Predicting Disasters with NLP

A text classification model that uses twitter data to determine if a tweet is about a real disaster or not. The dataset can be found here.

NOTE: This notebook was originally created via google colab and viewing it via this link might be better in terms of readibility.

Introduction

Twitter has become an important communication channel in times of emergency. With smartphones, people are able to announce an emergency they’re observing in real-time. Because of this, more agencies are interested in programatically monitoring Twitter (i.e. disaster relief organizations and news agencies).

Methodology

In this work, I utilize some of the state-of-the-art transformer models, specifically the BERT model and the Universal Sentence Encoder model. After fine-tuning of the models, I seek to gain more understanding of the models by using model interpretation techniques to understand these black-box models.