Towards a Pictorially Grounded Language for Machine-Aided Translation.
Samir Kumar Borgohain, Shivashankar B. Nair · 2010
In this paper, we present a translation system that can aid people not well versed in each other’s languages to communicate their ideas through an intermediate pictorial knowledge representation system called the Pictorially Grounded Language (PGL). Machine-aided translation augmented by PGL culminates in symbols of both the languages – the source and the target being grounded on a common set of images and animations. The system discards the conventional intermediate language representation and substitutes it with the PGL. Since PGL is a graphic language comprising of a sequence of images and animations and since both the parties can see these being rendered, they remain convinced that the ideas emanating in one language (source) are the same as that being generated in the other (target), thus ensuring reliability of translation. The mechanism of translation is also scalable and can be used to translate a language into a larger set of languages while at the same time is capable of preserving their inherent meanings. The paper also presents the results of translations augmented by the corresponding PGL versions that tend to improve the comprehension in the target language.