Semi-supervised Stance Detection of Tweets Via Distant Network Supervision

TL;DR

SANDS is a semi-supervised stance detection method that leverages social network homophily and deep feature views to effectively classify stance in social media tweets with minimal labeled data.

Contribution

The paper introduces SANDS, a novel semi-supervised stance detector utilizing distant network supervision and multiple feature views, outperforming existing baselines on large, real-world datasets.

Findings

SANDS achieves macro-F1 scores of 0.55 and 0.49 on US and India datasets.

Outperforms 17 baseline models, especially on minority and noisy data.

Ablation studies reveal the importance of textual and network signals.

Abstract

Detecting and labeling stance in social media text is strongly motivated by hate speech detection, poll prediction, engagement forecasting, and concerted propaganda detection. Today's best neural stance detectors need large volumes of training data, which is difficult to curate given the fast-changing landscape of social media text and issues on which users opine. Homophily properties over the social network provide strong signal of coarse-grained user-level stance. But semi-supervised approaches for tweet-level stance detection fail to properly leverage homophily. In light of this, We present SANDS, a new semi-supervised stance detector. SANDS starts from very few labeled tweets. It builds multiple deep feature views of tweets. It also uses a distant supervision signal from the social network to provide a surrogate loss signal to the component learners. We prepare two new tweet datasets comprising over 236,000 politically tinted tweets from two demographics (US and India) posted by over 87,000 users, their follower-followee graph, and over 8,000 tweets annotated by linguists. SANDS achieves a macro-F1 score of 0.55 (0.49) on US (India)-based datasets, outperforming 17 baselines (including variants of SANDS) substantially, particularly for minority stance labels and noisy text. Numerous ablation experiments on SANDS disentangle the dynamics of textual and network-propagated stance signals.