StatSheets: dataset and reproducibility kit for spreadsheet table understanding

Le résumé fourni par la source

Code and dataset accompanying the CIKM 2026 paper "Structured Prediction for Scalable Spreadsheet Table Understanding: From Cell Types to Table Ranges". Includes the StatSheets benchmark – 737 manually annotated sheets from 14 public statistical organizations, in six languages and five file formats, with complete cell-level cell-type annotations and 818 annotated table ranges – together with the cell-type classification and table detection pipelines and the scripts needed to reproduce all experiments. This archive bundles three differently licensed components. The source code (cell-type-classification/, table-detection/) is released under the MIT License. The StatSheets annotations (dataset/annotations/) are released under CC-BY 4.0. The source spreadsheets (dataset/spreadsheets/) remain the property of their original publishers and are subject to their respective terms of use; dataset/manifest.csv records the origin URL of each file. See DATA-LICENSE.md for details.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.