Shared Tasks

The VarDial workshop has a history of hosting well-attended shared tasks on various dialects and languages. In 2024, we will organize a classic language identification shared task as well as an unshared task on causal commonsense reasoning.

Please fill out this registration form to participate!

The DIALECT-COPA shared task on dialectal causal commonsense reasoning

The shared task invites the community to propose, develop, and test approaches for adapting models for causal commonsense language understanding to three dialects of South-Slavic languages: the Slovenian Cerkno dialect, the Croatian Chakavian dialect, and the Serbian, Macedonian and Bulgarian Torlak dialect. Training and development data based on the COPA (Choice of plausible alternatives, Roemmele et al. 2011) dataset are available for four related standard languages (Slovenian, Croatian, Serbian, Macedonian) and two out of the three testing dialects (Cerkno, Torlak), the Chakavian dialect serving as a surprise dialect.

Details


Organizers:
Nikola Ljubešić, Jožef Stefan Institute, Ljubljana

Ivan Vulić, University of Cambridge

Goran Glavaš, University of Würzburg

DSL-ML - Multi-label classification of similar languages

The DSL-ML task is a multi-label extension of the classic "Discriminating similar languages" task that has been popular with VarDial since the beginnings of the workshop. The motivation behind this new task formulation is that some texts do not present any linguistic markers to unambiguously determine their origin. It therefore makes sense to predict several possible labels for such texts.

The 2024 DSL-ML task is based on multi-label conversions of existing datasets from five different macro-languages: English, Spanish, Portuguese, French and BCMS (Bosnian, Croatian, Montenegrin, Serbian).

Details

Organizers:
Adrian Chifu, Aix-Marseille University
Radu Ionescu, University of Bucharest
Aleksandra Miletić, University of Helsinki
Filip Miletić, University of Stuttgart
Yves Scherrer, University of Oslo

Important dates

Paper submission: https://softconf.com/naacl2024/VarDial2024/