HTRC provides access to data from the HathiTrust corpus in several forms across its suite of tools and services for computational text analysis. Data is periodically synced from HathiTrust, but not all HTRC tools and services are updated on the same schedule. Additionally, copyright, user agreements, and security concerns impact data availability and format.
HTRC algorithms and HTRC Data Capsules are capable of analyzing the entire HathiTrust corpus, and additionally make use of each volume’s MARC bibliographic and METS metadata. Both the HTRC algorithms and Capsule-environments draw from the HTRC Data API described below.
The HTRC makes available also two datasets, the HTRC Extracted Features Dataset and a dataset of Word Frequencies in English Language Literature, 1700-1922. HTRC Extracted Features includes metadata and extracted page-level data (words and word counts) for 15.7 million volumes.
HathiTrust+Bookworm visualizes data for 15.7 million volumes.
The following table disambiguates textual data access, including availability and format, within the HTRC ecosystem.
|Tool or service|
# volumes available (as of 8/19)
|rights status||data access mechanism||file format||data format||permissions required|
|HTRC Analytics algorithms||17 million||All||Via HTRC Workset; researcher runs tool without accessing underlying data||(no file access)|
Uncorrected OCR text data; only results are exposed to researcher (not the underlying data)
|HTRC Analytics account|
|HT+Bookworm tool||15.7 million||All||Via web-interface; researcher visualizes data without accessing underlying data||(no file access)||Unigrams (single words), based on HTRC Extracted Features dataset; underlying data not exposed to researcher||(none)|
|HTRC Data Capsule||6.5 million for everyone; 17 million for affiliates of HT member institutions upon approval or request||Public domain for everyone; In-copyright limited to affiliates of HT member institutions upon approval of request||HTRC Data API||Zipped text files in PairTree directory structure||Uncorrected OCR text data||HTRC Analytics account|
|HTRC Extracted Features dataset||15.7 million||All||rsync||JSON files in PairTree directory structure||Volume- and page-level metadata and part-of-speech tagged page-level "bags of words"||(none)|
|HathiTrust dataset request||6.5 million, dependent on institutional agreements||Public domain; accessibility of Google-digitized volumes dependent on whether researcher's home institution has signed agreement||rsync|
Zipped text files in PairTree directory structure
|Uncorrected OCR text data||Custom dataset request application|
|HathiTrust Data API||~800 thousand available; practical limit for retrieval is 10 thousand volumes||Public domain volumes not digitized by Google||HathiTrust Data API||Zipped text files in PairTree directory structure||Uncorrected OCR text data and page images||Key required to use the API outside of the Web client: read the documentation|
This table outlines the differences between the HTRC Data API and HathiTrust Data API. As a researcher-accessible service, HTRC Data API functions within the HTRC Data Capsules environment.
HTRC Data API
HathiTrust Data API
|purpose||to provide access to HathiTrust text data within an HTRC Data Capsule AND to serve high-performance large-scale algorithms and programs (not publicly-accessible)||to provide public users some volume retrieval capabilities|
|data available||entire HathiTrust corpus||public domain volumes not digitized by Google|
|use||In-capsule via the HTRC Workset Toolkit||Get more information from HathiTrust|
|bulk retrieval of volumes||yes||up to 10,000 volumes|
|metadata available||METS||METS, MARC|