The goal of Cruciverb-IT is twofold: on the one hand, it provides an interactive environment where anyone can experiment with AI-assisted crossword solving; on the other, it serves as a demonstration platform to explore how modern language models handle lexical knowledge, clue interpretation, and word retrieval in Italian.
The application generates crossword grids automatically and lets players choose different grid sizes (from 5×5 to 13×13). During the game, users can interact with the system through three main features:
- Ask the AI to obtain candidate answers for a clue generated by the model;
- Verify your solution to check which letters are correct or incorrect;
- Compare with the AI solution, evaluating how the system performs against the human player.
The model powering the system is publicly available on Hugging Face under the name cruciverb-it/crossword-space-mpnet-base-ade, making the project fully reproducible and accessible for further research.
Click here to try the demo.
References
Ciaccio C., Sarti G., Miaschi A., Dell’Orletta F., Nissim M. (2026) “Cruciverb-IT at EVALITA 2026: Overview of the Crossword Solving in Italian Task”. In Proceedings of the Ninth Evaluation Campaign of Natural Language Processing and Speech Tools for Italian. Final Workshop (EVALITA 2026), 26-27 February, 2026, Bari.
Ciaccio C., Sarti G., Miaschi A., Dell’Orletta F. (2025) “Crossword Space: Latent Manifold Learning for Italian Crosswords and Beyond”. In Proceedings of the Eleventh Italian Conference on Computational Linguistics (CLiC-it 2025), 24-26 September, 2025, Cagliari.
(Please cite the papers above if you use Cruciverb-IT in your research)