A morphological analyser for K’iche’

Resumen

This paper describes the development of a free/open-source computational morphological description for K’iche’, a Mayan language spoken in Guatemala. The language is of the agglutinative morphological type, with both prefixing and suffixing morphology. Both the nominal and verbal morphology are moderately complex. K’iche’ is under-resourced and this is the first publication describing a computational tool for the language, and one of the first publications describing a computational tool for any language of the Mayan group. We use the Helsinki Finite-State Toolkit (HFST) for implementing the finite-state transducer. An automatic evaluation of the coverage of our implementation shows that the coverage is adequate, between 86% and 96% on range of freely available corpora. A manual evaluation gives a recall of over 90% over a hand-annotated test set. Both the analyser and the hand-annotated test set are available under a free/open-source licence.