Grammatical Error Correction: A Survey of the State of the Art

Abstract

Grammatical Error Correction (GEC) is the task of automatically detecting and correcting errors in text. The task not only includes the correction of grammatical errors, such as missing prepositions and mismatched subject-verb agreement, but also orthographic and semantic errors, such as misspellings and word choice errors respectively. The field has seen significant progress in the last decade, motivated in part by a series of five shared tasks, which drove the development of rule-based methods, statistical classifiers, statistical machine translation, and finally neural machine translation systems which represent the current dominant state of the art. In this survey paper, we condense the field into a single article and first outline some of the linguistic challenges of the task, introduce the most popular datasets that are available to researchers (for both English and other languages), and summarise the various methods and techniques that have been developed with a particular focus on artificial error generation. We next describe the many different approaches to evaluation as well as concerns surrounding metric reliability, especially in relation to subjective human judgements, before concluding with an overview of recent progress and suggestions for future work and remaining challenges. We hope that this survey will serve as comprehensive resource for researchers who are new to the field or who want to be kept apprised of recent developments.

Exploring Sentence-level Revision Capabilities of LLMs in English for Academic Purposes Writing Assistance

“…GEC involves two main processes: identifying the errors and correcting them. These errors can be lexical, such as spelling mistakes; grammatical, including issues with tense, voice, or subject-verb agreement; or related to usage, such as non-idiomatic expressions [9].…”

Sequence Tagging Approach in Grammar Error Detection: Identifying Areas of Improvement for the State-of-the-Art

“…Another popular ML-driven approach is sequence modeling, which predicts error corrections through processing sequential data in a text (Bryant et al, 2023). In such an approach, errors are considered to be consist of a sub-sequence of tokens in a longer token sequence, which can be identified by a combination of internal and contextual features (Gamon, 2011).…”

“…The validation of GEC models commonly include comparisons with benchmark models and human evaluation in datasets (Bryant et al, 2023;Davis et al, 2024;Sottana et al, 2023). For automated validation, tools such as ERRor ANnotation Toolkit (ERRANT, Bryant, Felice, & Briscoe, 2017;Zhou et al, 2023) are employed for the automatic alignment of original and corrected sentence pairs and subsequent detection of corrections made.…”

“…Buttressed by technological progress, these GED and GEC systems have evolved from basic rule-based to advanced machine learning-based models (Bryant et al, 2023;Yuan, 2017). Yet, the validation of these models often remains predominantly quantitative, primarily focusing on metrics such as precision, recall, and F-scores, without delving deeper into the underlying reasons for success and failure in detection or correction.…”

“…2 Literature Review GEC has evolved substantially over the past decades, driven by the complexities and the diverse needs in language learning applications. Starting from rule-based systems, which were limited by their inflexibility and inability to handle noisy data, the field has transitioned towards more robust machine learning (ML) and hybrid models (Bryant et al, 2023;Yuan, 2017). These newer models bring a better capacity for handling complex grammatical structures and diverse error types prevalent in learner English, as reported in various shared tasks (Felice, Yuan, Andersen, Yannakoudakis, & Kochmar, 2014;Yuan, Stahlberg, Rei, Byrne, & Yannakoudakis, 2019).…”