A Knowledge-Based Semantic Kernel for Text Classification

Typically, in textual document classification the documents are represented in the vector space using the “Bag of Words” (BOW ) approach. Despite its ease of use, BOW representation cannot handle word synonymy and polysemy problems and does not consider semantic relatedness between words. In this paper, we overcome the shortages of the BOW approach by embedding a known WordNet-based semantic relatedness measure for pairs of words, namely Omiotis, into a semantic kernel. The suggested measure incorporates the TF-IDF weighting scheme, thus creating a semantic kernel which combines both semantic and statistical information from text. Empirical evaluation with real data sets demonstrates that our approach successfully achieves improved classification accuracy with respect to the standard BOW representation, when Omiotis is embedded in four different classifiers.