Contextual spell checking for Tamil Language

Abstract

A spell checker is a basic necessity for composing text in any language. However, for the language like Tamil, due to its complex nature, a little amount of reportable work have been done compare to languages like English and other Latin based languages. There are no fully functional spell checking systems available for Tamil language, especially as an open source software. This paper proposes a contextual spell correction feature to an existing spell checker that is developed using a hybrid approach by the authors. The existing spell checker is integrating the dictionary check, Canti check, crowd sourcing and suggestion generation. In addition to that a contextual spelling correction approach in which, mistakes that arise due to confusion in following sets of letters {Ḻ akaram, ḷ akaram, lakaram}, {ṇ akaram, nakaram, ṉ akaram} and {ṟ akaram, rakaram} is proposed in this paper. A bigram language model is used to make suggestions for the words with confusing letters. Performance of the proposed approach is 89.13% using a human evaluation.