Abstracts and public data

Stored abstracts remain in the local literature database. The published explorer contains an availability flag, source identifiers and locally computed category labels, but no stored abstract fields or quoted abstract classification evidence. Historical public review copies also redact abstract excerpts. Original local registers and evidence remain intact; their recorded hashes refer to those local originals. publication-policy.json records hashes of transformed public copies.

Opening paper details requests the abstract directly from the reader's browser using the exact OpenAlex work ID or DOI, with Crossref DOI fallback. Requests use no credentials, no website proxy and no persistent application cache. Returned text lives only in the current page's memory and is rendered as text. Navigation cancels stale requests. If a source is unavailable or has no accessible abstract, the reader can retry or follow the publication/source links.

The publication build applies the same projection to historical CSV/JSON/HTML copies, removes abstract fields, replaces explicitly marked excerpts, and checks unmarked text against the locally stored abstract inventory. Bibliographic titles remain metadata even where an abstract repeats its title. The content audit checks structured values independently so concatenated adjacent titles are not mistaken for abstract text.

Search covers titles, authors, identifiers, venues and topics. Abstract full-text search and abstract snippets in result lists are no longer part of the public snapshot. Abstract-based category suggestions remain labelled as suggestions; their quoted text is withheld. Downloaded CSV and BibTeX and personal notes never receive fetched abstracts.