CISUC at IDPT2021: Traditional and Deep Learning for Irony Detection in Portuguese Hugo Gon¸calo Oliveira [0000-0002-5779-8645] , Jos´ e Pereira [0000-0001-8649-9006] , and Guilherme Cruz CISUC, Department of Informatics Engineering, University of Coimbra, Coimbra, Portugal hroliv@dei.uc.pt, {jose,gjcruz}@student.dei.uc.pt Abstract. These notes describe the participation of the CISUC team in the IDPT 2021 shared task. Irony detection was tackled as a text clas- sification task, where both traditional and transformer-based (BERT) approaches were explored. The former performed ok, but not everything went well, and the results achieved by BERT were not evaluated, due to an issue with our official submissions. Nevertheless, we still discuss some of the options taken, identify important features, and present validation results in the training data. Keywords: Irony Detection · Portuguese · Text Classification · Trans- formers · Logistic Regression 1 Introduction Irony is a rhetorical device where interpretation should not be literal [18], because its meaning diverges significantly from, and is often the opposite [7], of the intended meaning. Irony detection is a subtask of Natural Language Processing aiming at the automatic classification of texts as ironic or not, and is extremely relevant for tasks like Sentiment Analysis and Opinion Mining [14]. But irony detection can be challenging, even for humans, who often rely on visual clues, like facial expression or tone [7], for recognising irony. This is especially true when irony is expressed through text only, despite studies on identifying textual clues for irony detection [1]. Irony detection has been tackled by several Natural Language Process- ing (NLP) researchers, who adopted different approaches. In 2018, there was a SemEval task on Irony Detection in English Tweets [18] that covered the binary classification of tweets as ironic or not. Best systems adopted a deep learning approach, e.g., a densely LSTM neural network, based on pre-trained static word embeddings, with syntactic and sentiment features [19]. But there were also more traditional approaches, e.g., an ensemble classifier with Logistic IberLEF 2021, September 2021, M´alaga, Spain. Copyright © 2021 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).