Comparative study on linear and non linear models for spell correction
sathya bapuju Sai, K Abirami · 2014
Machine learning techniques are provided with small amount of data to learn and training models are expected to evolve in course of time with continuous and incremental learning. Each technique has different requirements and performance. In this article, an evaluation of two different machine learning techniques and how these could be used for isolated spelling correction have been explained. Measures of similarities based on distance and probability is used as features to train the model. Learning can be made specific to every language model by providing necessary data from the domain. Index Terms – Machine Learning, Line of best fit, Measures of similarity, n-grams, Levenshtein distance.