astrina

Pulsar Data — human-verified labels for AI

Real people judge whether a review matches reality. Their weighted consensus — with every vote, tag and confidence bound — becomes training and evaluation data your models can trust.

Get an API keyTalk to us
0Confirmed
0Disputed
36Votes
1Verifiers
What every row carries/api/v1/labels
The public review: subject type and key, star rating, full text, timestamp.
The human verdict: confirmed or disputed, the trust-weighted score, the raw vote histogram (−2…+2).
Statistical confidence: Wilson interval bounds over the vote split.
Reason tags voters attached to negative votes (scam, fake, outdated, exaggerated).
Integrity flags: star-court escalation for contested splits; brigading-quarantined votes never count.
Try the APIOne endpoint, stable cursor pagination, JSON. Create a key in your dashboard and pull the stream:
curl
curl -H "X-Api-Key: YOUR_KEY" \
  "https://astrina.io/api/v1/labels?since=0&limit=50&verdict=confirmed"

Parameters: since (cursor, exclusive) · limit (1–100) · verdict = confirmed | disputed | any.

Get an API key

Honest volume, honest terms

The dataset is young and grows with the verification network — the counters above are live, never projected. Commercial licensing (volume, exclusivity, refresh cadence and price) is agreed per deal. Talk to us

What this page answers