Hybrid approach for emotion detection in text using spatial-temporal features

Nida Hakak; Mahira Kirmani

doi:10.37591/joarb.v5i1.1542

Abstract

In recent years, the incredible exploratory growth of online data in social media mainly Twitter has to lead to growing attention of researchers towards Affective analysis of social media streams. Around 500 million tweets are generated per day all around the globe. Classification of these tweets into different affective classes is an arduous task. We propose a novel approach of classifying a tweet into Ekman’s six basic emotion classes using neural network architectures (convolutional and recurrent neural networks). We collected data for specific events from Twitter and evaluated our classifier. Results obtained lead us to the conclusion that neural network architectures outperform the supervised learning classifiers. We achieved 80% accuracy for our proposed classifier which is higher than the baseline classifiers.Keywords: Artificial intelligence, emotion analysis, neural networks, emotion corpora, sentiment analysisCite this ArticleNida Hakak, Mahira Kirmani. Hybrid Approach for Emotion Detection in Text using Spatial-Temporal Features. Journal of Advancements in Robotics. 2018; 5(1): 34–44p.