Loading...
Loading...
Contributors:
Dataset URI:
10.5281/zenodo.23382381This dataset preserves Version 1 of the research infrastructure developed within the Do Arquivo ao Algoritmo / From the Archive to the Algorithm project for the processing, evaluation, preservation, and curation of historical OCR derived from nineteenth-century Brazilian provincial reports. Version 1 constitutes a stabilized baseline rather than a completed or definitive representation of the corpus. It preserves raw Tesseract OCR outputs, corpus inventories, computational provenance, processing reports, reconstruction evidence, software, environment documentation, and two human-curated pilot cases. The preserved corpus comprises 2,500 reports, representing 195,909 expected digitized pages, of which 195,701 produced successful OCR outputs and 208 were recorded as processing errors. The physical raw-OCR preservation layer also retains 45 known excess TXT and JSON files associated with documented source duplication, rather than silently removing them. The repository is designed to make distinctions among source documents, machine-generated OCR, human-curated textual representations, validation states, and later analytical enrichment explicit and auditable. The human-curated layer included in Version 1 represents a first preserved human-review state and should not be interpreted as definitive transcription or ground truth. The deposit also documents the computational and curatorial architecture through which the data were produced, including scripts, execution reports, inventories, checksums, environment information, provenance reconstruction, and a collaborative diplomatic-transcription protocol. This architecture supports reproducibility, representation management, and future collaborative curation. A central methodological concern of the project is the concept of historiographical recoverability: the extent to which historically relevant information remains computationally recoverable across different textual representations, rather than treating OCR quality solely as a measure of global textual accuracy. The digitized source images themselves are not redistributed in Version 1 because redistribution rights for the complete source-image corpus have not been established. Provenance relationships to those sources are nevertheless preserved in the deposited metadata. Version 1 stabilizes the preservation and curation architecture; it does not close the research corpus. Future versions may incorporate additional validated textual states, structural classifications, report and section typologies, thematic annotations, controlled vocabularies, and other research-derived layers while preserving explicit version and derivation relationships.
Keyword(s):
Brazil; collaborative curation; computational provenance; digital curation; digital history.
MeSH Term(s):
Grants and Funding:
Associated Publication(s):
No Associated Publication Data Available
Supplementary Data:
No Supplementary Data Available
Related Dataset(s):
No Related Dataset Data Available
The National Library of Medicine (NLM) provides information about biomedical datasets. Inclusion on an NLM Web page does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: NLM Web Policies (nih.gov)