Pheme Stance Detection

The Pheme project worked on rumour and veracity in web text. One critical part of this framework was the ability to detect how commenters were reacting to social media claims. This is also know as their stance towards a claim.

This page provides access to the Pheme stance detection software developed in WP4. The tool beat state-of-the-art performance on the RumourEval dataset using classical ML methods like SVM and decision trees in combination with word embedding-derived features.

The system and data are available upon request (contacts in GATE: Kalina Bontcheva, Leon Derczynski)