RUSSIA-UKRAINE CONFLICT TWEETS SENTIMENT ANALYSIS USING BI-DIRECTIONAL LSTM NETWORK FOR POST-TRAUMATIC STRESS DISORDER EARLY DETECTION
Eman Sedqy Shlkamy1 Khaled Maher 2 and Ahmed Ahmed Hesham Sedky 3
1 Department of Information Systems, College of Computing and Information Technology, Arab Academy for Science Technology and Maritime Transport, Alexandria, Egypt
Abstract. Sentiment analysis techniques have a vital role in analyzing people’s opinions. The continuous and rapid growth of data posted on social media sites drives people’s opinions. However, most of the research focuses on analyzing sentiment to determine how the war will affect the global economy. As a consequence, in international conflict research, national leaders and other powerful figures are typically given more attention than public opinions and emotions. This paper aims to go over sentiment analysis, and focus on analyzing public emotions and opinions during the Russia-Ukraine Conflict to early detect Post-Traumatic Stress Disorder (PTSD) symptoms. This is the first study to provide a model that represents the intention to analyze how the Russian-Ukraine military conflict affected mental health. This can prevent people against mental illnesses and suicide, as well as point the way forward for future study in this field. The method utilized in this study is a single bidirectional LSTM network for English tweet sentiment analysis, with positive, negative, and neutral classifications as a multi-class classification strategy aimed at detecting PTSD indicators. Natural language processing (NLP) is used to extract emotional content from text data via sentiment analysis. By developing a Deep Learning (ML) model using text data, we hope to identify individuals with PTSD via sentiment analysis. Text sentiment analysis was used to learn Natural language processing (NLP) is used to extract emotional content from text data via sentiment analysis. By developing a Deep Learning (ML) model using text data, we hope to identify individuals with PTSD via sentiment analysis. We used one bidirectional long short-term memory (Bi-LSTM) layer in combination with the global Max pooling ID approach and achieved an accuracy of 91.64% based on sentiment analysis of text. In terms of accuracy, the results of the proposed framework outperform earlier state-of-the-art investigations. Creating an early detection model can aid in reducing the symptoms of post-traumatic stress disorder (PTSD).
Keywords: Sentiment Analysis · Neural network · Bidirectional LSTM · Deep neural network· Natural language processing (NLP)· Post-Traumatic Stress Disorder (PTSD)