CLEME2.0: Towards Interpretable Evaluation by Disentangling Edits for Grammatical Error Correction

TL;DR

CLEME2.0 introduces an interpretable, aspect-based evaluation metric for grammatical error correction that improves human consistency and outperforms existing metrics.

Contribution

The paper presents CLEME2.0, a novel reference-based GEC evaluation metric that disentangles correction aspects for better interpretability and robustness.

Findings

Achieves state-of-the-art results on human judgment datasets.

Improves human consistency over existing metrics.

Effectively exposes GEC system qualities and drawbacks.

Abstract

The paper focuses on the interpretability of Grammatical Error Correction (GEC) evaluation metrics, which received little attention in previous studies. To bridge the gap, we introduce **CLEME2.0**, a reference-based metric describing four fundamental aspects of GEC systems: hit-correction, wrong-correction, under-correction, and over-correction. They collectively contribute to exposing critical qualities and locating drawbacks of GEC systems. Evaluating systems by combining these aspects also leads to superior human consistency over other reference-based and reference-less metrics. Extensive experiments on two human judgment datasets and six reference datasets demonstrate the effectiveness and robustness of our method, achieving a new state-of-the-art result. Our codes are released at https://github.com/THUKElab/CLEME.