Hearst Patterns Revisited: Automatic Hypernym Detection from Large Text Corpora

TL;DR

This paper compares pattern-based and distributional methods for hypernym detection, finding that pattern-based approaches outperform distributional ones on standard benchmarks due to their contextual constraints.

Contribution

The study provides a comprehensive evaluation showing the superiority of pattern-based methods over distributional approaches in hypernym detection tasks.

Findings

Pattern-based methods outperform distributional methods on benchmarks.

Contextual constraints are crucial for hypernym detection.

Simple pattern-based models are highly effective.

Abstract

Methods for unsupervised hypernym detection may broadly be categorized according to two paradigms: pattern-based and distributional methods. In this paper, we study the performance of both approaches on several hypernymy tasks and find that simple pattern-based methods consistently outperform distributional methods on common benchmark datasets. Our results show that pattern-based models provide important contextual constraints which are not yet captured in distributional methods.