Manually Annotating Gender Biased Language in University of Edinburgh Heritage Collections Archival Metadata Descriptions

These datasets contains metadata descriptions extracted from the University of Edinburgh's Heritage Collections (HC) Archives' catalogue in to create an annotated dataset for training text classification models to detect gender biased language.Four descriptive metadata fields were extracted for all collections, subcollections, and items in the HC Archives' online catalog. The ``Title'' field is the name of the archival record, which either documents a single or group of archival material. The ``Biographical / Historical'' field contains information about the people, time period, and places associated with the collection, subcollection, or item being described. The ``Scope and Contents'' field summarizes the contents of the collection, subcollection, or item to which the field belongs. Though not all records include the ``Processing Information'' field, those that do typically record the person who wrote the description for the collection, subcollection, or item's descriptive metadata fields, and the date the description was written.The datasets were manually annotated by five annotators according to the Taxonomy of Gendered and Gender Biased Language. The annotated datasets include the annotated text span, the description in which that text span appears, the label with which the text span was annotated, and a note explaining an annotator's rationale for applying the label to the text span.Please refer to the datasets' associated GitHub repositories and papers for further details on their creation and contents.