How to trolls talk?

We use machine learning to identify patterns in the language of Russian government trolls.

To analyze content, either click the button below or tag @TrackTheTrolls in reply to a Tweet.

For more information about the inner workings of our model, see our extended explanation.

Please enter your text below

Unfortunately, the troll content analyzer has been archived due to financial constraints and is no longer operational.

0/240

All percentages shown in the tool above indicate the scaled confidence (from -100% to +100%) of the model that a text is theoretically a Russian IRA troll Tweet (positive values) or an organic Tweet (negative values) in an evenly distributed dataset.

All percentages shown in the tool above indicate the scaled confidence (from -100% to +100%) of the model that a text is theoretically a Russian IRA troll Tweet (positive values) or an organic Tweet (negative values) in an evenly distributed dataset.
This is an educational tool; this model is not meant to detect real Russian trolls.
Our model is tuned for 2018 discourse on Twitter. It may be less accurate when analyzing content from different time periods.
Our model was built using scikit-learn. For more information, see our extended explanation.