Semantic Framework to Text Clustering with Neighbors

Abstract

Conventional document clustering techniques use bag-of-words to represent documents, an often unsatisfactory representation, as it ignores the relationships between words that do not co-occur literally. Including semantic knowledge in text representation we can establish the relations between words and thus result in better clusters. Here we apply neighbors and link concept with semantic framework to cluster documents. The neighbors and link provides the global information to compute the closeness of two documents than simple pair wise similarity. We have given a framework to represent text documents with semantic knowledge and proposed Shared Neighbor Based Semantic Text Clustering algorithm. Our experiments on Reuters, Classic and real-time datasets shows significant improvement in forming coherent clusters.