International Journal of Innovative Research and Development

Abstract

A lot of problems in natural language processing, data mining, information retrieval, and bioinformatics can be legitimated as string transformation.The task of the string transformation is once the  input string is given, the system generates the k most likely occuring output strings resultant to the input string.So this paper proposes a novel and probabilistic approach to string transformation which includes the use of a log linear model, a training method for the model and an algorithm for generating the top k candidates using a non-dictionary approach which helps the approach to be accurate as well as efficient. The log linear model can be stated as a conditional probability distribution of an output string along with a rule set for the transformation conditioned on an input string. The learning method employs maximum likelihood estimation for parameter estimation. The string generation is based on pruning algorithm which is guaranteed to generate the optimal top k candidates. The proposed method is applied to correction of spelling errors in string or queries. 

Keywords

Full Text