UCT researchers scoop 1st place at CICLing 2016 Conference

University of Cape Town researchers scoop 1st place in Verifiability, Reproducibility and Working Description Award at CICLing’16 conference

The software, data-sets, and their description accompanying the paper “Pluralising Nouns in isiZulu and Related Languages” received First Prize for the in Verifiability, Reproducibility and Working Description Award at the 17th International Conference on Intelligent Text Processing and Computational Linguistics that was held last week in Konya, Turkey. The criteria that reviewers take into account are, principally:
“the clarity, simplicity, completeness, and overall quality of the code accompanying the paper that allows to verify and exactly reproduce the claims of the paper”. That is, the software and the experimental results show easily and unambiguously the claims made in the paper, and anyone wishing to reuse the code can do so. (More detailed motivations for the award are described at the conference’s webpage)

The paper itself, co-authored by Joan Byamugisha, Maria Keet (both at UCT), and Langa Khumalo (UKZN), describes the design and implementation of an algorithm to automatically pluralise isiZulu nouns in the singular, which does not neatly follow the ‘standard’ noun class table. Several iterations of experimentation, error analysis, and updating the code with new pluralisation rules resulted in over 90% correctness of the plurals for isiZulu. The underlying patterns that resulted in new pluralisation rules were also observed in a related language, Runyankore (spoken in Uganda), suggesting generalisability of the approach and algorithms. The software that automates the pluralisation for isiZulu consists of a Python script for isiZulu and a java implementation for Runyankore, and can be downloaded from the project’s home page, as well as the data-sets, data analysis, and the paper itself.