Texty.org.ua

Texty.org.ua is a Kyiv-based data-journalism outfit, founded in 2010 and run as an independent non-profit, that describes itself as Ukraine’s first centre for data journalism. Where a firm like Molfar turns open-source material into target packages, Texty works the information battlefield: it builds machine-learning systems to detect and map the propaganda and manipulation flowing out of Russian media and Ukrainian Telegram channels. Its founders include editor Roman Kulchynsky and data lead Anatoliy Bondarenko.

The technical core is a text classifier the team has refined since 2018. After early prototypes built on a random forest and then a plain LSTM proved too weak on scarce training data, Texty moved to ULMFiT, a transfer-learning architecture fine-tuned for Ukrainian and Russian and implemented in PyTorch. Editors hand-labelled roughly eight thousand articles to teach it; the model then scored some two million. That engine powered “Hot Disinfo from Russia,” which recognised and tracked disinformation campaigns across channels and won a Sigma Award in 2020, and it now drives Topic Radar, an AI-assisted watch over around a thousand Russian websites and Telegram channels that surfaces weekly propaganda trends.

The work turned directly toward the war. Roughly a week before the February 2022 invasion, Texty published an interactive game to help readers tell Russian armour from Ukrainian, and it has used satellite imagery to study Russian reconnaissance activity alongside its disinformation monitoring. The combination of trained classifiers and conventional reporting has drawn repeated recognition, including further Sigma Awards and a European Press Prize, and marks Texty as one of the clearer examples of machine learning applied to the contest over wartime narrative.