Stemmata
Know what you may mine.
Text-and-data-mining permission is not a footnote. Across 350 live records measured from OpenAlex, only a fifth could be processed in full. Most tools hide this. Stemmata puts it on the card, with the reason written out: which licence, which publisher, which clause.
What it does
- Permission on every work. Full text, metadata only, or closed — each with the reason that produced it. When the reservation set behind a decision is older than 180 days, the card says so instead of pretending the answer is fresh.
- Search your saved library offline, without a model. A lexical index and a character-trigram vector work together; on inflected queries the vector leg raised recall from 7.4% to 44.1%.
- Notes stored on this device. A conflict-free merge engine underlies the notes, but this release does not provide synchronization between devices.
- Export that respects the answer. BibTeX, RIS and Markdown. A work marked closed does not leave in an export — eleven tests hold that line.
- Lineage. A stemma drawn from what a work cites — filled where you hold the full text, open where the publisher closed access. Works cited but not in your library are counted, not guessed at. Citations to a work are not drawn: the app only knows what your own library contains.
Try it before you build anything
Stemmata is empty until you open a library, because it ships with no data pretending to be yours. To see it working, download a real one:
ornek-kutuphane.jsonl — 140 works from OpenAlex: 17 full text, 64 metadata only, 59 access closed. Those proportions are not staged; they are what the publishers' licences actually say for that query.
Open it in the app with Open library, or tap Open a sample library on the first screen — the same file ships inside the app.
Where your data is
On your device. There is no Stemmata server, no account and no telemetry. Searching your saved library works offline. External catalogue lookups and DOI links, where available in your installed version, contact external services only when you choose those actions. The privacy policy explains what is sent.
Stemmata reports the permission a publisher has declared. It is a reading of published licence and opt-out data, not legal advice.