Exploiting stance hierarchies for cost-sensitive stance detection of Web documents

Fact checking is an essential challenge when combating fake news. Identifying documents that agree or disagree with a particular statement (claim) core task in this process. In context, stance detection aims at identifying the position (stance) of document towards claim. Most approaches address through classification models do not consider highly imbalanced class distribution. Therefore, they are particularly ineffective detecting minority classes (for instance, ‘disagree’), even though such instances crucial for tasks as fact-checking by providing evidence false claims. paper, we exploit hierarchical nature which allows us to propose modular pipeline cascading binary classifiers, enabling performance tuning on per step and basis. We implement our approach combination neural traditional highlight misclassification costs classes. Evaluation results demonstrate state-of-the-art its ability significantly improve important ‘disagree’ class.