Transliteration in Hindi

I. INTRODUCTION

Till date, many softwares have been made to transliterate from one language to another. There are two ways in which this is done. One involves transliterating only when a space is typed after the word, as in orkut. The other method involves transliterating simultaneously as the user types in the letters, so that he doesn't have to wait till he types a space, as in Gopi’s Unicode Hindi Converter tool.

The main problem in the transliteration lies in converting the letters which can correspond to different letters in the language to be transliterated. For example, in hindi, we have d which can convert into द and ड़. So, while in the former approach, the word once typed is matched through a dictionary and the word having highest probability is returned; as a word can’t be matched through a dictionary as it is still being typed, a different symbol for each letter is used in the transliterated language. As an example, we can use d for द and d: for ड़.

But this makes it pretty cumbersome for a layman to use different kinds of symbols, for he is quite acquainted with using the similar symbols. So, in this approach, we use the machine learning using the previous and next context of every letter the user types in which, in turn, can be written in two or more ways in transliterated language and thus, provide the most probabilistic letter. This approach solves both our problems of simultaneously transliterating the words and providing user friendly approach.

II. PROPOSED APPROACH

As described earlier, the basic approach is machine learning. We used varied context and different machine learning algorithms of Wagon and Weka speech tools. On the whole, our work can be divided into The Rule Based Step, The Idea Of Machine Learning, and Frequent Words’ Dictionary Matching.

II.I RULE BASED STEP

We can consider some ways in which users write and use them to give the output:

1. If there is an “a” at the end of a word, we can directly append an “a” at the...