Research
Measured, published,
and repeatable in method.
Every claim WriteX makes about its engine should trace to a published evaluation with its method and its limits. This is where those evaluations live.
How WriteX evaluates
We test the engine the way its users judge it: on real sentences, against a reference written by a linguist, with the competitors run under the same conditions. Every system receives the same text, the same instruction and the same scoring, and contributes one output per sentence. Detection is scored on exact spans, correction only where detection succeeded, and the whole sentence once more at the end. We publish precision, recall, F1 and F0.5 together, so the choice of headline metric is visible.
Three rules hold on every page. Each number carries its evidence status: Measured, Team-reported, Qualitative or Product capability. Comparisons name the competitor and the version, and say what it did well. And the limitations are on the page, not in a footnote: what the prompt constrained, which figures are cited estimates, and the fact that Lisan Research runs its own evaluations and does not call them independently verified.
The canonical copies of the reports, with the citation and the version history, sit on the Lisan research hub. To repeat an evaluation on your own material, the editor is free to start, and support@lisan.com is the address for methodology questions.
Every word your institution writes, checked.
Official letters to social posts, Arabic and English alike. One assistant that proofreads, enforces your style, and explains every correction, wherever your teams write.
12 error types · Institutional style guides · Works in Office, browsers, and the web editor