Machine Transliteration for Indian languages: Survey

Transliteration is an important aspect when conversing from one language to another. This paper addresses several models designed for the Transliteration of Indian languages. It deals with the methods and approaches used for the languages like Marathi, Hindi, and Punjabi. A Transliteration model is designed on the basis of grapheme, phoneme, correspondence and hybrid methods. Various models have been designed for top-level languages but for Indian languages, the researchers are yet in an infant state as fewer resources are available for them as compared to other languages. Initial research work was rule-based which means rules had to be created for every language taking into consideration of rules and grammar of that language. Later, technology transfer occurred, and alignment models and hybrid models gained popularity. Technological progress in the last few years have improved Transliteration results manifold. Advances in this area contributed to the advancement of communication technologies immensely. To improve data retrieval processing, research developments and Transliteration results, the future scope and requirements of several models are discussed briefly in this paper. The analysis of various models in this paper can be further used to improve research in this field.