Second Call for Participation: Shared Task in Parsing into Uniform Meaning Representation (UMR)

https://ufal.mff.cuni.cz/umr-parsing

Uniform Meaning Representation https://umr4nlp.github.io/web/ (UMR) is designed as a typologically-aware successor of AMR, with various modifications and additions, including explicit token-node alignment and document-level relations. Parsers previously proposed for AMR are thus not directly applicable to UMR, although they may serve as a baseline for development of UMR parsers. The few attempts at UMR parsing published so far focused mostly on English. The first shared task in UMR parsing aims to perform evaluation of parsing systems on multiple languages and to assess the current state of the art.

The shared task is collocated with the DMR 2006 workshop https://dmr2026.github.io/ (and with LREC), to be held on 11 May 2026 in Palma de Mallorca, Spain.

The shared task is now in its development phase, with training data available for six languages (Arapaho, Chinese, Czech, English, Latin, Navajo). The test phase is schedule to run between 16 and 27 February 2026. Registration is open until the end of the test phase. For more details, see the shared task homepage at https://ufal.mff.cuni.cz/umr-parsing.

Dan Zeman, Jan Štěpánek & the UMR community