Framework for Many To One Machine Translation

Abstract

The frequent domestic and international exchanges have created an opportunity for machine translation toflourish since human translators cannot cater for thetranslationdemand. However, there is slow pace in the development of machine translation tools using the one (source language) to one (target language) framework. The paper proposes a many (source languages) to one (target language) conceptual framework that will ensure faster and efficient development of machine translation tools using the Interlingua rule based machine translation approach. The many to one framework make use of shallow structure: lexical similarity and syntacticsimilarity and deep structure: a unique universal intermediate representation language. A formal computational grammar willbeneededand is exemplified by use of two under resourced languages: Swahili and Kikamba. Evaluation model is proposedat the end.