Published in the Proceedings of the Ninth Italian Conference on Computational Linguistics (CLiC-it 2023), Venice, Italy, CEUR Workshop Proceedings Vol-3596.
The paper builds a system that generates and checks Italian crossword clues, fine-tuning GPT-3 models and BERT on a large set of clue-answer pairs and using zero/few-shot prompting to draw clues from texts. Classifiers trained on human-labeled data filter out weak clues, and the evaluation showed the approach produces good educational crosswords.
Authors: Kamyar Zeinalipour, Tommaso Iaquinta, Asya Zanollo, Giovanni Angelini, Leonardo Rigutini, Marco Maggini, Marco Gori.
Read the paper | Code | Models and data
All the lab's language releases are listed in Language models and datasets.