A probabilistic spellchecker

Abstract

This thesis studies graphical Models applied to the correction of words. More particularly, hidden Markov models and Markov chains will be used in order to build a probabilistic spellchecker. Several ways to cluster words will be introduced: the batch
K-Means clustering algorithm with a specific distance measure and the Expectation- Maximization algorithm in order to learn a mixture of Markov chains. Moreover, a
solution for dealing with the suffixes and prefixes will be presented.

Date of Award	Sept 2000
Original language	English
Awarding Institution	Aston University