Measuring What Counts: The case of Rumour Stance Classification
This addresses a methodological problem for researchers in natural language processing and social media analysis by improving evaluation in rumour stance classification, though it is incremental as it focuses on metrics rather than new models.
The paper identifies that widely used evaluation metrics (accuracy and macro-F1) are not robust for the four-class imbalanced task of rumour stance classification, as they favor systems skewed toward the majority class, and proposes new metrics that are robust to imbalance and better recognize informative minority classes.
Stance classification can be a powerful tool for understanding whether and which users believe in online rumours. The task aims to automatically predict the stance of replies towards a given rumour, namely support, deny, question, or comment. Numerous methods have been proposed and their performance compared in the RumourEval shared tasks in 2017 and 2019. Results demonstrated that this is a challenging problem since naturally occurring rumour stance data is highly imbalanced. This paper specifically questions the evaluation metrics used in these shared tasks. We re-evaluate the systems submitted to the two RumourEval tasks and show that the two widely adopted metrics -- accuracy and macro-F1 -- are not robust for the four-class imbalanced task of rumour stance classification, as they wrongly favour systems with highly skewed accuracy towards the majority class. To overcome this problem, we propose new evaluation metrics for rumour stance detection. These are not only robust to imbalanced data but also score higher systems that are capable of recognising the two most informative minority classes (support and deny).