English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Data management has traditionally relied on synthetic data generators togenerate structured benchmarks, like the TPC suite, where we can controlimportant parameters like data size and its distribution precisely. Thesebenchmarks were central to the success and adoption of database managementsystems. But more and more, data management problems are of a semantic nature.An important example is finding tables that can be unioned. While any twotables with the same cardinality can be unioned, table union search is theproblem of finding tables whose union is semantically coherent. Semanticproblems cannot be benchmarked using synthetic data. Our current methods forcreating benchmarks involve the manual curation and labeling of real data.These methods are not robust or scalable and perhaps more importantly, it isnot clear how robust the created benchmarks are. We propose to use generativeAI models to create structured data benchmarks for table union search. Wepresent a novel method for using generative models to create tables withspecified properties. Using this method, we create a new benchmark containingpairs of tables that are both unionable and non-unionable but related. Wethoroughly evaluate recent existing table union search methods over existingbenchmarks and our new benchmark. We also present and evaluate a new tablesearch methods based on recent large language models over all benchmarks. Weshow that the new benchmark is more challenging for all methods thanhand-curated benchmarks, specifically, the top-performing method achieves aMean Average Precision of around 60%, over 30% less than its performance onexisting manually created benchmarks. We examine why this is the case and showthat the new benchmark permits more detailed analysis of methods, including astudy of both false positives and false negatives that were not possible withexisting benchmarks.

Generative Benchmark Creation for Table Union Search