Artificial Intelligence-based Natural Language Processing for sarcasm detection and classification on Arabic Corpus

Abstract

Sarcasm is a type of communication designed to harass or mock an individual using words against their accurate meaning. It signifies a negative sentiment but a positive sentiment. Sarcasm detection is challenging due to the gap between its intended and literal meaning and how sarcasm is expressed, specifically in Arabic, which has a complex and rich linguistic structure. Effective sarcasm detection is significant for Sentiment Analysis (SA) and can significantly improve the performance of various Natural Language Processing (NLP) applications. This study presents an Artificial Intelligence-based Natural Language Processing Driven Applied Linguistic using Bidirectional Temporal Convolutional Networks (AINLP-ALBTCN) technique on Sarcasm Detection in Arabic Corpus. The AINLP-ALBTCN technique concentrates on classifying and detecting sarcasm in Arabic Corpus. Initially, the AINLP-ALBTCN approach applies a series of data pre-processing steps to convert the input data into synchronized formats. Then, the word2vec embedding model is used to generate feature vectors. Furthermore, the BTCN technique is employed as a classification method to detect and classify sarcasm. Finally, the Sand Cat Swarm Optimization (SCSO) approach is chosen for hyperparameter optimization of the BTCN method, enhancing the performance of sarcasm detection. A wide range of experiments are conducted to demonstrate the promising outcomes of the AINLP-ALBTCN technique under the ArSarcasm dataset. The experimental validation of the AINLP-ALBTCN technique portrayed a superior accuracy value of 95.59 % over existing models in the sarcasm classification process.