Published in the Proceedings of ArabicNLP 2023 (First Arabic Natural Language Processing Conference), co-located with EMNLP 2023, Singapore (hybrid).
The paper presents the first AI-driven Arabic crossword puzzle generator, using GPT-4, GPT-3 variants and BERT with fine-tuning, few/zero-shot prompting and quality filtering. It builds on a dataset of over 50,000 clue-answer pairs to produce clues for educational crosswords.
Authors: Kamyar Zeinalipour, Mohamed Saad, Marco Maggini, Marco Gori.
Read the paper | Code | Models and data
All the lab's language releases are listed in Language models and datasets.