Marking Irony Activators in a Universal Dependencies Treebank: The Case of an Italian Twitter Corpus

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 5098–5105 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 5098 Marking Irony Activators in a Universal Dependencies Treebank: The Case of an Italian Twitter Corpus Alessandra Teresa Cignarella1,2, Manuela Sanguinetti1, Cristina Bosco1, Paolo Rosso2 1. Dipartimento di Informatica, Università degli Studi di Torino, Italy 2. Universitat Politècnica de València, Spain {cigna|msanguin|bosco}@di.unito.it, prosso@disc.upv.es Abstract The recognition of irony is a challenging task in the domain of Sentiment Analysis, and the availability of annotated corpora may be crucial for its automatic processing. In this paper we describe a fine-grained annotation scheme centered on irony, in which we highlight the tokens that are responsible for its activation (irony activators), and their morpho-syntactic features. As our case study we therefore introduce a recently released Universal Dependencies treebank for Italian which includes ironic tweets: TWITTIRÒ-UD. For the purposes of this study, we enriched the existing annotation in the treebank with a further level that includes irony activators. A description and discussion of the annotation scheme is provided with a definition of irony activators and the guidelines for their annotation. This qualitative study on the different layers of annotation applied on the same dataset can shed some light on the process of human annotation, and irony annotation in particular, and on the usefulness of this representation for developing computational models of irony to be used for training purposes. Keywords: Twitter, Italian, Irony, Irony Activators, Syntax, Universal Dependencies 1. Introduction Irony detection proves to be an important task in the do- main of Sentiment Analysis, due to the fact that the pres- ence of ironic devices in a text may reverse the polarity of an opinion, which can be expressed, for example, with pos- itive words but intending a negative meaning, therefore un- dermining the performance of Sentiment Analysis systems (Bosco et al., 2013; Reyes et al., 2013; Barbieri et al., 2014; Ghosh et al., 2015; Hernández Farías et al., 2015; Joshi et al., 2015). As a matter of fact, the recognition of irony and the identi- fication of pragmatic and linguistic devices that activate it have always been a controversial topic (Grice, 1975; Grice, 1978; Sperber and Wilson, 1981), and a challenging task for both human annotators and automatic tools. Additionally, the automatic treatment of such phenomenon is further complicated by the co-occurrence of simi- lar forms of speech such as sarcasm or satire (Hernán- dez Farías and Rosso, 2016; Joshi et al., 2017; Ravi and Ravi, 2017; Zhang et al., 2019) and by the text domain. The application of tools for irony detection has been espe- cially focused on micro-blogging platforms (such as Twit- ter), since irony is a pragmatic phenomenon widely ex- ploited by users online. In the last five years in particular, there has been a stronger attention from the NLP community to the topic of irony de- tection, and this is also attested by the proposal of various shared tasks for different languages and …

Content truncated to 500 words