en
Conference presentation
Open access
English

Exploring attitudes and perceptions regarding obesity on Twitter during the COVID-19 pandemic : A sentiment analysis and topic modeling approach

Presented atECO 2023, 30th European Congress on Obesity, Dublin, 17-10 May 2023
Publication date2023-05-12
Presentation date2023-05-12
Abstract

Introduction

The prevalence of obesity is increasing worldwide among people belonging to all age groups. It is associated with a high risk of medical, psychiatric and psychosocial morbidities. It is important to explore attitudes and perceptions regarding obesity to formulate effective health policies, prevention strategies, and treatment approaches. This study aims to explore the sentiments of the public, influential people and important organizations regarding obesity by leveraging the wealth of information available on social media platforms.

Methods

This study utilizes a huge dataset of tweets regarding obesity posted on the Twitter platform during the COVID-19 period (2019 to 2021). This dataset comprised a corpus of 25,580 tweets. All analyses were run in Python. Sentiment analysis was run using the XLM-Roberta-base model trained on approximately 198M multilingual tweets, accessed at the hugging face platform for transformer language models. Topic modeling was done using the BERTopic library which leverages cluster analytical approaches, transformers, and Term Frequency-Inverse Document Frequency (c-TF-IDF) to create dense clusters of interpretable topics.

Results

The Twitter API was used to extract 25,580 tweets regarding obesity. The number of tweets increased yearly from 11,841 in 2020 to 12,775 in 2021. Sentiment analysis revealed a significantly higher percentage of tweets (72.97%) represented negative sentiments, followed by neutral (18.78%) and positive (8.25%). Spikes in Twitter activity were associated with significant political events such as the Speaker of the United States House of Representatives negative remarks on President Trump’s struggle with obesity (May 19, 2020). Another important event included Boris Johnson’s obesity campaign which drew a lot of criticism from the public (July 27, 2020). On 23rd August 2021, Ben Shapiro’s comments on the refusal to vaccinate people with obesity for COVID-19 sparked outrage.

Topic modeling revealed 243 parsimonious clusters. These represented different topics of tweets on obesity. Important topics included childhood obesity, President Trump’s struggle with obesity, vaccination for COVID-19, Boris Johnson’s obesity campaign, glorifying and body shaming, racism and high obesity rates among Black Americans, smoking, illicit substance use, and alcohol consumption among people with obesity, environmental risk factors for obesity and polycystic ovary syndrome and surgical treatments.

Conclusion

Twitter is an important source to gauge obesity-related sentiments and attitudes of the public, celebrities, and influential organizations. Sentiments regarding obesity were predominantly negative. Obesity was associated with racism, poorer life choices, and social evils such as illicit substance use and alcohol consumption. Influential politicians’ negative portrayal of obesity among their colleagues may lead to poorer public sentiments. This has negative connotations for public health in general. The Conservative government’s campaign for curbing the “epidemic” of obesity in Britain attracted criticism from the public. Obesity being a risk factor for severe COVID-19 also led to a negative portrayal on social media.

eng
NotePublished in : Obesity Facts, 2023, 16(Suppl. 1):178
Citation (ISO format)
CORREIA, Jorge, SARMAD SHAHARYAR, Ahmad. Exploring attitudes and perceptions regarding obesity on Twitter during the COVID-19 pandemic : A sentiment analysis and topic modeling approach. In: ECO 2023. Dublin. 2023. doi: 10.1159/000530456
Main files (1)
Presentation
Identifiers
27views
27downloads

Technical informations

Creation02/01/2024 5:03:51 PM
First validation04/22/2024 8:48:02 AM
Update time04/22/2024 8:48:02 AM
Status update04/22/2024 8:48:02 AM
Last indexation05/06/2024 6:25:57 PM
All rights reserved by Archive ouverte UNIGE and the University of GenevaunigeBlack