Creating term associations using a hierarchical ART architecture

Abstract

In this work we address the problem of creating semantic term associations (key words) from a text database. The proposed method uses a hierarchical neural architecture based on the Fuzzy Adaptive Resonance Theory (ART) model. It exploits the specific statistical structure of index terms to extract semantically meaningful term associations; these are asymmetric and one-to-many due to the polysemy phenomenon. The underlying algorithm is computationally appropriate for deployment on large databases. The operation of the system is illustrated with a real database.