Domain Ontology Guided Feature-Selection for Document Categorization

We present a novel method employing a hierarchical domain ontology structure to select features representing documents. All raw words in the training documents are mapped to concepts in a domain ontology. Based on these concepts, a concept hierarchy is established for the training document space, using is-a relationships defined in the domain ontology. An optimum concept set may be obtained by searching the concept hierarchy with an appropriate heuristic function. This may be used as the feature space to represent the training dataset. The proposed method aims to solve some drawbacks suffered by text classification algorithms and feature selection algorithms. One major difficulty for text classification algorithms, especially for machine learning approaches, is the high dimensionality of the feature space. The second major difficulty is to obtain a training dataset of good quality, which is crucial to the performances of almost all text classifiers. Experimental results show that our method solves these problems more reasonably and more effectively than existing methods.