The utilization of parallel corpora for the extension of machine translation lexicons

There has recently been an increasing awareness of the importance of large collections of texts (corpora) used as resources in machine translation research. The process of creating or extending machine translation lexicons is time-consuming, difficult and costly in terms of human involvement. The contribution that corpora can make towards the reduction in cost, time and complexity has been explored by several research groups. This article describes a system that has been developed to identify word-pairs, utilizing an aligned bilingual (English-Afrikaans) corpus in order to extend a bilingual lexicon with the words and their translations that are not present in the lexicon. New translations for existing entries can be added and the system also applies grammar rules for the identification of the grammatical category of each word-pair. This system limits the involvement of the human translator and has a positive impact on the time, cost and effort needed to extend a bilingual lexicon.Keywords: alignment; bilingual corpora; corpus; extension; lexicon; machine translation; monolingual corpora; parallel corpor