Multi-label large margin hierarchical perceptron

Abstract

This paper looks into classification of documents that have hierarchical labels and are not restricted to a single label. Previous work in hierarchical classification focuses on the hierarchical perceptron (Hieron) algorithm. Hieron only supports single label learning. We investigate applying several standard multi-label learning techniques to Hieron. We then propose an extension of the algorithm (MultiHieron) that significantly outperforms all previously mentioned techniques. MultiHieron has a new aggregate loss function for multiple labels. Improvement is shown on the Aviation Safety Reporting System (ASRS) flight anomaly database and OntoNews corpus using both at and hierarchical categorisation metrics.