English to Hindi Statistical Machine Translation System

Abstract

Machine Translation is an important part of Natural Language Processing. It refers to using machine to convert one natural language to another. Statistical Machine Translation is a part of Machine Translation that strives to use machine learning paradigm towards translating text. Statistical Machine Translation consists of Language Model (LM), Translation Model (TM) and decoder. In this thesis, English to Hindi Statistical Machine Translation system has been developed. The development of Language Model, Translation Model and decoder is done by making use of software’s available in Linux environment. SR International’s Language Model (SRILM) for Language Model, GIZA++ and mkcls for Translation Model, Moses for decoding, has been used in this system. LM computes the probability of target language sentences. TM calculates the probability of target sentences given the source sentence and the decoder maximizes the probability of translated text of target language. A parallel corpus of 5000 sentences in English and Hindi has been used in training of the system. The system was evaluated using manual evaluation method and a geometric average score of 2.693, 2.93 on the parameters of fluency and adequacy respectively, were found.