Resumo

SOTOLONGO-PENA, Anakarla; ARCO, Leticia  e  BELLO, Rafael. Feature Selection and Ranking to Characterize Ironic Texts. Rev cuba cienc informat [online]. 2020, vol.14, n.4, pp. 67-84.  Epub 01-Dez-2020. ISSN 2227-1899.

Textual opinions impose great challenges to opinion mining applications since several problems are present; among them: writing opinions ironically or sarcastically. One of the trends that exist to detect irony is the classification based on features. In previous research a set of features that allow detecting irony in textual opinions is proposed; however, the calculation of these features is computationally costly. In this paper, we propose to study this set of features to detect a subset of it that discriminates between ironic and non-ironic short texts, without affecting the effectiveness of the classifiers. The main result of this work consists of obtaining a subset of features that can effectively detect irony, through the application of selection and feature ranking techniques, and the evaluation of several supervised learning techniques. The set obtained from seven features is enough to discriminate between ironic and non-ironic opinions, obtaining statistically comparable results with those obtained by using a larger and more complex set of features.

Palavras-chave : irony detection; opinion mining; feature selection; ranking of features; classification.

        · resumo em Espanhol     · texto em Espanhol     · Espanhol ( pdf )