Learning transformation rule in heterogeneous environment using log based posterior model

Transformation is a powerful technique which provides an effective way to access a website. Search engines today receives a list of natural language terms and retrieves the documents similar to the original query. String transformation systems have been widely used in several applications of natural language processing and it is used to identify the similarity between two elements or two strings. Measuring similarity or distance between two strings is a fundamental machine learning problem in K-nearest neighbor problems, clustering domain and neural networks. Learning the domain specific application leads to generate the transformation rule automatically and identify the best rule by appending the weight based on frequency of occurrence of data. This method of assigning weight has some inconvenience which is replaced by employing the probabilistic machine learning approaches. Nowadays generating and learning the rules from heterogeneous environment is required due to increase in web access. Moreover, existing learning algorithms are not suitable to learn the interrelated keywords. This paper introduces a new model and algorithm to generate the rule with probability estimation of weight from the training data and to learn the rules.