Beinborn (Natural Language Processing) has been awarded the Veni grant for her research proposal " Interpretability of Transfer in Multilingual Models". She is one of the 78 promising young scientists who have received an NWO Veni grant up to 280,000 euros. How has oil changed the relationships in the world? How does optimism arise? But also new methods to print large pieces of living muscle tissue and how to remove nanoplastics from drinking water. Just like every year, it is again a broad spectrum of research themes on which the newest Veni laureates will focus the coming years.