Cross-Linguistic Error Analysis in ESL Writing using NLP

Abstract

This study examines cross-linguistic error analysis in ESL writing using natural language processing. It focuses on the role of the writer’s first language (L1) in shaping error patterns and influencing grammatical error correction performance. Based on the FCE corpus, the research compares three learner groups: native speakers of Russian, Spanish, and Chinese. Error profiles were constructed for each group, and several models, including BERT, RoBERTa, T5, and Qwen, were evaluated. The results show that incorporating L1-specific information can improve correction quality and reduce disparities across learner groups. The practical value of the study lies in the possibility of creating adaptive learning systems that would allow teachers to analyze each learner’s performance and adjust teaching methods accordingly.