This pipeline was used to generate proposed disinformation trigger events.
Human labelling was then completed to seperate quality analysis from useless results.
This model was trained on those labels, in order to complete a filtering process automatically.
