Something's Fishy in the Data Lake: A Critical Re-evaluation of Table Union Search Benchmarks

Allaa Boutaleb, Bernd Amann, Hubert Naacke, Rafael Angarita

Keywords: table union search, data discovery, related table search, table representation learning, table similarity, information retrieval

TL;DR: We identify structural limitations in Table Union Search benchmarks that can both inflate and hinder method performance, and propose criteria for future benchmarks.

Abstract: Recent table representation learning and data discovery methods tackle table union search (TUS) within data lakes, which involves identifying tables that can be unioned with a given query table to enrich its content. These methods are commonly evaluated using benchmarks that aim to assess semantic understanding in real-world TUS tasks. However, our analysis of prominent TUS benchmarks reveals several limitations that allow simple baselines to perform surprisingly well, often outperforming more sophisticated approaches. This suggests that current benchmark scores are heavily influenced by dataset-specific characteristics and fail to effectively isolate the gains from semantic understanding. To address this, we propose essential criteria for future benchmarks to enable a more realistic and reliable evaluation of progress in semantic table union search.

Include In Proceedings: Yes

Submission Number: 7