Applying Basic Features from Sentiment Analysis for Automatic Irony Detection

Abstract

Hernández Farías, I.; Benedí Ruiz, JM.; Rosso, P. (2015). From the sentiment analysis perspective such utterances represent a challenge being a polarity reversor (usually from positive to negative). This paper presents an approach to address irony detection from a machine learning perspective. Our model considers structural features as well as, for the first time, sentiment analysis features such as the overall sentiment of a tweet and a score of its polarity. The approach has been evaluated over a set classifiers such as: Naïve Bayes, Decision Tree, Maximum Entropy, Support Vector Machine, and for the first time in irony detection task: Multilayer Perceptron. The results obtained showed the ability of our model to distinguish between potentially ironic and non-ironic sentences.

Irony Detection in Twitter

“…Experimental results show that emotIDM outperforms the irony detection models presented in [Riloff et al 2013;Reyes et al 2013;Barbieri et al 2014;Hernández Farías et al 2015] over the same datasets.…”

“…There is now a consistent body of work on computational models for irony and sarcasm detection in social media [González-Ibáñez et al 2011;Reyes et al 2013;Wang 2013;Riloff et al 2013;Barbieri et al 2014;Ptáček et al 2014;Hernández Farías et al 2015], and in particular in Twitter, which can be considered the most widely used source of information to experiment with irony detection. In this article we also address the task of detecting irony in tweets, by identifying a set of discriminative features to automatically differentiate an ironic text from a non-ironic one.…”

“…In line with most of the current approaches and with some theoretical accounts Whalen et al 2013], irony is here considered an umbrella term that covers also sarcasm, being the issue of discriminating between the two devices a further challenge for figurative language processing. Our irony detection model, called emotIDM, extends the model proposed in [Hernández Farías et al 2015] with new features, in particular experimenting the use of a wide range of psycholinguistic and affective features concerning affective information, with the main aim to answer to our main research questions: (1) Does information about different facets of affect help in distinguishing among ironic and non-ironic tweets? (2) Which facets of affect seem to be more important in order to address our classification task?…”

“…Summarizing, the main contributions of this paper are the following: a) We propose a new approach to irony detection emotIDM based on [Hernández Farías et al 2015] that exploits affective information as features to represent ironic tweets; b) We evaluate emotIDM carrying out a battery of binary classification experiments over a set of Twitter corpora, developed in different way both for what concerns the selection criteria for samples of irony/sarcasm and the annotation methodology. This is important in order to validate the robustness of the model and to better compare results with state-of-the-art; c) We demonstrate that affective information helps in distinguishing among ironic and non-ironic tweets, presenting a comparative evaluation of the performances over the various corpora, and a feature analysis in order to identify the most useful features in emotIDM.…”

Sentiment Detection, Recognition and Aspect Identification

“…Hernández-Farías, et al [11] presented an irony detection model capable of detecting irony sentiments through ironic and non-ironic sentences. It considered a set of structural features that associate both text characteristics and knowledge.…”