Sentiment Analysis of Social Media Data on Ebola Outbreak Using Deep Learning Classifiers

Area of Science:

• Public Health
• Computational Linguistics
• Epidemiology

Background:

• The Ebola virus disease (EVD) outbreak in Uganda highlighted the need to understand public sentiment on social media.
• Effective health interventions require comprehending online discussions and sentiments regarding EVD.

Purpose of the Study:

• To analyze public sentiment towards Ebola virus disease (EVD) using deep learning on social media data.
• To compare the performance of Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Bidirectional Encoder Representations from Transformers (BERT) models for EVD sentiment classification.

Main Methods:

• Utilized deep learning techniques to classify sentiments (positive, neutral, negative) in 8395 Ebola-related tweets.
• Evaluated a 6-layer CNN, a 6-layer LSTM, and an 8-layer BERT model for sentiment analysis.
• Assessed model performance using various evaluation metrics.

Main Results:

• The BERT model demonstrated superior performance compared to CNN and LSTM models.
• Achieved a classification accuracy of 95% using the BERT model for EVD sentiment analysis.
• Findings underscore the effectiveness of Transformer-based architectures in natural language processing tasks.

Conclusions:

• Deep learning, particularly BERT, is highly effective for analyzing public sentiment during disease outbreaks.
• Understanding social media sentiment can aid in targeted public health communication and intervention strategies for EVD.
• Transformer models offer a robust approach for sentiment analysis in public health contexts.