ArabiGEE

ArabiGEE: Hierarchical taxonomy for Arabic grammatical error explanation with LLM evaluation support

Researchers introduce ArabiGEE, the first structured taxonomy for Arabic grammatical error explanation (GEE), organizing 27 error types, 140 correction types, and 324 explanations across orthographic, morphological, syntactic, and lexical dimensions. The taxonomy is applied to annotate existing Arabic grammatical error correction corpora and is used to benchmark LLMs on Arabic GEE tasks. The work addresses a gap in Arabic NLP tooling by moving beyond free-form explanation generation toward structured, evaluable outputs.