Attribution Matters: Recovering Hidden Contributions through CARE-Aligned Annotation

Biodiversity knowledge in unstructured text is fragmented, siloed, and unevenly attributed. This poster presents a CARE-aligned human-in-the-loop annotation approach to recovering hidden contributors in transcriptions from Livingstone Online. 

Taxonomic records routinely credit only publication authors, omitting those who located, collected, prepared, depicted, and preserved specimens. Initiatives such as Bionomia, One More Voice, and Wikidata improve attribution through persistent identifiers but depend on explicit naming.

Structural omissions such as passive voice (“was collected”), collective descriptors (“the crew”), and possessive proxies (“my guide”) obscure contributions of women in science, Indigenous experts, local guides, and other knowledge stewards. As digital collections increasingly train AI systems, these patterns of erasure risk being reproduced and amplified at scale.

This pilot pipeline combines AI-based entity detection and community validation to produce W3C Web Annotation Data Model (WADM) outputs published in an InvenioRDM repository with persistent identifiers, supported by annotation environments such as annostor.

New methods for encoding hidden contributors as provenance-linked data:

• improve understanding of nomenclatural and taxonomic processes, 

• contribute to the history of biodiversity science, and 

• connect otherwise disconnected scientific events through a graph of attribution.

The pipeline advances FAIR and CARE principles and produces data for reuse through PID “roundtripping” and related infrastructure processes. Foregrounding attribution as a key pillar of taxonomic description further supports the Digital Extended Specimen vision and enables community-driven annotation to transform historical records into interoperable knowledge that honors all contributors.