Translation-Based Dictionary Alignment for Under-Resourced Bantu Languages

Abstract

Despite a large number of active speakers, most Bantu languages can be considered as under- or less-resourced languages. This includes especially the current situation of lexicographical data, which is highly unsatisfactory concerning the size, quality and consistency in format and provided information. Unfortunately, this does not only hold for the amount and quality of data for monolingual dictionaries, but also for their lack of interconnection to form a network of dictionaries. Current endeavours to promote the use of Bantu languages in primary and secondary education in countries like South Africa show the urgent need for high-quality digital dictionaries. This contribution describes a prototypical implementation for aligning Xhosa, Zimbabwean Ndebele and Kalanga language dictionaries based on their English translations using simple string matching techniques and via WordNet URIs. The RDF-based representation of the data using the Bantu Language Model (BLM) and - partial - references to the established WordNet dataset supported this process significantly.