Computer science professor Iryna Gurevych, who conducts research on artificial intelligence, computational linguistics and machine learning models at TU Darmstadt, has received a major scientific award. In the interview, she explains reasons and takes us through some ongoing projects.
TU Darmstadt: Professor Gurevych, the renowned international scientific Association for Computational Linguistics (ACL) recently accepted you together with eight other scientists from top universities worldwide as new “Fellows”. What was the reason given and how do you rate this award?
Prof. Gurevych: The ACL is the most important international scientific society for computational linguistics. It recognizes me for theoretical and practical contributions to deep learning models for analyzing texts, to computational argumentation analysis, and for contributions around the ACL and the German NLP community.
In general, the ACL Fellowship program recognizes members whose research makes an exceptional contribution to the field of computational linguistics, whether in the form of scientific and technical excellence, service to the community, or service to society. The number of new Fellows appointed in any one year should not exceed 0.2 percent of the current ACL membership.
You are doing research on artificial intelligence, computational linguistics, and machine learning models. Please give a brief insight into your projects and some application examples.
My research group does broad research on methods for automatic language processing. One example is automated question answering in different languages. This involves Big Data on the Internet - to train very large language models that are used in the analyses, but also as a very large knowledge repository from which we obtain answers to important questions in an automated way.
The basis for this is cross-linguistic models of text similarity, which we use to find meaning-similar questions on the one hand and the relevant answer components, if necessary in different languages, on the other. This enables efficient retrieval and reuse of human knowledge. Other examples from our work are intelligent AI language assistants to check facts for their truth content, or to search for evidence in large amounts of scientific literature. Ethical aspects of language processing and AI in general play an increasingly important role in our research.
How is your scientific exchange with international colleagues going?
This encompasses various levels. On the one hand, we exchange ideas at top annual conferences, where my research group, as one of the leading German institutes for language processing, is regularly present with several papers. On the other hand, we cooperate with researchers worldwide, for example with the University of Cambridge, the University of Pennsylvania, the University of Bar-Ilan, but also with companies like Google or DeepMind.
You have also just taken over the office of “Vice-President-elect” in the ACL and will become ACL President in 2023. What is on your to-do list?
The tasks include the preparation and organizational support of ACL 2024 as the most important top conference with currently around 5,000 participants as well as the further development of ACL in terms of content and structure. Prior to Corona, the very strong growth of the professional society, the enormous increase in the number of submissions and related challenges in the review process were on the agenda.
In the meantime, there is also the fact that we fundamentally need new virtual and hybrid formats for the conferences, and associated with this, completely new competencies in virtual communication. This year, for example, we used gather.town as a huge virtual community space for exchange. It was a lot of fun.
The questions were asked by Jörg Feuck. The original interview can be found on the homepage of the TU Darmstadt (German version only).