CxGGEC: Construction-Guided Grammatical Error Correction

ACL ARR 2025 February Submission6640 Authors

Abstract: The grammatical error correction (GEC) task aims to detect and correct grammatical errors in text to enhance its accuracy and readability. Current GEC methods primarily rely on grammatical labels for syntactic information, often overlooking the inherent usage patterns of language. In this work, we explore the potential of construction grammar (CxG) to improve GEC by leveraging constructions to capture underlying language patterns and guide corrections. We first establish a comprehensive construction inventory from corpora. Next, we introduce a construction prediction model that identifies potential constructions in ungrammatical sentences using a noise-tolerant language model. Finally, we train a CxGGEC model on construction-masked parallel data, which performs GEC by decoding construction tokens into their original forms and correcting erroneous tokens. Extensive experiments on English and Chinese GEC benchmarks demonstrate the effectiveness of our approach.

Paper Type: Long

Research Area: Machine Learning for NLP

Research Area Keywords: grammar error correction; construction grammar; language model

Contribution Types: Model analysis & interpretability, NLP engineering experiment, Theory

Languages Studied: English

Submission Number: 6640