A Retrospective Analysis of the Fake News Challenge Stance Detection Task

TL;DR

This paper critically analyzes the 2017 Fake News Challenge stance detection task, revealing biases in evaluation metrics, proposing a new metric, and introducing a novel model and dataset to improve future fake news detection research.

Contribution

It provides an in-depth analysis of FNC-1 systems, proposes a new evaluation metric, and introduces a feature-rich stacked LSTM model and a new dataset for better stance classification.

Findings

FNC-1's evaluation metric favors majority class

The new F1-based metric changes system rankings

The proposed stacked LSTM performs well on minority classes

Abstract

The 2017 Fake News Challenge Stage 1 (FNC-1) shared task addressed a stance classification task as a crucial first step towards detecting fake news. To date, there is no in-depth analysis paper to critically discuss FNC-1's experimental setup, reproduce the results, and draw conclusions for next-generation stance classification methods. In this paper, we provide such an in-depth analysis for the three top-performing systems. We first find that FNC-1's proposed evaluation metric favors the majority class, which can be easily classified, and thus overestimates the true discriminative power of the methods. Therefore, we propose a new F1-based metric yielding a changed system ranking. Next, we compare the features and architectures used, which leads to a novel feature-rich stacked LSTM model that performs on par with the best systems, but is superior in predicting minority classes. To understand the methods' ability to generalize, we derive a new dataset and perform both in-domain and cross-domain experiments. Our qualitative and quantitative study helps interpreting the original FNC-1 scores and understand which features help improving performance and why. Our new dataset and all source code used during the reproduction study are publicly available for future research.