Hierarchical Entity Typing via Multi-level Learning to Rank

TL;DR

This paper introduces a hierarchical entity classification method that leverages ontological structures during training and prediction, achieving state-of-the-art accuracy by using a multi-level learning-to-rank loss and a coarse-to-fine decoder.

Contribution

It presents a novel multi-level learning-to-rank loss and a coarse-to-fine decoding approach for hierarchical entity typing, improving accuracy over previous methods.

Findings

Achieved state-of-the-art results on multiple datasets.

Significantly improved strict accuracy in hierarchical classification.

Demonstrated effectiveness of ontological structure integration.

Abstract

We propose a novel method for hierarchical entity classification that embraces ontological structure at both training and during prediction. At training, our novel multi-level learning-to-rank loss compares positive types against negative siblings according to the type tree. During prediction, we define a coarse-to-fine decoder that restricts viable candidates at each level of the ontology based on already predicted parent type(s). We achieve state-of-the-art across multiple datasets, particularly with respect to strict accuracy.