INAOE-UPV at SemEval-2018 Task 3: An Ensemble Approach for Irony Detection in Twitter

Proceedings of the 12th International Workshop on Semantic Evaluation (SemEval-2018), pages 594–599 New Orleans, Louisiana, June 5–6, 2018. ©2018 Association for Computational Linguistics INAOE-UPV at SemEval-2018 Task 3: An Ensemble Approach for Irony Detection in Twitter Delia Iraz´u Hern´andez Far´ıas1, Fernando S´anchez-Vega1 Manuel Montes-y-G´omez1,2, and Paolo Rosso2 1 Instituto Nacional de Astrof´ısica, ´Optica y Electr´onica (INAOE), Mexico 2 PRHLT Research Center, Universitat Polit`ecnica de Val`encia, Spain {dirazuherfa,fer.callotl,mmontesg}@inaoep.mx prosso@dsic.upv.es Abstract This paper describes an ensemble approach to the SemEval-2018 Task 3. The proposed method is composed of two renowned methods in text classification together with a novel ap- proach for capturing ironic content by exploit- ing a tailored lexicon for irony detection. We experimented with different ensemble settings. The obtained results show that our method has a good performance for detecting the presence of ironic content in Twitter. 1 Introduction Social media provide a perfect scenario for ex- ploiting language beyond its literal sense by us- ing figurative language devices such as, for ex- ample, irony. Correctly identifying the real inten- tion behind user-generated content is a big chal- lenge for different areas related to computational linguistics. For example, in Sentiment Analy- sis (SA), the presence of irony could undermine the performance of systems dedicated to this task (Hern´andez Far´ıas and Rosso, 2016). There are several disciplines studying irony from different perspectives. The most prevalent definition is that from Grice (1975), stating that the function of irony is to effectively communicate the opposite of the literal interpretation a given utterance. Nowadays, with the growing interest in irony detection, there are several approaches1 for ad- dressing such an interesting task. Probably, the most widely used is that exploiting characteristics extracted from the text (such as n-grams, punctu- ation marks, part-of-speech labels, among others) on its own (Riloff et al., 2013; Pt´aˇcek et al., 2014). Inherent aspects of irony such as its very subjec- tive component have also been considered (Reyes et al., 2013; Barbieri et al., 2014; Hern´andez Far´ıas 1For a more comprehensive overview of irony detection, see (Joshi et al., 2017). et al., 2016). Other methods have opted for taking advantage of information coming from the con- text in which a given utterance is produced (Ra- jadesingan et al., 2015). There are also some ap- proaches exploiting deep learning techniques and word embeddings (Poria et al., 2016; Ghosh and Veale, 2016; Joshi et al., 2016; Nozza et al., 2016). A less explored strategy for addressing irony de- tection is the use of ensemble methods. Fersini et al. (2015) and Liu et al. (2014) compared the performance of ensemble approaches against tra- ditional classifiers; the best results were obtained by the ensemble strategy setting. In this paper we describe our participation to the SemEval-2018 Task 3: Irony detection in En- glish tweets (Van Hee et al., 2018). The INAOE- UPV system explores the use of an ensemble ap- proach that considers different combinations of three methods. The main contribution of our ap- proach lies on the use of a list of potentially …

Content truncated to 500 words