Using Decision Trees for Coreference Resolution

TL;DR

This paper introduces RESOLVE, a decision tree-based system for coreference resolution in business texts, demonstrating superior performance over manual rules and offering insights into useful knowledge types.

Contribution

The paper presents RESOLVE, a novel decision tree approach for coreference resolution, outperforming manual rules and enabling exploration of relevant knowledge.

Findings

Decision trees outperform manual rules in coreference resolution.

RESOLVE achieves higher evaluation metrics in two of three measures.

Framework facilitates understanding of knowledge useful for coreference tasks.

Abstract

This paper describes RESOLVE, a system that uses decision trees to learn how to classify coreferent phrases in the domain of business joint ventures. An experiment is presented in which the performance of RESOLVE is compared to the performance of a manually engineered set of rules for the same task. The results show that decision trees achieve higher performance than the rules in two of three evaluation metrics developed for the coreference task. In addition to achieving better performance than the rules, RESOLVE provides a framework that facilitates the exploration of the types of knowledge that are useful for solving the coreference problem.