An argumentative approach for discovering relevant opinions in Twitter with probabilistic valued relationships

Abstract

Twitter is one of the most widely used social networks when it comes to sharing and criticizing relevant news and events. In order to understand the major opinions accepted and rejected in different domains by Twitter users, in a recent work we developed an analysis system based on valued abstract argumentation to model and reason about the social acceptance of tweets, considering different information sources from the social network. Given a Twitter discussion, the system outputs the set of accepted tweets from the discussion, considering two kinds of relationship between tweets: criticism and support. In this paper, we introduce and investigate a natural extension of the system, in which relationships between tweets are associated with a probability value, indicating the uncertainty that the relationships hold. An important element in our system is the notion of an uncertainty threshold, which characterizes how much uncertainty on probability values we are willing to tolerate: given an uncertainty threshold $\alpha$, we reject criticism and support relationships with probability below $\alpha$. We also extend our analysis system by incorporating support propagation when computing the social relevance of tweets. To this end, we extend the abstract argumentation framework with a new valuation function that propagates the support between tweets by taking into account not only the social relevance of tweets but also the probability that the support relationship holds, provided that it is above the specified uncertainty threshold $\alpha$. In order to test these new extensions, we analyze different Twitter discussions from the political domain. Our analysis shows that the social support of the accepted tweets is typically much stronger than the one for the rejected tweets. Also, the set of accepted tweets seems to be very stable with respect to changes to the social support of the tweets, and therefore even when considering support propagation we mainly observe differences in such set when using the more permissive probability thresholds.

This work was partially funded by the Spanish MICINN Projects TIN2014-53234-C2-2- R, TIN2015-71799-C2-2-P and ENE2015-64117-C5-1- R. This research article has received a grant for its linguistic revision from the Language Institute of the University of Lleida (2017 call). The authors would like to thank anonymous reviewers for providing helpful comments to improve the paper.