An Optimized Arabic Sarcasm Detection in Tweets using Artificial Neural Networks

Abstract

This paper presents an optimized Arabic sarcasm classification model using artificial neural networks in conjunction with particle swarm optimization. Artificial Neural Networks (ANNs) are used to learn the extracted feature representation of a given text. Term frequency with inverse document frequency (TFIDF) is adapted for feature extraction and text transformation into numerical values. Particle Swarm Optimization (PSO) selects the most relevant features to optimize classification performance. Experiments show that the classification accuracy is optimized after using PSO from 82.12% to 86.85%.