Noun Class Disambiguation in Runyankore and Related Languages

Proceedings of the 29th International Conference on Computational Linguistics, pages 4350–4359 October 12–17, 2022. 4350 Noun Class Disambiguation in Runyankore and Related Languages Joan Byamugisha IBM Research Africa 45 Juta Street, Braamfontein Johannesburg, South Africa joan.byamugisha@ibm.com Abstract Bantu languages are spoken by communities in more than half of the countries on the African continent by an estimated third of a billion people. Despite this populous and the amount of high quality linguistic research done over the years, Bantu languages are still computationally under-resourced. The biggest limitation to the development of com- putational methods for processing Bantu lan- guage text is their complex grammatical struc- ture, chiefly in the system of noun classes. We investigated the use of a combined syn- tactic and semantic method to disambiguate among singular nouns with the same class pre- fix but belonging to different noun classes. This combination uses the semantic general- izations of the types of nouns in each class to overcome the limitations of relying only on the prefixes they take. We used the nearest neigh- bors of a query word as semantic generaliza- tions, and developed a tool to determine the noun class based on resources in Runyankore, a Bantu language indigenous to Uganda. We also investigated whether, with the same Run- yankore resources, our method had utility in other Bantu languages, Luganda, indigenous to Uganda, and Kinyarwanda, indigenous to Rwanda. For all three languages, the com- bined approach resulted in an improvement in accuracy, as compared to using only the syn- tactic or the semantic approach. 1 Introduction Over the last three decades, there has been an increase in the development of computational re- sources for Bantu languages. Much of this work applies the knowledge gained on Bantu language linguistics to digitize textual resources and develop software tools for text processing for a single lan- guage or for a group of languages. Among the textual resources created are the SAWA English- Kiswahili parallel corpus for machine learning (De Pauw et al., 2011) and the labelled and unlabelled Runyankore datasets (Byamugisha, 2020). The text processing resources include morphological tools, such as morphological analyzers for isiZulu (Bosch and Pretorius, 2003, 2004), isiXhosa, seSwati, and isiNdebele (Bosch et al., 2008); text generation tools, such as a morphological generator for isiZulu (Bosch and Pretorius, 2003) and surface realizers for isiZulu (Keet et al., 2017) and Runyankore (Byamugisha et al., 2017a,b); part-of-speech tag- gers for Kiswahili, Ciluba, Northern Sotho, and isiZulu (De Pauw et al., 2012); and noun pluraliza- tion tools for isiZulu and Runyankore (Byamugisha et al., 2016), chiShona, isiXhosa, Kikuyu, Kin- yarwanda, and Luganda (Byamugisha et al., 2018). Despite these efforts, Bantu languages are still among the most computationally under-resourced languages in the world. This is due to their com- plex grammatical structure, mainly the noun class system, verb morphology, and agglutinative mor- phology. In this paper, we focus on the noun class system only, the hallmark of Bantu nominal mor- phology (Katamba, 2003). This system places ev- ery noun in a language into …

Content truncated to 500 words