Spellchecker for Malayalam using finite state transition models

Finite-state machine is used as a core technology in many fields of natural language processing. The applications include speech recognition and generation, spelling correction, fact extraction, information retrieval and approaches to translation. This paper describes the design of finite state machine in development of a Malayalam spellchecker. Conventional spell checkers has the disadvantage of large dictionary size which can be reduced using finite state models. Morphological factors like word categories and their inflections evolve automatically during training of these models if we train the model with an appropriate training set.