The dataset is provided in a csv file format, containing tabular data with 1831 records extracted from approximately 50 articles. Each record consists of a distinct combination of descriptors (composition, processing route), targets (properties), and reference information.