SIGMORPHON 2020 Shared Task 0: Typologically Diverse Morphological Inflection Ekaterina Vylomova@ Jennifer WhiteQ Elizabeth SaleskyZ Sabrina J. MielkeZ Shijie WuZ Edoardo PontiQ Rowan Hall MaudslayQ Ran ZmigrodQ Josef ValvodaQ Svetlana ToldovaE Francis TyersI;E Elena KlyachkoE Ilya YegorovM Natalia KrizhanovskyK Paula CzarnowskaQ Irene NikkarinenQ Andrew KrizhanovskyK Tiago PimentelQ Lucas Torroba HennigenQ Christo Kirov5 Garrett Nicolaiá Adina WilliamsF Antonios Anastasopoulosì Hilaria CruzL Eleanor Chodroff7 Ryan CotterellQ;D Miikka Silfverbergá Mans HuldenX @University of Melbourne QUniversity of Cambridge ZJohns Hopkins University EHigher School of Economics MMoscow State University KKarelian Research Centre 5Google AI áUniversity of British Columbia FFacebook AI Research ìCarnegie Mellon University IIndiana University LUniversity of Louisville 7University of York DETH Zürich XUniversity of Colorado Boulder
[email protected] [email protected] Abstract 1950 and more recently, List et al., 2016), gram- matical features, and even abstract implications A broad goal in natural language processing (NLP) is to develop a system that has the capac- (proposed in Greenberg, 1963), each language nev- ity to process any natural language. Most sys- ertheless has a unique evolutionary trajectory that tems, however, are developed using data from is affected by geographic, social, cultural, and just one language such as English. The SIG- other factors. As a result, the surface form of MORPHON 2020 shared task on morpholog- languages varies substantially. The morphology ical reinflection aims to investigate systems’ of languages can differ in many ways: Some ability to generalize across typologically dis- exhibit rich grammatical case systems (e.g., 12 tinct languages, many of which are low re- in Erzya and 24 in Veps) and mark possessive- source.