Evaluating the Performance of Chinese Multi-Label Grammatical Error Detection Using Deep Neural Networks

Tzu Mi Lin, Chao Yi Chen, Lung Hao Lee, Yuen Hsien Tseng

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In this paper, we describe the process of building a benchmark data set for Chinese multi-label grammatical error detection tasks, comparing the performance of 10 representative neural network models. Experimental results reveal that no matter which deep learning model is used, the performance is still limited which confirms the difficulty of the multi-label detection task. Our constructed datasets and evaluation results will be publicly released on the GitHub repository (https://github.com/NCUEE-NLPLab/CMLGED) to promote further research to facilitate technology-enhanced Chinese learning.

Original languageEnglish
Title of host publication30th International Conference on Computers in Education Conference, ICCE 2022 - Proceedings
EditorsSridhar Iyer, Ju-Ling Shih, Ju-Ling Shih, Weiqin Chen, Weiqin Chen, Mas Nida MD Khambari, Mouna Denden, Rwitajit Majumbar, Liliana Cuesta Medina, Shitanshu Mishra, Sahana Murthy, Patcharin Panjaburee, Daner Sun
PublisherAsia-Pacific Society for Computers in Education
Pages524-526
Number of pages3
ISBN (Electronic)9789869721493
StatePublished - 28 Nov 2022
Event30th International Conference on Computers in Education Conference, ICCE 2022 - Kuala Lumpur, Malaysia
Duration: 28 Nov 20222 Dec 2022

Publication series

Name30th International Conference on Computers in Education Conference, ICCE 2022 - Proceedings
Volume1

Conference

Conference30th International Conference on Computers in Education Conference, ICCE 2022
Country/TerritoryMalaysia
CityKuala Lumpur
Period28/11/222/12/22

Keywords

  • deep learning
  • Grammatical error detection
  • multi-label classification

Fingerprint

Dive into the research topics of 'Evaluating the Performance of Chinese Multi-Label Grammatical Error Detection Using Deep Neural Networks'. Together they form a unique fingerprint.

Cite this