Data

The full Uplink archive — 1,838 articles from 145 issues, October 1990 through March 2007 — is available as JSON. It documents how computer-assisted reporting was actually practiced as it evolved: the tools, the data sources, and the stories, issue by issue for sixteen years.

Downloads

data/uplink-articles.json — the complete corpus in one file, including full text (~8 MB). Wrapped as {name, description, publisher, source, generated_at, count, articles: [...]}.
article/<slug>.json — one JSON file per article, alongside its HTML page, with related-article links added (example). Every article page links to its JSON version.

Field dictionary

Field	Type	Description
`id`	`string`	Permanent identifier, minted once into the corpus. Never changes, even if the headline is later corrected.
`slug`	`string`	URL slug derived from year, month, and headline. Currently identical to id, but regenerated at build time — use id for longitudinal work.
`url`	`string`	Canonical URL of the article page on this site.
`year / month / monthNum`	`number / string / number`	Issue date. monthNum is 1–12 for sorting; some issues cover two months (e.g. "July/August"), in which case monthNum reflects the first.
`headline`	`string`	Article title as printed.
`kicker`	`string`	Section label or eyebrow above the headline (e.g. "ENERGY"). Often empty.
`author_name`	`string`	Byline name. Empty for unsigned items (about a third of the corpus — many short listings ran without bylines).
`author_title`	`string`	The author’s job title only.
`affiliation`	`string`	The author’s publication or organization.
`summary`	`string`	One-paragraph abstract, generated during archival processing.
`keywords`	`string[]`	Free-text tags assigned during archival processing.
`topics`	`string[]`	One to three entries from a controlled 12-topic taxonomy (see data/taxonomy.json in the repository).
`technologies`	`string[]`	Software, hardware, formats, and services mentioned in the article, normalized against a controlled vocabulary (data/technologies.json) and alias map (data/technology_aliases.json). Useful for tracing tool adoption over time.
`full_text`	`string`	Complete article text, OCR-extracted from the original scanned PDFs. Occasional OCR artifacts remain.
`provenance`	`object`	How the record was produced: extracted_by (extraction model or "pre-v2" for legacy records), extracted_at, and verified (whether the extraction passed anchor checks against the source text).
`similar`	`{slug, score}[]`	Per-article JSON only: top five related articles by TF-IDF cosine similarity, computed at build time.

Example

Count technology mentions by year with curl and jq:

curl -s https://dwillis.github.io/uplink/data/uplink-articles.json \
  | jq -r '.articles[] | .year as $y | .technologies[] | "\($y) \(.)"' \
  | sort | uniq -c | sort -rn | head

Notes for researchers

Use id as the stable key when tracking articles across corpus revisions.
topics and technologies are controlled vocabularies; keywords are free text.
The underlying per-issue source files, extraction pipeline, and data-quality report live in the GitHub repository. Original publication: NICAR (National Institute for Computer-Assisted Reporting).