Text Categorization with Semantic Commonsense Knowledge

Most of text categorization research exploit bag-of-words text representation. In this approach, however, all contextual information contained in text is neglected. Therefore, capturing semantic similarity between text documents that share very little or even no vocabulary is not possible. In this paper we present an approach that combines well established kernel text classifiers with external contextual commonsense knowledge. We propose a method for computing semantic similarity between words as a result of diffusion process in ConceptNet semantic space. Evaluation on a Reuters dataset show a substantial improvement in precision of classification.