Using Natural Language Processing to Explore Social Media Opinions on Food Security: Sentiment Analysis and Topic...

Area of Science:

• Public Health Nutrition
• Data Science
• Computational Social Science

Background:

• Social media offers valuable large-scale data for public health insights.
• Natural language processing (NLP) tools like sentiment analysis and topic modeling are underutilized in food security research.
• Understanding public discourse on food security is crucial for public health nutrition.

Purpose of the Study:

• To explore the utility of NLP tools for analyzing social media data on food security.
• To identify public sentiment and key topics related to food security discussed on social media.
• To assess the evolution of food security conversations over time.

Main Methods:

• Collected 38,070 tweets related to food security from Australian users (2019-2021).
• Applied sentiment analysis (Valence Aware Dictionary and Sentiment Reasoner) and topic modeling (Latent Dirichlet Allocation).
• Compared sentiment, topics, and engagement (likes, retweets, replies) across years and in relation to events like COVID-19.

Main Results:

• Overall positive sentiment towards food security, with variations across years and higher positivity during COVID-19 lockdowns.
• Identified 10 key topics, including "Food insecurity and health," "Use of food banks," and "Australian food insecurity."
• Negative tweets received higher engagement; "Giving to food banks" had the most positive sentiment, while "Global food insecurity" had the most negative.

Conclusions:

• Sentiment analysis and topic modeling effectively track evolving conversations on food security via social media.
• Future applications require public health expert interpretation and broader datasets for evidence-based decision-making.
• NLP offers a powerful tool for monitoring public health nutrition issues like food security.