Transliterating Bangla to English: a rule-based phonetic approach and a suggestion system

Abstract

Transliteration of Bangla to English is a major problem in Bangladesh. People use various spellings of the same word creating confusion for chat, social media posts, and other things. In this paper, I have developed a new way of transliterating Bangla words into English letters using a phonetic approach. At first, I gathered 110,000 Bangla words from various sources. Then, I converted Bangla words into the International Phonetic Alphabet (IPA) to capture their sound. Subsequently, I converted IPA symbols into English letters using an IPA chart for English dialect. Then, I made modifications to some words based on certain rules proposed by my linguists keeping user-friendliness in mind. I also took experts’ advice on my transliterations to make sure these are correct and sound natural. Next, I developed a suggestion system where for one misspelled transliterated word, it will give n number of possible correct words. I wanted to have a proper balance of linguistical accuracy with user-friendliness. This data can help in language learning, computer programs that understand language, and working with digital text.