Automated Mining Of Names Using Parallel Hindi-English Corpus

Abstract

Machine transliteration has a number of applications in a variety of natural language processing related tasks such as machine translation, information retrieval and question-answering. For automated learning of machine transliteration, a large parallel corpus of names in two scripts is required. In this paper we present a simple yet powerful method for automatic mining of Hindi-English names from a parallel corpus. An average 93% precision and 85% recall is achieved in mining of proper names. The method works even with a small corpus. We compare our results with Giza++ word alignment tool that yields 30% precision and 63% recall on the same corpora. We also demonstrate that this very method of name mining works for other Indian languages as well.