Disambiguating Prepositions for Machine Translation using Lexical Semantic Resources

In this paper, the problem of mapping prepositions in the context of machine translation is discussed. For instance, each sense of a polysemous English preposition maps to a different lexical postposition in Hindi. In a fully automatic machine translation system this problem may be resolved in at least two ways-using transfer rules with semantic features or mapping the preposition to an underlying concept and finding its lexical equivalent in the target language. The two methods are compared and a solution is suggested for disambiguation of prepositions. The paper also discusses the lexical semantic resources available like WordNet and other lexicons which may be used for our purpose. The need for developing such resources for Indian languages is also emphasized.