Emotion Recognition and Sarcasm Mining using Rule-based and Deep Neural Networks - Archive ouverte HAL Accéder directement au contenu
Poster De Conférence Année : 2021

Emotion Recognition and Sarcasm Mining using Rule-based and Deep Neural Networks

Résumé

Since the health crisis linked to covid, people spend more time on the internet. In other words, social networks have become a space to make themselves heard and also to exchange, to express themselves on current events. Speakers use many linguistic and language processes, whether explicit or implicit, such as feelings, emotions, irony and sarcasm, known for their argumentative and even persuasive effectiveness. Therefore, the focus of this research is two-fold. The first objective is to analyze the reaction of citizens on different subjects related to confinement, covid treatment and vaccination. More specifically, we collected a French tweet corpus from March 2020 based on the trending hashtags #confinement, #chloroquine, #vaccination, #stopcovid. The second objective is to compare the performance of two approaches (rule-based and deep neural networks) for emotion recognition and sarcasm mining. Only six basic emotions are applied to this study and are classified into two classes: positive emotions (happiness, surprise), negative emotions (anger, disgust, fear, sadness). For the rule-based approach, we manually defined a set of rules to extract emotion patterns, identify the sentiment polarity as well as sarcasm detection. Besides, due to the performance of deep learning on a variety of NLP tasks, we developed different models such as CNN, LSTM + Attention and Bert model. The usefulness of these models can be explained by the fact that their representations robustly encode lexical and syntactic information. However, much work remains to be done to understand the extent to which we could adapt existing models to our complex problems. Thus, we used diverse techniques like: deconvolution, attention mechanism, and integrated gradient to understand the decision of models. 
Fichier principal
Vignette du fichier
Emotion Recognition and Sarcasm Mining using Rule-based and Deep Neural Networks.pdf (1.1 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04368363 , version 1 (31-12-2023)

Identifiants

  • HAL Id : hal-04368363 , version 1

Citer

Abdelouafi El Otmani, Trang Lam, Julien Longhi. Emotion Recognition and Sarcasm Mining using Rule-based and Deep Neural Networks. CMC2021: 8th Conference on Computer-Mediated Communication (CMC) and Social Media Corpora, Oct 2021, Nijmegen, Netherlands. ⟨hal-04368363⟩
4 Consultations
4 Téléchargements

Partager

Gmail Facebook X LinkedIn More