ValenTo: Sentiment Analysis of Figurative Language Tweets with Irony and Sarcasm

Proceedings of the 9th International Workshop on Semantic Evaluation (SemEval 2015), pages 694–698, Denver, Colorado, June 4-5, 2015. c⃝2015 Association for Computational Linguistics ValenTo: Sentiment Analysis of Figurative Language Tweets with Irony and Sarcasm∗ Delia Iraz´u Hern´andez Far´ıas Universitat Polit`ecnica de Val`encia Pattern Recognition and Human Language Technology dhernandez1@dsic.upv.es Emilio Sulis, Viviana Patti, Giancarlo Ruffo, Cristina Bosco University of Turin Dipartimento di Informatica {sulis,patti,ruffo,bosco}@di.unito.it Abstract This paper describes the system used by the ValenTo team in the Task 11, Sentiment Anal- ysis of Figurative Language in Twitter, at Se- mEval 2015. Our system used a regression model and additional external resources to as- sign polarity values. A distinctive feature of our approach is that we used not only word- sentiment lexicons providing polarity anno- tations, but also novel resources for dealing with emotions and psycholinguistic informa- tion. These are important aspects to tackle in figurative language such as irony and sar- casm, which were represented in the dataset. The system also exploited novel and stan- dard structural features of tweets. Considering the different kinds of figurative language in the dataset our submission obtained good re- sults in recognizing sentiment polarity in both ironic and sarcastic tweets. 1 Introduction Figurative language, which is extensively exploited in social media texts, is very challenging for both traditional NLP techniques and sentiment analysis, which has been defined as “the computational study of opinions, sentiments and emotions expressed in text” (Liu, 2010). There is a considerable amount of works related to sentiment analysis and opinion mining (Pang and Lee, 2008; Liu, 2010; Cambria et al., 2013). In particular, the linguistic analysis ∗ The National Council for Science and Technology (CONACyT-Mexico) has funded the research work of the first author (218109/313683 grant). of social media (microblogging like Twitter espe- cially) has become a relevant topic of research in different languages (Rosenthal et al., 2014; Basile et al., 2014) and several frameworks for detecting sentiments and opinions in social media have been developed for different application purposes. In a sentiment analysis setting, the presence in a text of figurative language devices, such as for in- stance irony, can work as an unexpected polarity re- verser, by undermining the accuracy of the systems (Bosco et al., 2013). Therefore, several efforts have been recently devoted to detect and tackle figura- tive language phenomena in social media, follow- ing a variety of computational approaches, mostly focussing on irony detection and sarcasm recogni- tion (Davidov et al., 2010; Gonz´alez-Ib´a˜nez et al., 2011; Riloff et al., 2013) as classification tasks. Buschmeier et al. present an analysis of fea- tures, previously applied in irony detection, in a dataset from a product reviews corpus from Amazon (Buschmeier et al., 2014). Veale and Hao present a linguistic approach to separate ironic from non- ironic expressions in figurative comparisons over a corpus of web-harvested similes (Veale and Hao, 2010). Concerning Twitter, the problem of irony de- tection is addressed in (Reyes et al., 2013), where a set of textual features is used to recognize irony at …

Content truncated to 500 words