English to Tamil machine translation system using parallel corpus

The present research work makes use of the statistical machine translation approach for English to Tamil rather than the other approaches of machine translation such as rule based and example based. Because of the complexities in other approaches which will be discussed briefly in the later chapters. English to Tamil language translation is built here by making use of Statistical Machine Translation (SMT). Main goal of this system is to undertake translation with minimum human efforts. There are many tools pertaining to LM, TM, decoder for undertaking SMT. SMT has three major parts of the system, Language Model, Translation Model and searching (decoder). The LM computes the probabilities with respect to the target language. The TM computes the probabilities regarding the substitution of target language word with source language word. For development of LM, SRI international¿s SRILM Language Model toolkit is used. GIZA++ is used for creation of Translation Model. For decoding stage, Moses software has been used. The system is based upon Linux operating system. It will accept English sentence from the terminal and produce output in Tamil.