Reading tips
You should not need a special dictionary for this site. If a page still feels dense, these short reminders may help.
“Finding” or “lead”
Something worth opening and checking. Not a conviction, and not proof of bad faith.
“Rewrite” (sometimes called a pivot in research notes)
A stretch of time when a large share of wording that had been stable for a long time was replaced and the replacement stayed. Open the before-and-after view and read it yourself.
“Candidate” vs “confirmed”
A candidate is a time window selected for closer inspection. It may come from one sharp interval or from sustained loss across roughly twelve months. Confirmed means the underlying revision history also shows the required collapse of long-lived wording. Neither term means the rewrite was biased.
“Persistence-weighted loss” and “PWR mass”
Persistence-weighted loss is the share of wording lost across an interval, with more weight given to text that had survived in more earlier snapshots. PWR mass is the absolute weighted amount lost. The percentage describes how much of that article changed; the mass helps compare the scale of candidate rewrites without letting a tiny article outrank a large one on percentage alone.
“Durable spine” and “durable-spine drop”
The durable spine is the more persistent half of the wording present at the start of a candidate window. Its drop is the percentage-point decline in how much of that wording survives across the whole window. It measures established wording loss, not whether the resulting text is better or worse.
“Rewrite episode” and “peak interval”
A rewrite episode is a bounded window in which established wording was substantially replaced. Its peak interval is the single snapshot-to-snapshot interval with the largest persistence-weighted loss; it may be only one part of the wider episode.
“Coarse scan” vs “exact check”
The coarse scan compares sampled article snapshots to find candidate windows. The exact check then searches the revisions inside each investigated window to test whether its durable spine really collapsed. A candidate can therefore be rejected without implying that the article never changed.
“Redline” and “receipt”
A redline shows wording removed and added between two versions. A receipt is the structured record behind a result: the window or revisions inspected, measurements made, and decision reached. The redline supports reading the change; the receipt supports auditing how WikiDrift classified it.
“Snapshot,” “mature interval,” and “token”
A snapshot is a version selected near a sampling date for interval comparison. A mature interval starts only after the article has enough tracked text for the persistence measurement; earlier intervals are excluded, not treated as negative findings. A token is one tracked unit of text, usually a word or punctuation mark.
“Overview”
The short briefing at the top of an article page. Begin there; use the other tabs only if you want detail.
“Framing”
How the writing presents a subject — more critical, more sympathetic, or more neutral — beyond the bare facts. Two languages can agree on a date and still frame the story differently.
“Facts check”
A short list of basic questions (who / when / where / how many). Editions may agree, differ, contradict, or simply not say enough to tell (insufficient). That is not a score of who is right.
“Citations changed”
Which domains and types of sources (books, news, journals, web pages) grew or shrank across a rewrite. We show the mix only. We do not mark sources as trustworthy or untrustworthy.
“Vocabulary shift”
Which words became more or less common. Helpful for noticing a topic change; easy to over-read. Ignore it if the rewrite text already makes the story clear.
“Concentrated authorship”
A large share of today’s text comes from a small number of accounts. Common on specialist pages. Context only — not evidence of a conspiracy.
“Exact-event attribution”
For one confirmed event, the public edit history is read as an ordered sequence between stable before and after states. The receipt separates every observed addition, removal, and restoration from the removals and replacement wording still standing at the after state. Raw per-revision counts reproduce the displayed shares. It describes observable revision history, not bias, motive, coordination, or misconduct.
“Heavily contested” vs “not heavily contested”
Whether editors spent a lot of effort undoing each other. A quiet rewrite can still matter. A loud fight does not prove anyone is wrong.
“Found via discovery”
This page showed up because large deletions on another related article pointed investigators here. It was then checked on its own history. The trail never condemns a page by itself.
“Control” topics
Quiet subjects (for example basic science) included so we can see what ordinary long-term editing looks like — and avoid treating every big rewrite as a scandal.
If you only do three things
- Read Overview.
- Open Rewrite and read the actual text change.
- Click through to Wikipedia when you want the original version.