CTFL-GEC (Thai L2 Learners GEC Corpus )

The CTFL-GEC Corpus is a manually annotated dataset created from essays written by Thai learners of English as a second language. The original writings were collected from the Center for Thai as a Foreign Language (CTFL), then segmented into sentences and corrected using a minimal-edit approach. Each sentence pair consists of the original learner-written sentence and its corrected version.The corpus contains 4,047 sentences with 8,583 annotated errors categorized into Grammar, Spelling, Punctuation, and Fluency. The annotation follows a structured format using in-text tags: error => correction:::Type. This resource is designed for grammatical error correction (GEC) tasks and supports training and evaluation of sequence-to-sequence models for Thai.