Stance Classification through Proximity-based Community Detection

Abstract

Numerous domains have interests in studying the viewpoints expressed online, be it for marketing, cybersecurity, or research purposes with the rise of computational social sciences. Current stance detection models are usually grounded on the specificities of some social platforms. This rigidity is unfortunate since it does not allow the integration of the multitude of signals informing effective stance detection. We propose the SCSD model, or Sequential Community-based Stance Detection model, a semi-supervised ensemble algorithm which considers these signals by modeling them as a multi-layer graph representing proximities between profiles. We use a handful of seed profiles, for whom we know the stance, to classify the rest of the profiles by exploiting like-minded communities. These communities represent profiles close enough to assume they share a similar stance on a given subject. Using datasets from two different social platforms, containing two to five stances, we show that by combining several types of proximity we can achieve excellent results. Moreover, we compare the proximities to find those which convey useful information in term of stance detection.

Your Stance is Exposed! Analysing Possible Factors for Stance Detection on Social Media

“…Furthermore the study of [48] used the interactions between users with focus on the retweet and reply network as way to cluster the users with the same views. Similarly, [22] used the users interactions and the textual features to model the users stance by proximity graphs.…”

Cross-issue correlation based opinion prediction in cyber argumentation

“…Researchers have also used social media data to predict user reactions on different social events, such as the 2015 Paris Terror Attack [33]. Many researchers used social media data to predict users' opinions on important issues/people using different algorithms (see [12,38,57] for examples). These works mostly looked at predicting an individual or group's opinion on a single issue using the related textual content on that issue only.…”

“…Typically users participate only in a few issues and do not discuss other issues in the system. Existing opinion analysis models focus mostly on analyzing user opinion on the participated issues only [11,12]; often, the scope of such analysis is limited. These missing opinion values on the non-participated issues may be crucial, and discarding these values may yield an incomplete analysis of the underlying discussion.…”

Les téléphones mobiles, un outil de désinformation ? La circulation des informations peu fiables dans Twitter lors de la campagne présidentielle française de 2017

“…Dans le cadre de cet article, nous cherchons à prolonger ce résultat. Après avoir évalué si la propagation de la désinformation est plus fortement associée à l'usage de l'application mobile de Twitter, il s'agira d'observer les éventuelles différences entre les principales communautés politiques françaises : peut-on considérer que les formes de participation développées à partir de l'application mobile de Twitter tendent à amplifier la tendance de certaines communautés politiques, en l'occurrence Les Républicains et le (Fraisier et al, 2018). Une fois en possession de cet échantillon de (re)tweets, il a été nécessaire de détecter les communautés politiques dans le but de pouvoir analyser la circulation des sources d'information au sein de chacune d'elle.…”

“…Par exemple, nous avons isolé la communauté politique du PS en repérant le groupe de comptes Twitter qui contient les tweets émis par le candidat Benoît Hamon et les autres cadres du Parti socialiste. Au terme de ce travail, une vérification manuelle a été réalisée pour tester la robustesse de cette méthode de détection des communautés et pour décrire plus précisément la composition de ce jeu de données (Fraisier et al, 2018).…”