User-Chatbot Conversations During the COVID-19 Pandemic: Study Based on Topic Modeling and Sentiment Analysis

Area of Science:

• Public Health
• Human-Computer Interaction
• Computational Social Science

Background:

• Chatbots show promise for public health initiatives, but user interactions during the COVID-19 pandemic remain under-researched.
• Understanding these interactions is vital for developing effective health emergency support services.

Purpose of the Study:

• To investigate COVID-19 pandemic-related topics discussed by online users with a social chatbot.
• To compare user sentiment across five culturally diverse countries regarding these discussions.

Main Methods:

• Analysis of 19,782 COVID-19 related conversation utterances from SimSimi (a social chatbot) between 2020-2021.
• Natural language processing for topic identification and sentiment analysis across the United States, United Kingdom, Canada, Malaysia, and the Philippines.

Main Results:

• Identified 18 topics across five themes: questions about COVID-19, preventive behaviors, outbreak details, health impacts, and pandemic life.
• Users treated chatbots as information sources for health queries and for emotional connection during lockdowns.
• Negative sentiment was associated with masks, lockdowns, and pandemic worries; positive sentiment dominated small talk. US users expressed more negative sentiment than Asian users.

Conclusions:

• User-chatbot interactions reveal informational and emotional needs during health crises.
• Chatbots have potential for delivering accurate health information and emotional support.
• Future research should explore support strategies aligned with public health policies.