HathiTrust today announces the release of a significantly expanded open dataset, the HathiTrust Research Center (HTRC) Extracted Features (EF) Dataset, Version 1.0. This dataset provides researchers with open access to data extracted from the full text of the HathiTrust Digital Library (HTDL) at an unprecedented scale.
The Extracted Features Dataset opens the complete HathiTrust collection for investigations into historical and cultural trends, the rise and fall of topics within the corpus, and the evolution of words and writing structures in publications dating from the 16th to the late 20th century. The EF Dataset provides quantitative information about word and line counts, parts of speech, and other details within each page of every volume in the HTDL. In addition to these larger-scale investigations, the EF Dataset also allows researchers to closely analyze the contents of a given volume or subset of volumes.
The data is extracted from 13.7 million volumes found in the HTDL, representing over 5 billion pages consisting of over 2 trillion tokens (words). A preliminary release of the EF Dataset, drawn from a much smaller subset comprising only HathiTrust’s public domain collection, has already enabled novel research from scholars in economics, history, linguistics, literary studies and sociology, among other fields.
“The Extracted Features Dataset creates opportunities for scholarship and teaching that were previously impossible,” said J. Stephen Downie, co-director of HathiTrust Research Center and Associate Dean for Research and Professor at the School of Information Sciences, University of Illinois at Urbana-Champaign. “We look forward to seeing how the scholarly community takes advantage of the EF dataset in their research, labs, and classrooms.”
“We launched the HathiTrust Research Center to help researchers to fully mine the entire collection of texts found in HathiTrust,” said Michael Furlough, HathiTrust’s executive director. “This release provides a novel and effective way to do so by generating relevant data from the entire corpus.”