Peeking inside the Black-Box: Reinforcement Learning for Explainable and Accurate Relation Extraction

Abstract

We introduce CogRE, a novel framework for relation extraction (RE), enhancing RE from both accuracy and explainability. The framework has two key components: (i) a reasoning mechanism that formulates relation extraction as a series of text-processing steps inspired by cognitive science, and (ii) an optimization process driven by a novel reinforcement learning (RL) reward function. Our framework introduces relation keywords and rewards generating such keywords using an automatically constructed keywords dictionary. This design addresses the lack of language-based explanations in traditional RE and provides supervision for explanation during RL training. Our experiments show that CogRE improves explanation quality by addressing two common failure patterns in one-shot RE: poor attention focus and limited one-shot learning capability. For example, our cognitive-structured reasoning with Qwen2.5-15B-Instruct on One-shot NYT29 achieves 24.65% F1, surpassing prior reasoning-based designs. Optimizing this approach with RL using our reward further improves performance by +23.46% (absolute). Further, models trained on NYT29 with our reward achieve a +16.9% F1 gain on out-of-distribution WIKIDATA. Finally, human evaluation shows that our best model generates relational keywords closely aligned with gold labels, increasing human explanation quality ratings by 54% (relative).