[Lingtyp] SIGTYP 2022 Shared Task on the Prediction of Cognate Reflexes

[Johann-Mattis List]

Dear all,

In 2022, SIGTYP is hosting a shared task on predicting words and 
morphemes from cognate words in related languages (Cognate Reflex 
Prediction Task). We provide 20 cognate-coded multilingual wordlists (10 
for development, and 10 withheld for testing) from which we have 
withheld varying amounts of cognate sets from which the systems provided 
by participants should later predict individual words and morphemes. The 
code we provide allows to transparently access the data from its 
standardized format, compute sample predictions with a baseline method, 
and to evaluate results with three evaluation metrics.

For more information on cognate reflex prediction, we recommend the 
following study (available in open access):

Bodt, T. and J.-M. List (2022): Reflex prediction. A case study of 
Western Kho-Bwa. Diachronica 39.1. 1-38. 
https://doi.org/10.1075/dia.20009.bod

For more information, please check the GitHub repository, which contains 
the data, code to process the data, and additional information.

* https://github.com/sigtyp/ST2022

Important dates are:

* Feb 21, 2022: Training data and baseline are released at 
https://github.com/sigtyp/ST2022/tree/v0.2
* Apr 11, 2022: Test data released at 
https://github.com/sigtyp/ST2022/tree/v1.0
* Apr 25, 2022: System submissions are due
* May 13, 2022: System description papers are due
* May 20, 2022: Camera-ready papers due

To register for the task:

* 
https://docs.google.com/forms/d/e/1FAIpQLSdKvUOOUQSUpbOCWGR6a3zXUTEhhDqTRIXwhlgPwGPrpNpPcQ/viewform?usp=sf_link

We hope for many interesting contributions!

Sincerely,

Mattis List (on behalf of all co-organizers)

-- 
PD Dr. Johann-Mattis List
Research Group Leader "Computer-Assisted Language Comparison"
Department of Linguistic and Cultural Evolution
Max Planck Institute for Evolutionary Anthropology
Deutscher Platz 6
04103 Leipzig
https://lingulist.de
https://digling.org/calc/