Contextual information integration for stance detection via cross-attention

TL;DR

This paper proposes a novel method for stance detection that integrates contextual information from diverse sources as text, improving accuracy and robustness, especially for unseen targets, by leveraging cross-attention mechanisms and large language models.

Contribution

It introduces a new approach to incorporate heterogeneous contextual information as text into stance detection models, outperforming baselines and enhancing robustness to noise.

Findings

Outperforms competitive baselines on a large stance detection benchmark.

More robust to noisy context and reduces unwanted label-target correlations.

Independent of the pretrained language model used.

Abstract

Stance detection deals with identifying an author's stance towards a target. Most existing stance detection models are limited because they do not consider relevant contextual information which allows for inferring the stance correctly. Complementary context can be found in knowledge bases but integrating the context into pretrained language models is non-trivial due to the graph structure of standard knowledge bases. To overcome this, we explore an approach to integrate contextual information as text which allows for integrating contextual information from heterogeneous sources, such as structured knowledge sources and by prompting large language models. Our approach can outperform competitive baselines on a large and diverse stance detection benchmark in a cross-target setup, i.e. for targets unseen during training. We demonstrate that it is more robust to noisy context and can regularize for unwanted correlations between labels and target-specific vocabulary. Finally, it is independent of the pretrained language model in use.