Reading Between the Clues: A Fine-Grained Analysis of Italian Crossword Solving Systems

Download

• Paper

Abstract

Existing evaluations of crossword solving systems predominantly rely on aggregate metrics, offering limited insight into how different systems behave with respect to the linguistic properties of clues and answers. In this work, we propose a fine-grained evaluation on the Italian language along two complementary annotation axes: a syntactic axis, characterizing each clue according to an improved categorization of the cruciverbese register, and a lexical-semantic axis, capturing the clue-answer relation through cosine distance, named entities, and degree of polysemy. We release an enriched version of an existing Italian crossword dataset annotated along both axes, introduce two novel architectures for crossword clue answering, and conduct an extensive comparative analysis of a broad set of existing systems together with our newly proposed models, surfacing systematic differences in how architectural choices interact with the linguistic properties of clue-answer pairs.

Citation

@article{ciaccio2026clues,
  title={Reading Between the Clues: A Fine-Grained Analysis of Italian Crossword Solving Systems},
  author={Ciaccio, Cristiano and Zanollo, Asya and Miaschi, Alessio and Sarti, Gabriele and Dell’Orletta, Felice and Nissim, Malvina},
  booktitle={Proceedings of the Twelfth Italian Conference on Computational Linguistics (CLiC-it 2026)},
  year={2026}
}