Title: Aladeeb: AI-Powered Arabic Poetry and Fantasy Story Generator Using Transformer Models (BERT/GPT) Author: Oday Yaser Alnaddaf Date: April 2026 University: Arab International University, Faculty of Informatics and Communication Engineering Abstract:This project presents Aladeeb, an AI system that generates metered classical Arabic poetry and fantasy stories from user-provided keywords or short prompts. The system fine-tunes AraBERT for poetry understanding and GPT-2-Arabic for story generation, combined with a rule-based post-processor to enforce Khalilian prosodic meters (achieving 86% meter accuracy). A custom dataset of 50,000 annotated Arabic poems and 20,000 fantasy short stories was collected. Experimental evaluation shows a human coherence score of 4.2/5 for poetry and 4.0/5 for stories, with perplexity dropping from 78.4 to 42.1 after fine-tuning. The system is deployed as a lightweight Flask web application that runs on CPU (poetry in ~4.2s, story in ~8.5s). This work demonstrates that hybrid neural-symbolic approaches and domain-specific fine-tuning enable accessible creative text generation for Arabic literature. Keywords: Arabic poetry generation, fantasy story generation, Transformer, BERT, GPT, Arabic prosody, natural language processing This work was conducted at Arab International University (AIU), Syria. The official website of the university is: https://www.aiu.edu.sy
Oday Naddaff (Sat,) studied this question.