Deep learning based topic and sentiment analysis: COVID19 information seeking on social media

Area of Science:

• Public Health
• Computational Social Science
• Epidemiology

Background:

• Social media platforms facilitate widespread information exchange and reveal user emotions through text.
• Analyzing social media data over time and space provides critical insights into public health issues, including disease outbreaks and mental health trends.
• Timely insights from social media analysis can inform the development of effective public health strategies and resource allocation.

Purpose of the Study:

• To analyze a large Spatio-temporal tweet dataset related to COVID-19 in Australia.
• To gain insights into the COVID-19 pandemic outbreak and public discussion across different Australian states and cities over time.
• To compare social media-derived insights with independently observed phenomena, such as government-reported case data.

Main Methods:

• Spatio-temporal analysis of a large tweet dataset.
• Volume analysis to track discussion frequency.
• Topic modeling to identify key themes in public discourse.
• Sentiment detection to gauge public emotional responses.
• Semantic brand score to assess the nature of discussions.

Main Results:

• Identification of distinct patterns in COVID-19 related discussions across Australian states and cities.
• Correlation between social media sentiment and public health events.
• Temporal trends in public engagement with pandemic-related information.
• Insights into public perception and concerns regarding COVID-19.

Conclusions:

• Social media data offers a valuable resource for understanding public responses to health crises like COVID-19.
• Spatio-temporal analysis of tweets can provide real-time insights into pandemic dynamics.
• Integrating social media data with official reports enhances situational awareness and response strategies.