Intelligent Fusion of Structural and Citation-Based Evidence for Text Classification

This paper investigates how citation-based information and structural content (e.g., title, abstract) can be combined to improve classification of text documents into predefined categories. We evaluate different measures of similarity, five derived from the citation structure of the collection, and three measures derived from the structural content, and determine how they can be fused to improve classification effectiveness. To discover the best fusion framework, we apply Genetic Programming (GP) techniques. Our empirical experiments using documents from the ACM digital library and the ACM classification scheme show that we can discover similarity functions that work better than any evidence in isolation and whose combined performance through a simple majority voting is comparable to that of Support Vector Machine classifiers.