Public Sentiments in User Generated Content amid COVID-19 Pandemic in Ghana

ABSTRACT

Towards post COVID-19 pandemic, a natural language processing (NLP) technique was leveraged to understand the sentiments of Ghanaians through their public discourse in tweets during the lockdown period in Ghana. With NLP resources, feature words were extracted from the tweets and fed into three machine learning algorithms to track public sentiments in the tweets. The algorithms, support vector machines (SVM), naïve-bayes (NB) and artificial neural network (ANN) were evaluated to ascertain their efficacies. Frequently occurring words used by Ghanaians during the lockdown period were extracted to provide more insight into public sentiments. The study revealed that negative sentiments prevailed throughout the COVID-19 lockdown among Ghanaians. However, positive sentiments were surprisingly high at some points during the lockdown period. The result of evaluating the machine learning classifier yielded SVM as the best performing classifier though the other classifiers performed beyond the acceptable threshold. With these findings, it is envisioned that this study will be adopted by policymakers, as a guide, towards public management of public sentiments in pandemics. © 2022 IST-Africa Institute and Authors.