A Hybrid Approach Using Phrases and Rules for Hindi to English Machine Translation

Abstract

The present work focuses on developing a hybrid approach for developing a machine translation (MT) scheme for automatic translation of Hindi sentences to English. Development of machine translation (MT) systems for Indian languages to English almost invariably suffers from the limited availability of linguistic resources. As a consequence, statistical, rule-based or example-based approaches have not been found to generate good quality translation for all types of sentences. Moreover, purely statistical or example based schemes do not leverage the semantic association (i.e. association of the noun case-endings (karakas)) with each other and also with the verb phrase to generate the English translation. The scheme proposed in this work is based on the syntactic and semantic roles of the phrases of an input Hindi Sentence. The proposed solution is a hybrid scheme involving phrase-based, rule-based and statistical translation approaches. In our approach we try to identify the difficulties associated with the translation of different types of input sentences, viz. simple declarative, negative and interrogative sentences and propose a translation approach for them.

English-to-Hindi Mapping of Compositional Verb-Particle Constructions

“…English and Hindi, including some other Indian languages, have been compared in the context of divergence studies in recent decades, particularly in the context of machine translation applications (Dave et al, 2001;Sinha & Thakur, 2008;Gupta & Chatterjee, 2003;Chatterjee & Balyan, 2011 among others). However, we have not found a comparative study of EVPCs with Hindi (except for an initial study by Chatterjee and Balyan, 2011).…”