Human-annotated dataset of keyword quality in open data

This dataset contains human annotations of keyword quality in open data metadata. It comprises a sample of 300 datasets from the Belgian bel-data-gov collection obtained through the European Data Portal. The dataset includes descriptive metadata, assigned keywords, human-identified concepts expressed in the metadata, and human assessments of keyword relevance and completeness. Four annotators annotated the datasets according to a structured annotation guideline. Relevance assessments examine how well assigned keywords correspond to the dataset titles and descriptions, while completeness assessments examine the extent to which the keywords cover the concepts identified in the metadata. The dataset provides a reusable benchmark for research on keyword quality assessment, metadata quality, and dataset discoverability, and can support the development and evaluation of automated methods against human judgments.