Singlish to Sinhala Transliteration using Rule-based Approach

Machine translation is a difficult task in natural language processing. All text mining systems face major challenges in terms of lexicon, proper noun handling, and high-tech terminology. This paper describes a transcription technique for text mining in social media posts from English to Sinhala. This study proposes a system for Singlish to Sinhala transliteration that employs a rule-based approach with an error correction module. This method can both transliterate Sinhala words written in English characters (Singlish) and filter out original English words from the text. When compared to the Google input tool, the system has higher accuracy and a lower error rate. Thus, this study would broaden the transliteration knowledge base by introducing a novel set of transliteration rules and employing an ensemble method with a news corpus to improve transliteration accuracy.