Browsing by Author "Jamil, Muhammad Luqman"
Now showing 1 - 1 of 1
Results Per Page
Sort Options
- Event Detection and Tracking Detection of Dangerous Events on Social MediaPublication . Jamil, Muhammad Luqman; Pais, Sebastião Augusto Rodrigues Figueiredo; Cordeiro, João Paulo da CostaOnline social media platforms have become essential tools for communication and information exchange in our lives. It is used for connecting with people and sharing information. This phenomenon has been intensively studied in the past decade to investigate users’ sentiments for different scenarios and purposes. As the technology advanced and popularity increased, it led to the use of different terms referring to similar topics which often result in confusion. We study such trends and intend to propose a uniform solution that deals with the subject clearly. We gather all these ambiguous terms under the umbrella of the most recent and popular terms to reach a concise verdict. Many events have been addressed in recent works that cover only specific types and domains of events. For the sake of keeping things simple and practical, the events that are extreme, negative, and dangerous are grouped under the name Dangerous Events (DE). These dangerous events are further divided into three main categories of action-based, scenario-based, and sentiments-based dangerous events to specify their characteristics. We then propose deep-learning-based models to detect events that are dangerous in nature. The deep-learning models that include BERT, RoBERTa, and XLNet provide valuable results that can effectively help solve the issue of detecting dangerous events using various dimensions. Even though the models perform well, the main constraint of fewer available event datasets and lower quality of certain events data affects the performance of these models can be tackled by handling the issue accordingly.