An approach to integrate translation memory in MT system for English to Indian language

Although there is a requirement for English to Indian language translation, there are no proper and enough tools for the translators to use, because of the complexities involved in it. The research in developing Machine translation systems is going on but they are not yet to the stage of product or useful systems development. Major reason is the word level, phrase level and semantic level translation complexities involved because of the different cultures and use of the languages and non availability of good quality Translation Memories. Translation memories are widely used by translation industry because of their ease of use, ability to create glossaries, terminologies and phrasal dictionaries in a domain. It also helps maintain writing styles throughout the series of documents. Also, Human Machine Interface is better for human translators. On the other hand Machine Translation (MT) systems when tuned properly produce better translations for simpler sentences. Complex and compound sentences are difficult for MT to translate solely by machines. Post editing by human is required. This paper proposes integration of these two methods to get the respective advantages of both the systems providing better translation support systems for translators to work faster without compromising quality and help human translator to maintain proper writing styles intact. The concept was validated with a small experiment using a widely used TM tool (Wordfast) and English to Indian Language Rule Based Machine Translation system provided by IIT Kanpur.