A Crash Course in Automatic Grammatical Error Correction

This repository contains materials for our tutorial on automatic grammatical error correction: R. Grundkiewicz, C. Bryant, M. Felice: A Crash Course in Automatic Grammatical Error Correction, COLING 2020.

Links and materials:

• Tutorial proposal with the recommended reading list
• Slides (Parts I-V)
• Bibliography
• List of GEC resources

Tutorial outline

Part I: Introduction

• About the tutorial
• Task definition
• Challenges

Part II: Historical and recent approaches

• Rule-based methods
• Language models
• Error-type classifiers
• Statistical machine translation
• Deep neural networks
• Shared tasks

Part III: Data and evaluation

• Data annotation
• Error corpora
• Evaluation metrics
• Human evaluation

Part IV: Neural grammatical error correction

• Neural approach to GEC
• GEC as low-resource NMT
• Data sparsity
• Correction efficacy
• Beyond the NMT framework

Part V: Recent and future work

• Findings from the BEA-2019 shared task
• Towards unsupervised GEC
• Non-English languages
• Future work

List of GEC resources

Data sets, evaluation scripts and other resources related to the field of automatic grammatical error correction.

Datasets

Publicly available error corpora for English:

• W&I+LOCNESS Corpora [paper] [download v2.1]
• FCE Data Set [paper] [download v2.1]
• JFLEG (JHU FLuency-Extended GUG) Corpus [download]
• AESW (Automated Evaluation of Scientific Writing Data Set) [download v1.2]
• NUCLE (NUS Corpus of Learner English) [paper] [download v3.3]
• Annotated Test Data from the CoNLL 2013 & 2014 Shared Task [CoNLL-2013] [CoNLL-2014]
• 8 additional annotations for the CoNLL 2014 data set [paper] [download]
• Lang-8 Learner Corpora [download v2.0] [download pre-processed]
• The WikEd Error Corpus [download v1.0]
• GMEG data sets [paper] [download]
• CWEB (Corrected Websites) data sets [paper] [download]

System outputs:

• System outputs from the CoNLL 2014 Shared Task [download]
• System outputs from the BEA 2019 Shared Task [download]

Publicly available error corpora for other languages:

• Chinese: NLPCC 2018 Task 2 GEC [paper] [website], NLPTEA 2016 CGED [paper] [website]
• Czech: AKCES-GEC [paper] [website]
• German: Falko & MERLIN Corpora [paper] [website]
• Polish: PlEWiC [website]
• Russian: RULEC-GEC dataset [paper] [website]

Metrics

• M2Scorer [software] [paper]
• ERRANT [software] [paper]
• GLEU [software] [paper #1] [paper #2]

Shared Tasks

• BEA 2019 Shared Task: Grammatical Error Correction [website] [paper]
• NLPCC 2018 Shared Task 2 - Grammatical Error Correction for Chinese [website] [paper]
• Automated Evaluation of Scientific Writing Shared Task 2016 [website]
• The Second QALB Shared Task on Automatic Text Correction for Arabic 2015 [paper]
• CoNLL-2014 Shared Task: Grammatical Error Correction [website] [paper]
• CoNLL-2013 Shared Task: Grammatical Error Correction [website] [paper]

Other materials

• http://nlpprogress.com/english/grammatical_error_correction.html
• Slides from the COLING2014 tutorial on Automatic GEC for Language Learners

(This list is incomplete, please feel free to open a pull request if you would like to add something to the list)

Reference

@inproceedings{grundkiewicz-etal-2020-crash,
  title = "A Crash Course in Automatic Grammatical Error Correction",
  author = "Grundkiewicz, Roman and Bryant, Christopher and Felice, Mariano",
  booktitle = "Proceedings of the 28th International Conference
               on Computational Linguistics: Tutorial Abstracts",
  month = dec,
  year = "2020",
  address = "Barcelona, Spain (Online)",
  publisher = "International Committee for Computational Linguistics",
  url = "https://www.aclweb.org/anthology/2020.coling-tutorials.6",
  pages = "33--38",
}