DART: a Dataset of Arguments and their Relations on Twitter

DART: a Dataset of Arguments and their Relations on Twitter Tom Bosc, Elena Cabrio, Serena Villata INRIA Sophia Antipolis, University of Nice Sophia Antipolis, CNRS - I3S 930 route des Colles - Bˆat. Les Templiers, 06903 Sophia Antipolis CEDEX, France tom.bosc@inria.fr, {cabrio, villata}@i3s.unice.fr Abstract The problem of understanding the stream of messages exchanged on social media such as Facebook and Twitter is becoming a major challenge for automated systems. The tremendous amount of data exchanged on these platforms as well as the specific form of language adopted by social media users constitute a new challenging context for existing argument mining techniques. In this paper, we describe a resource of natural language arguments called DART (Dataset of Arguments and their Relations on Twitter) where the complete argument mining pipeline over Twitter messages is considered: (i) we identify which tweets can be considered as arguments and which cannot, and (ii) we identify what is the relation, i.e., support or attack, linking such tweets to each other. Keywords: Argument mining, social media, textual entailment 1. Introduction Understanding and interpreting streams of messages ex- changed on social networks such as Facebook and Twitter in an automated way is becoming a major challenge in Nat- ural Language Processing (NLP). This tremendous amount of data provides a huge amount of information to be mined, in particular for monitoring opinions about controversial is- sues in political scenarios and to manage reputation analy- sis of brands advertising. In this scenario, users express their viewpoints by means of textual messages posted on- line. In this paper, we present a dataset to support the development of frameworks tackling the argument mining pipeline over Twitter data, as highlighted in (Gabbriellini and Torroni, 2012). More precisely, in this paper, we an- swer the following research questions: • How to distinguish from a stream of tweets those tex- tual exchanges that are arguments and those that are not? • How to identify the relationship, i.e., attack or support, among two (or more) tweets to study the evolution of a certain discussion? To answer these questions we rely on argumentation the- ory (Rahwan and Simari, 2009), and more precisely, on the argument mining pipeline (Lippi and Torroni, 2016). Ar- guments are supposed to support, contradict, explain state- ments, and are used to support decision making. Abstract argumentation theory provides a framework that allows to detect winning arguments from a set of arguments and the relations among them. Applying the argument min- ing pipeline to Twitter means addressing the following is- sues: (i) the set of tweets to be analyzed has to be identified, i.e., a selection of potential tweet-arguments that express an idea or an opinion on a given topic, (ii) the identified tweet- arguments are coupled with each other forming meaningful pairs, and (iii) the relations, i.e., support or attack, among the tweet-arguments of each pair have to be identified. Fi- nally, identifying and linking arguments together allows for a synthetic view of the ongoing discussion. Differently from texts extracted from traditional sources, …

Content truncated to 500 words