Learning to generate and evaluate fact-checking explanations with transformers

Darius Feher, Abdullah Salem Khered, Hao Zhang, Riza Batista-Navarro, Viktor Schlegel

Abstract: Highlights•A dataset for fact-checking which includes explanations written by journalists.•Transformer models for generating human-accessible fact-checking explanations.•Multi-dimensional annotations reflecting explanation quality judgements.•A metric learning model scoring explanations aligned with these judgements.