Towards the Development of Balanced Synthetic Data for Correcting Grammatical Errors in Arabic: An Approach Based on Error Tagging Model and Synthetic Data Generating Model

Ahlam Alrehili, Areej Alhothali · 2025

Read Towards the Development of Balanced Synthetic Data for Correcting Grammatical Errors in Arabic: An Approach Based on Error Tagging Model and Synthetic Data Generating Model by Ahlam Alrehili, Areej Alhothali (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-05312.txt.

Description

Ahlam Alrehili and Areej Alhothali build artificial training data for Arabic grammar correction in a paper from 2025. Systems that fix grammar need many example errors, but Arabic has few labeled resources. Their approach has two models. The first tags a correct sentence with the error types it could plausibly carry, sorting it into twenty-six classes. The second model then rewrites the sentence with those errors planted in.

The tagging model reached top scores on two standard Arabic test sets. Training a corrector on the artificial data set a new best score on one set. The pair generated more than thirty million sentence pairs in total. They offer the method as a template for other low-resource languages.

Catalog record

CREATORAhlam Alrehili, Areej Alhothali
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE80 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-05312

Browse the full library · About the Tower of Babel