Emergent Self Organizing Maps for Text Cluster Visualization by Incorporating Ontology Based Descriptors

摘要

Despite various advantages of traditional feature vector model for document representation, the well-known inherent deficiency in this model is "sovereign term assumption". This deficiency makes it impossible to identify syntactically different but semantically related terms. In this paper, we demonstrate the use of semantic similarity measure for quantifying the relationship between related terms. Identifying such relationships help in reducing the difference between related documents. In this work, we use only noun terms for enriching the representation model. The natural visualization of clusters is investigated in this study using Emergent Self Organizing Map (ESOM). Experimental results show that incorporation of semantic relationship enhances the accuracy of clustering results.