Semantic Based Intuitive Topic Search Engine
P. Ajitha, G. Gunasekaran · International Review on Computers and Software (IRECOS) · 2014
This paper deals with the classification which involves supervised learning. In other words classification is made possible by building a knowledge base against which any data could be categorized. The scope is finite in this case and hence building the knowledge base needs to be precise else the classification would be inefficient. Past trends have suggested many useful methodologies and at the same time it has its inherent problems. This paper captures few of those and has made an attempt to improvise on the same. Conventional classification is just based on feature extraction frequency and not on its relevance and it has its own limitations like polysemy and synonymy. Proposed model makes an attempt on those lines to mitigate the same. We have proposed a model which involves Semantic based approach to extract only the relevant features with more frequency. This is based on the pattern extracted from a set of documents, rather than just from a single document. This averages out any inconsistency present in the set of documents and thereby creating a search space that is balanced and cohesive in nature. The training of the data has been done using PMI score, which is tested using set of novel documents.