Data
Everything the viewers use is published here. If you want to check a number, reproduce a chart, or build something else on top of this, start with these files.
What is published#
Derived facts only. Each record says which work, which passage, which journal, which year, and which article DOI. There is no article text of any kind. The verbatim citation context extracted during processing is confidential under the JSTOR Text Analysis Support terms and is dropped before any shareable file is written.
| File | Contents |
|---|---|
[meta.json](/data/viewer/meta.json) | Journals, year range and histogram, corpus totals, thresholds used. |
[view_a.json](/data/viewer/view_a.json) | Per work: placed count, uncertainty band, unplaceable mass, placement rate, tier, distinct articles, collision partners. |
[view_a_filter.json](/data/viewer/view_a_filter.json) | Per work: sparse journal × year matrix of placed citations. The time chart's source. |
[works_index.json](/data/viewer/works_index.json) | Work list with tiers and totals. |
view_b/<work>.json | Per work: section-level cells and per-citation records with DOI links. |
Licence#
The derived data is CC BY 4.0 — use it for anything, including commercially, provided you attribute the source. See licensing for the reasoning and for how the other components are licensed.
The underlying JSTOR full text is not redistributable and is not available here. It never leaves the machine that processes it.
Reproducing the build#
The aggregator that produces these files runs locally against the JSTOR delivery. Given the delivery, build_viewer_data.py regenerates every file above, and prints invariant self-checks so a bad run announces itself.
Without the delivery you cannot regenerate the data from scratch — but you can verify every published number against these files, and the methods record documents each transformation between the two.
Caveat before reuse#
Placement rates vary enormously across works, and a raw count from view_a.json means something different for Republic (95% placed) than for Crito (20% placed). If you reuse these numbers, carry the tier and placement rate with them. See the two tiers.