The corpus consists of ten tabular datasets from surveys conducted by official organizations such as CIS, CEA, CRS, and 40dB. It includes a total of 200 question–answer pairs, organized as tuples (dataset, question, answer), which facilitates its expansion and reuse.
Language(s)
Spanish
Dataset description link
Year
2025
Domain
Diverse
Annotations
Each instance in Spa-DataBench is organized as a tuple (dataset, question, answer), where the answer corresponds to a gold standard label. The labeling information includes the type of expected answer, classified as boolean, categorical, numerical, or a list (of numbers or categories).
Data access
Public
Publication
Osés Grijalba, J. et al. 2024. Towards Quality Benchmarking in Question Answering over Tabular Data in Spanish. Procesamiento del Lenguaje Natural, 73, pp. 283-296.
NLP Topic
Number of units
300
Documents
300
Size
300.00MB
Training set size
120
Test set size
100
Development set size
80
