A study published in *Scientific Reports* compared domain-specific and general-purpose transformer models for classifying Arabic poets. Arabic poetry has long been central to Arab cultural and intellectual life, serving as artistic expression and a repository of collective memory. The task of poet classification is a form of authorship attribution, aiming to determine which poet from a predefined set wrote a given verse based on stylistic evidence. Arabic poetry remains challenging for Natural Language Processing due to its morphological richness, structural complexity, and poetic language features like figurative expression and prosodic structure. Earlier studies relied on traditional machine learning algorithms such as Support Vector Machines and Naïve Bayes with manually engineered lexical features, which had limited ability to generalize across heterogeneous poetic styles and historical periods. Deep neural architectures like Long Short-Term Memory and Bidirectional LSTM networks later improved sequential language modelling by learning hierarchical representations directly from data.