ArabiGEE: A Hierarchical Taxonomy for Arabic Grammatical Error Explanation

Researchers have introduced ArabiGEE, a hierarchical taxonomy designed to bring structured grammatical error explanations to the Arabic language, according to a paper posted on the arXiv preprint server [1]. The taxonomy is described as the first comprehensive Arabic grammatical error explanation framework grounded in explicit error types [1]. It organizes explanations across orthographic, morphological, syntactic, and lexical dimensions through a hierarchical structure, departing from approaches that treat explanation generation as free-form text [2]. The system comprises 27 error types, 140 correction types, and 324 associated explanations [2]. The authors applied ArabiGEE to manually annotate portions of existing Arabic grammatical error correction corpora [2]. They also demonstrated how structured grammatical explanations can support the automatic evaluation of large language models on Arabic grammatical error explanation tasks [2]. Large language models, which are neural networks trained on vast amounts of text for natural language processing, can generate, summarize, and analyze text but are susceptible to biases from their training data [8]. The paper was submitted to arXiv on June 9, 2026 [1]. arXiv, an open-access repository of electronic preprints founded in 1991, hosts scientific papers across fields including computer science, mathematics, and physics and surpassed two million articles by the end of 2021 [6]. The repository is not peer-reviewed; submissions are approved after moderation [6]. The authors have made their code and data publicly available [2]. The work appears on arXiv alongside community-developed tools offered through arXivLabs, a framework launched in 2020 that allows collaborators to build experimental features directly on the platform [5]. These tools include citation explorers and code-finding services, all developed under guidelines that require partners to share arXiv’s values of openness, community, and user data privacy [5].