Comparison of NLP Techniques for Sentiment Analysis on Social Data (Application Case: War in GAZA)

Abstract

The field of sentiment analysis (SA) has experienced a significant resurgence with the advancement of artificial intelligence (AI) techniques in recent years. Its use has been widely associated with the analysis of public opinions. Given the global attention on the war in Gaza, this study aims to apply and compare various natural language processing (NLP) techniques for sentiment analysis on public comments from social media related to this conflict. Different classification approaches were employed, including traditional machine learning, deep learning, and transfer learning. The results indicated that the majority of comments expressed negative sentiments towards the war. Notably, the DistilBERT classifier achieved the highest classification accuracy at 89%, slightly outperforming the LSTM model, which achieved an accuracy of 88%. The findings of this study will serve to inform and stimulate future research in this evolving field.