You can simply rely on communities for a robust characterization of stances

Abstract

We show that the structure of communities in social me- dia provides robust information for weakly supervised approaches to assign stances to tweets. Using as seed the SemEval 2016 Stance Detection Task annotated data, we retrieved a high number of topically related tweets. We then propagated information from the manually an- notated seed to the retrieved tweets and thus obtained a bigger training corpus. Classifiers trained with this bigger, weakly supervised dataset reach similar or better performance than those trained with the manually annotated seed. In addition, they are more robust with respect to common manual annotator errors or biases and they have arguably more coverage than smaller datasets. Weakly supervised approaches, most notably self- supervision, commonly suffer from error propagation. Interestingly, communities seem to provide a structure that constrains error propagation. In particular, weakly supervised classifiers that incorporate community struc- ture are more robust with respect to class imbalance. Additionally, this is a straightforward, transparent ap- proach, using standard tools and pipelines, cheaper and faster than methods like crowd sourcing for manual an- notations. Thus it facilitates adoption, interpretability and accountability.

Strength in coalitions: Community detection through argument similarity

“…Concerning the other topic inherent to our work, Furman et al [19] present a method to find discursive communities in social media. The authors analyze small and comprehensive annotated datasets using standard tools like graphs, an algorithm for Modularity Maximization, and a supervised classifier.…”

“…The identification of communities in social media and the detection of stances in Tweets has become increasingly important in recent times [15,18,19,26,28] as a result of the tangible effect that these platforms have on the public opinion. In this domain, identifying communities implies analyzing the position of contributing agents concerning a particular topic or their respective argumentative stance; several tools can be used for this purpose, for instance [19] describes an approximation solution based on a supervised classifier that finds stances, and classifies them, over a graph representation. Most of the explored work on identifying stances focuses on the classification of tweets as "in favor" (support), "against" (dispute), or "neutral" (comments or questions) regarding a previous tweet in a conversation [35,37].…”