Automated Taxonomy Building by Adopting Discriminant and Characteristic Capabilities

Taxonomies are becoming essential in several fields, playing an important role in a large number of applications, particularly for specific domains. Taxonomies provide efficient tools to people by organizing a huge amount of information into a small hierarchical structure. Taxonomies were originally built by hand, but nowadays the technology permits to produce a vast amount of information. Consequently, recent research activities have been focused on automated taxonomy generation. In this paper, we propose a novel approach for automatically build a taxonomy, starting from a set of categories. We deem that, in a hierarchical structure, each node should intuitively be represented with proper meaningful and discriminant features, instead of considering a fixed feature space. Our proposal relies on two metrics able to identify the most meaningful features. Our conjecture is that a feature could significantly change its discriminant power (hence, its role) along the taxonomy levels. Hence, we devise a greedy algorithm able to build a taxonomy by identifying the meaningful terms for each level. We perform preliminary experiments that give rise to the usefulness of the proposed approach.