The lecture entitled “Free lunch”
Many natural language processing tasks require either “knowing how it should be done” via manually encoded linguistic rules or large amounts of explicitly annotated “right answers” and supervised machine learning. Mark Fišel and his research group are focused on finding natural language processing solutions that do not require annotated examples or linguistic knowledge encoding, a “free lunch” of sorts. This research direction aims to support low-resource languages, including Estonian, that do not have as many linguistic resources and annotated data as resource-rich languages like English.
In his lecture Mark Fišel will talk about transfer learning and model parameter reuse in machine learning-based natural language processing, covering topics such as machine translation, translation quality estimation, style adaptation and grammatical error correction.
Mark Fišel is a professor of natural language processing at the Institute of Computer Science, Faculty of Science and Technology at the University of Tartu. His research is focused on multilinguality [AK1] and computational language learning in low-resource settings, as well as language technology solutions for Estonian. He received his PhD in computer science at the University of Tartu in 2011 and did his post-doctoral research at the University of Zurich in 2011–2015. He has been the head of the Chair of Natural Language Processing at the University of Tartu since 2015.