Abstract
An auto-generated dataset of Curie and Néel temperatures is presented, containing 56,037 records extracted from 108,181 published scientific articles. Each record contains the extracted chemical entity and associated extracted temperature. The dataset was auto-generated by mining text from the papers using the 'chemistry-aware' natural-language-processing toolkit, ChemDataExtractor 2.2.2, in conjunction with the Snowball v2 parser, which has been adapted to extract Curie and Néel temperatures from scientific text. This dataset is the first of its kind to be generated using the Snowball v2 parser, with its evaluation yielding a precision of 72% and a recall of 61%. The public availability of this dataset will aid in the design, prediction and analysis of magnetic materials.