Unsupervised Explanation Generation via Correct Instantiations

While large pre-trained language models (PLM) have shown their great skills at solving discriminative tasks, a significant gap remains when compared with humans for explanation-related tasks. Among them, explaining the reason why statement is wrong (e.g., against commonsense) incredibly challenging. The major difficulty finding conflict point, where contradicts our real world. This paper proposes Neon, two-phrase, unsupervised explanation generation framework. Neon first generates corrected instantiations of (phase I), then uses them to prompt PLMs find point and complete II). We conduct extensive experiments on two standard benchmarks, i.e., ComVE e-SNLI. According both automatic human evaluations, outperforms baselines, even those human-annotated instantiations. In addition negative prediction, we further demonstrate that effective generalizing different scenarios. resources are available at: https://github.com/Shark-NLP/Neon.