CN1C

CN1C brings propaganda-technique detection to any web page. Select text and click the CN1C button, or scan a whole article, to see: - A verdict (propaganda / neutral) with confidence - The coarse intent behind the text (e.g. distort reality, shift blame, delegitimize opponent, manufacture consent, confuse & distract) - The exact spans of text driving that classification, highlighted inline in the page, color-coded by intent - Entity framing — who's cast as protagonist, antagonist, or innocent, with a live role-map overlay connecting each entity on the page to its role Powered by an XLM-R multi-task model developed at the Quality & Usability Lab, TU Berlin, as part of the CN1C research project (a continuation of "Check News in 1 Click", EACL 2024). Multilingual: English, German, French, Italian, Romanian, Russian, Ukrainian. By default, analysed text is sent to the CN1C model server at checknewsin1.click for classification. When the CN1C model finds no propaganda, a research LLM at the GWDG Academic Cloud cross-checks the verdict so that topics outside the training data are not missed. You can point the extension at a different model server (including one running on your own machine) in the popup settings. See the full privacy policy at https://checknewsin1.click/privacy.html. This is a research prototype. It is strongest on Russia–Ukraine-war-related news text; general text is out of distribution for the current model.