UdS submission to the shared task of CoNLL 2019

The UdS submission to the 2019 Conference on Computational Language Learning (2019) shared task, Cross-Framework Meaning Representation Parsing (MRP 2019), placed fourth overall out of 18 teams, as well as 1st overall on both the PSD graph bank and the Little Prince subset. The team was one of nine to successfully submit results for all graphbanks. The team was comprised of current LST Masters students Matthias Lindemann, Mario Mina, and Pia Weißenhorn; Ph.D. candidate Jonas Groschwitz; postdoc Lucia Donatelli; former postdoc Meaghan Fowlie (now professor at Utrecht University); and professor Alexander Koller.
The team presented a compositional neural parser based on previous group work (Lindemann et al., 2019; Groschwitz et al., 2018), which used a decomposition algorithm to successfully parse across all graph banks in the shared task. The work was presented at the CoNLL conference in Hong Kong, November 2-4, and published in the conference proceedings.