Detecting hazardous events from social media based on deep learning

• Main Author: Lim, Xu Kang
• Other Authors: Mao Kezhi
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/149531
• Institution: Nanyang Technological University

For this project, the main objective is to build a network that is capable of detecting posts made during a hazardous event and hence be able to detect hazardous events. Today, where social media has grown to such size and prevalence, a lot of valuable information can be drawn from this platform if one were to be able to harness it. Previous efforts have been made to make use of that data for not only event detection but also various other uses like sentiment analysis. Previously, many methods have been used to achieve those goals, machine learning methods like SVM (Support-Vector Machine), Naïve Bayes and Random Forests have been used with varied success. However, with many new developments in the field of NLP (Natural Language Processing), the objective of this project is to evaluate the results that can be achieved with these new SoTA (State-of-the-Art) models. For this project, a dataset from Twitter will be scraped, filtered, cleaned, and fed into a baseline model to ascertain a baseline performance. It will then be fed into various other more advanced models to compare the differences in performance. Although, there is more work to be done for a fully functioning application for hazard detection, results for the models were very encouraging.