Stance Evolution and Twitter Interactions in an Italian Political Debate

Abstract

The number of communications and messages generated by users on social media platforms has progressively increased in the last years. Therefore, the issue of developing automated systems for a deep analysis of users' generated contents and interactions is becoming increasingly relevant. In particular, when we focus on the domain of online political debates, interest for the automatic classification of users' stance towards a given entity, like a controversial topic or a politician, within a polarized debate is significantly growing. In this paper we propose a new model for stance detection in Twitter, where authors' messages are not considered in isolation, but in a diachronic perspective for shedding light on users' opinion shift dynamics along the temporal axis. Moreover, different types of social network community, based on retweet, quote, and reply relations were analyzed, in order to extract network-based features to be included in our stance detection model. The model has been trained and evaluated on a corpus of Italian tweets where users were discussing on a highly polarized debate in Italy, i.e. the 2016 referendum on the reform of the Italian Constitution. The development of a new annotated corpus for stance is described. Analysis and classification experiments show that network-based features help in detecting stance and confirm the importance of modeling stance in a diachronic perspective.

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

“…Our experiments are applied on the SemEval stance detection benchmark dataset [40], which contains a set of over 4,000 tweets labeled by stance towards five different topics. The five topics covers multiple domains not just politics, which makes the dataset ideal to examine the generalisability of the stance detection models, unlike most of work in literature that typically focus on studying one political topic at a time [14,33,34,36]. Our results show that training a classification model on pure user network features outperforms the state-of-the-art baseline system [40] which is trained on multiple features extracted from the tweets text content.…”

“…detection to analyze social media as main component of investigating the users aligns toward a given topic or entity [3,7,33,34,36,40].…”

“…Some users prefer to interact quietly on social media using other means of interactions instead of directly posting or sharing contents, such as following others and liking posts [27]. Most of stance detection studies used the network representation of the active users only and overlooked the silent users [14,34,36].…”

Searching for polarization in signed graphs: a local spectral approach

“…opposition). Referendum [20] records the 2016 Italian Referendum on Twitter: an edge is negative if two users are classified with different stances, and positive otherwise. Bitcoin is a trust network of Bitcoin users.…”

Discovering Polarized Communities in Signed Networks

“…High-landTribes 1 represents the alliance structure of the Gahuku-Gama 1 konect.cc Table 1: Signed networks used: number of vertices and edges; ratio of negative edges (ρ − = |E − | |E + ∪E − | ); L 1 -norm of the eigenvector corresponding the largest eigenvalue of A (∥v∥ 1 ); and, ratio of non-zero elements of A (δ = 2 |E + ∪E − | |V |(|V |−1) ). [32] records Twitter data about the 2016 Italian Referendum: an interaction is negative if two users are classified with different stances, and is positive otherwise. Slashdot 2 contains friend/foe links between the users of Slashdot.…”