Unsupervised User Stance Detection on Twitter

Abstract

We present a highly effective unsupervised framework for detecting the stance of prolific Twitter users with respect to controversial topics. In particular, we use dimensionality reduction to project users onto a low-dimensional space, followed by clustering, which allows us to find core users that are representative of the different stances. Our framework has three major advantages over pre-existing methods, which are based on supervised or semi-supervised classification. First, we do not require any prior labeling of users: instead, we create clusters, which are much easier to label manually afterwards, e.g., in a matter of seconds or minutes instead of hours. Second, there is no need for domain-or topic-level knowledge either to specify the relevant stances (labels) or to conduct the actual labeling. Third, our framework is robust in the face of data skewness, e.g., when some users or some stances have greater representation in the data. We experiment with different combinations of user similarity features, dataset sizes, dimensionality reduction methods, and clustering algorithms to ascertain the most effective and most computationally efficient combinations across three different datasets (in English and Turkish). We further verified our results on additional tweet sets covering six different controversial topics. Our best combination in terms of effectiveness and efficiency uses retweeted accounts as features, UMAP for dimensionality reduction, and Mean Shift for clustering, and yields a small number of high-quality user clusters, typically just 2-3, with more than 98% purity. The resultant user clusters can be used to train downstream classification. Moreover, our framework is robust to variations in the hyper-parameter values and also with respect to random initialization.

News Consumption in Time of Conflict: 2021 Palestinian-Israel War as an Example

“…After filtration, we were left with 357,725 tweets from 130,110 different users. Using unsupervised stance detection Darwish et al (2019) over the most active 20,000 users, we attempted to identify underlying groups with varying stances. The method was able to identify the stances of 8,034 users, with 6,495 siding with the Palestinian side and 1,539 siding with the Israeli side.…”

“…Label propagation is also an effective semisupervised method that propagates labels in a network based on follow or retweet relationships (Borge-Holthoefer et al, 2015;Weber, Garimella, and Batayneh, 2013) or the sharing of identical tweets (Darwish, 2018;Kutlu, Darwish, and Elsayed, 2018;. More recent work projects user onto a two dimensional space and then uses clustering to perform unsupervised stance detection (Darwish et al, 2019), with the best setup involving the use of UMAP for projection, mean shift for clustering, and the retweeted accounts as user features. Though this method may have relatively low recall, it has nearly perfect precision.…”

Embeddings-Based Clustering for Target Specific Stances: The Case of a Polarized Turkey

“…Other methods for user stance detection include: collective classification (Duan et al, 2012), where users in a network are jointly labeled, and projecting users into a lower dimensional user space prior to classification (Darwish, Magdy, and Zanouda, 2017). More recent work projects user onto a two dimensional space and then uses clustering to perform unsupervised stance detection (Darwish et al, 2019). In their work, the best setup used UMAP for projection, mean shift for clustering, and the retweeted accounts as user features.…”

Embedding-based Qualitative Analysis of Polarization in Turkey

“…Other methods for user stance detection include: collective classification [11], where users in a network are jointly labeled, and projecting users into a lower dimensional user space prior to classification [9]. More recent work projects user onto a two dimensional space then uses clustering to perform unsupervised stance detection [10].…”

“…Such are consistent with election results. Tirit is popular in cities such as Konya 7 , Samsun 8 , and Sanliurfa 9 , and cig kofte (raw meatball) is popular in SanliUrfa 10 . Voters in Konya, Samsun and Sanliurfa, gave Erdogan 74%, 61%, and 65% of the votes respectively 11 .…”