Pluralizing Nouns across Agglutinating Bantu Languages

Proceedings of the 27th International Conference on Computational Linguistics, pages 2633–2643 Santa Fe, New Mexico, USA, August 20-26, 2018. 2633 Pluralizing Nouns across Agglutinating Bantu Languages Joan Byamugisha University of Cape Town Cape Town South Africa jbyamugisha@cs.uct.ac.za C. Maria Keet University of Cape Town Cape Town South Africa mkeet@cs.uct.ac.za Brian DeRenzi Dimagi Cape Town South Africa bderenzi@dimagi.com Abstract Text generation may require the pluralization of nouns, such as in context-sensitive user inter- faces and in natural language generation more broadly. While this has been solved for the widely- used languages, this is still a challenge for the languages in the Bantu language family. Plural- ization results obtained for isiZulu and Runyankore showed there were similarities in approach, including the need to combine morphology with syntax and semantics, despite belonging to dif- ferent language zones. This suggests that bootstrapping and generalizability might be feasible. We investigated this systematically for seven languages across three different Guthrie language zones. The first outcome is that Meinhof’s 1948 specification of the noun classes are indeed inadequate for computational purposes for all examined languages, due to non-determinism in prefixes, and we thus redefined the characteristic noun class tables of 29 noun classes into 53. The second main result is that the generic pluralizer achieved over 93% accuracy in coverage testing and over 94% on a random sample. This is comparable to the language-specific isiZulu and Runyankore pluralizers. 1 Introduction The need for generating plurals to ease the use of software tools is well known in several areas. In Natural Language Understanding, plurals are used during parsing to distinguish between atomic elements (simple nouns) and collectives (Schubert, 2015), while in Natural Language Generation, pluralization is essential during quantification (Schubert, 2015) and number agreement. For instance, to inflect properly for the grammatical number in end-user tools, like a calendar properly stating to set off an alarm “1 hour” before and not “1 hours”, but also for more advanced tools, such as automatically generating patient discharge notes instructing how many pills the patient has to take. Automated pluralization may use regular expressions of the ending of the nouns, alike for English (Conway, 1998), or extensive linguistic resources to devise pluralization, as for German (Nakisa and Hahn, 1996). Pluralization for Runyankore and isiZulu, members of the Bantu language family, showed that neither of the above approaches were feasible. It required a combined morphology with syntactic and semantic approach, where the latter was covered by using a combination of the noun and noun class to pluralize, along with a few similar refinements of Meinhof’s noun class definition (Byamugisha et al., 2016). A surprising observation was that a similar approach could be used despite the regions where these languages are spoken being thousands of kilometers apart (Uganda and South Africa) and they belong to different Guthrie zones (Guthrie, 1948; Maho, 2009). This suggests there may be sufficient similarities not only for closely related languages, but also for languages in other Guthrie zones. We thus aim to investigate the generalizability of pluralization for …

Content truncated to 500 words