A Sinhala and Tamil Extension to Generic Environment for Context-aware Correction

摘要

There are several types of research available on spell checkers for European languages and Indian languages. However, low resourced languages like Tamil & Sinhala have limited research in this problem space, maybe, because of its highly inflectional and morphologically rich nature. There is no fully functional context-aware spell-checking system, especially as an open source. A Generic Environment for context-aware spell correction approach is extended for resource-scarce languages: Sinhala and Tamil in this paper. Experimental results show that our system detects the error in spelling well and provides the most suitable suggestions for correcting the misspelled words with a minimum of 85% accuracy for Tamil and 70% for the Sinhala Language. This is the first ever context-aware spell corrector for the Sinhala language. Compared to prior Tamil language context-aware spell correctors this leaps in 1) modularized architecture and 2) increased coverage and accuracy. Moreover, this study produced a Tamil and Sinhala spell correction benchmark dataset. Both the dataset and the tools are available for public use.