We are very happy that Professor Eric Gaussier will give an invited talk at the twelfth BioASQ workshop.
Title: Generalisation and Deep Neural Networks: The Case of NLP and IR
Speaker: Eric Gaussier, The University Grenoble Alpes
Abstract: Deep neural networks, such as language models pre-trained on large text collections, currently represent the dominant paradigm in Natural Language Processing (NLP) and Information Retrieval (IR). However, there are still many questions about their performance and how they function. In particular, while they have led to significant improvements in almost all NLP and IR tasks, several studies have highlighted their limitations in terms of generalization, which are linked to their difficulty in correctly handling new collections or new tasks. We will explore these limitations in our presentation and discuss some of the approaches being considered to overcome them.
Biosketch: Eric Gaussier received his Ph.D. in Computer Science from University Paris Diderot in 1995, where he conducted research in statistical machine translation. Following his doctoral studies, he worked at Xerox Research Centre Europe, where he made contributions to the development of algorithms and models for text analysis and information retrieval. He then joined, in 2006, the University Grenoble Alpes as a computer science Professor. His research focuses on machine learning models for natural language processing and information retrieval, as well as on causality. He was Director of the Grenoble Computer Science Lab (LIG) from 2016 to 2020 and is Director of the Grenoble Institute of Artificial Intelligence since 2019.
