Goran Glavaš has received the best (long) paper award at one of the most renowned conferences in NLP, the 16th Conference of the European Chapter of the Association for Computational Linguistics (EACL) for the paper “Is Supervised Syntactic Parsing Beneficial for Language Understanding Tasks? An Empirical Investigation”. The work is a collaboration with Dr. Ivan Vulić from the Language Technology Lab, University of Cambridge.
The paper investigates the role of explicit syntactic supervision (i.e., inductive linguistic biases in the form of manually annotated treebanks) in the times of large pretrained models (BERT & co.) and whether such signal still benefits downstream natural language understanding (NLU) tasks. The work examines both monolingual English setups as well as cross-lingual downstream transfer by means of multilingual LMs.