The OnForumS corpus from the Shared Task on Online Forum Summarisation at MultiLing 2015

The OnForumS corpus from the Shared Task on Online Forum Summarisation at MultiLing 2015 Mijail Kabadjov∗, Udo Kruschwitz∗, Massimo Poesio∗, Josef Steinberger†, Marc Poch‡, Hugo Zaragoza‡ ∗University of Essex, Colchester, UK, {malexa,udo,poesio}@essex.ac.uk †University of West Bohemia, Pilsen, Czech Republic, jstein@kiv.zcu.cz ‡Websays, Barcelona, Spain {marc.poch,hugo.zaragoza}@websays.com Abstract In this paper we present the OnForumS corpus developed for the shared task of the same name on Online Forum Summarisation (OnForumS at MultiLing’15). The corpus consists of a set of news articles with associated readers’ comments from The Guardian (English) and La Repubblica (Italian). It comes with four levels of annotation: argument structure, comment-article linking, sentiment and coreference. The former three were produced through crowdsourcing, whereas the latter, by an experienced annotator using a mature annotation scheme. Given its annotation breadth, we believe the corpus will prove a useful resource in stimulating and furthering research in the areas of Argumentation Mining, Summarisation, Sentiment, Coreference and the interlinks therein. Keywords: Online Forums, Summarization, Argument Structure, Sentiment Analysis 1. Introduction Internet or online forums are discussion websites where people can hold asynchronous conversations in the form of posted messages. Much work has been devoted in recent years on mining and analysing online forums – as in search (Bhatia and Mitra, 2010; Seo et al., 2009), question an- swering (Ding et al., 2008; Hong and Davison, 2009), clas- sification of argumentative propositions (Park and Cardie, 2014) and automatic summarisation (Giannakopoulos et al., 2015) – and such work, to a large extent is fuelled by the creation and public release of online forums corpora of var- ious kinds (e.g., (Wang et al., 2011) and the boards.ie Forums Dataset as part of the ICWSM 2012 conference1). One type of online forums which is increasingly popular is that of readers’ discussions taking place on news publish- ers sites, such as The Guardian2 or Le Monde3. There is strong interest in such forums, their mining and analysis, by a broad range of information seekers, as are journalists, news editors and trend and media monitors. Yet, to our knowledge currently there is no news-forums corpus easily available, most likely due to copyright restrictions on such data, but also due to its novelty. Furthermore, the high volume of reader-supplied comments on news-forums suggests the need for automated meth- ods to summarise this content, a problem addressed by the shared task on Online Forum Summarisation (OnForumS) at MultiLing 20154 (Kabadjov et al., 2015; Giannakopoulos et al., 2015). In addition to summarisation, the OnForumS 1http://www.icwsm.org/2012/submitting/ datasets/ 2http://www.theguardian.com/ 3http://www.lemonde.fr/ 4http://www.sigdial.org/workshops/ conference16/sessions.html task addressed the problem of argument structure identifi- cation which is particularly relevant in news forums. In this paper we present the corpus that emerged as a result from the OnForumS shared task and which is aimed at fill- ing the gap in availability of news-forum corpora. The cor- pus consists of a set of news articles with associated com- ments from The Guardian (English), which has a comment moderating system in place, and La Repubblica5 (Italian). The choice of the sources was mainly driven …

Content truncated to 500 words