Converting Afrikaans to Dutch for technology recycling

HLT resource development for a resource scarce language (L2) can be expedited by recycling existing technologies for a closely related language (L1). To improve the success of L1 technologies on L2 data, one can convert L2 data to make it appear more L1-like. We explore this possibility by developing an Afrikaans-to-Dutch lexical conversion module and using it as pre-processing step before applying a Dutch part of speech tagger to Afrikaans data. The accuracy of the Dutch tagger increased from 62.6%, when tagging raw Afrikaans data, to 80.6% when tagging converted Afrikaans data. We therefore conclude that, at least in the case of Dutch and Afrikaans, the use of lexical conversion as a pre-processing step for technology recycling merits further investigation. Keywords-technology recycling; lexical conversion; Part of speech tagging; Afrikaans; Dutch