Mobile Application for Transliteration of Roman Script to Gurmukhi Script

Abstract

Language transliteration is one of the important areas in Natural Language Processing. Accurate transliteration of named entities plays an important role in the performance of machine translation and cross-language information retrieval processes. As Punjab has biggest NRI population and they frequently visit to their motherland. The new generation of these NRI people does not have understanding to the Punjabi Language. They know the words but grammatically they don’t know how to write Punjabi. With the help of this mobile application, NRI Punjabi can send a SMS to the transliteration service and receive the SMS as output in the desired Language (Punjabi). There are many challenges in transliteration of Roman script to Gurmukhi script, because there is large character gap in both the scripts. As in Roman script, there are 21 consonant and 5 vowels only, but in Gurmukhi script there are 41 consonants and 19 vowels. So it is a bit difficult to map these characters. There are different approaches for transliteration like rule based approach, statistical approach, etc. Rule based approach is used for the transliteration purpose, in this system and is discussed in this thesis. J2ME Environment is used to develop this mobile application and is connected to a Web Application, which in turn transliterate the user input. For Transliteration Purpose, more than 24 rules are being developed, which are used by web application while transliterating. Experiments conducted on application have shown that the performance is sufficiently good. Most commonly used words are taken from daily life and processed by this application. Output generated by processing these words is shown in chapter 5. Accuracy of this system comes out to be 80% for these daily use sentences. The wrong word creation was mainly because of ੱ (Adhak), ੱ (Bindi) and ੱ (Tippi). The accuracy of system can be further improved with the help of dictionary.