Case Studies

The dashboard evaluation is ongoing using three priority COVID-19 Case Studies:

Scottish Government Dialogue Public Consultation

The case study uses data from the Scottish Government Dialogue platform to analyse public sentiments in relation to the government’s COVID-19 response. The online platform was open between 5 May to 11 May 2020 and received more than 4,000 ideas and 18,000 comments. The dataset was used to develop, train and validate a novel AI-based Topic and Theme Classification Model and, further, to develop and validate automated sentiment classification of public comments. The topic modelling results are used to generate insights for policymakers and to identify new or emerging themes in relation to the COVID-19 pandemic. These insights will help to inform and support future public engagement activities.

Link to Scottish Government Dialogue Platform.

Impact of COVID-19 on BAME communities in Scotland

This preliminary case study aims to analyse the overall sentiments on Twitter in relation to the “Beyond the data – Understanding the impact of COVID-19 on BAME groups” report published by Public Health England in June 2020. More than 5000 tweets from the UK were collected using the “#bamereport” keyword from 1 June until 10 June 2020. These were analysed using a novel deep learning (DL) model trained on a large publicly available language corpus. In addition, approximately 2000 tweets were collected using the “#BAME” keyword over a 5-month period (February-June 2020) and further analysed to determine variations and trends in public sentiment over time. The model has been optimised using a COVID-19 specific lexicon and the analysis widened to include other ethnicity and COVID-19 related keywords.

Link to Public Health England: Beyond the data – Understanding the impact of COVID-19 on BAME groups report.

Public attitudes towards COVID-19 vaccination in the UK and the US

This case study aims to analyse public sentiments towards COVID-19 vaccination in the UK and the US, using data from two popular social media platforms, Twitter and Facebook. Publicly available data from both platforms were extracted for the period of 1 March to 15 October 2020 using relevant keywords identified by domain experts. Posts were filtered geographically for the UK and the US, resulting in approximately 30,000 and 80,000 tweets, and 20,000 and 90,000 Facebook posts respectively. The sentiments of the tweets and posts were analysed over time using a hybrid ensemble deep learning-based approach, and a random ~10% sample manually annotated to validate the AI model. Geographical and temporal sentiment graphs, sentiment word-clouds and statistical analysis measures were used to identify statistically significant trends and potential determinants of positive and negative public sentiment and opinion.

The results indicate that the public opinion on social media, both in the UK and the US, was mainly optimistic about vaccine development and its future prospects over the studied time period. However, common concerns around vaccine-related issues, such as safety and availability, were also evident.

Publications

[1] Hussain Z, Sheikh Z, Tahir A, Dashtipour K, Gogate M, Sheikh A, Hussain A. Artificial intelligence-enabled social media analysis for pharmacovigilance of COVID-19 vaccinations in the United Kingdom: Observational Study. JMIR Public Health Surveill. 2022 Feb 8. doi: 10.2196/32543. Epub ahead of print. PMID: 35144240.

[2] Hussain A, Sheikh A (2021) Opportunities for AI-enabled social media analysis of public attitudes towards Covid-19 vaccines, NEJM Catalyst Innovations in Care Delivery, DOI:10.1056/CAT.20.0649

[3] Hussain A, Tahir A, Hussain Z, Sheikh Z, Gogate M, Dashtipour K, Ali A, Sheikh A, (2021), Artificial intelligence-enabled analysis of UK and US public attitudes on Facebook and Twitter towards COVID-19 vaccinations, Journal of Medical Internet Research, DOI: 10.2196/26627

[4] Cresswell K, Tahir A, Sheikh Z, Hussain Z, Hernández A D, Harrison E, Sheikh A, Hussain A, (2021), Artificial Intelligence-enabled analysis of social media data to understand public perceptions of COVID-19 contact tracing apps, Journal of Medical Internet Research, DOI: 10.2196/26618

Project Datasets:

We have published our collected datasets, which contain over 13 million Twitter messages and 5.6 million Facebook posts along with annotated results of our AI-based categorisation process. The datasets are publicly available via the ENU repository: https://www.napier.ac.uk/research-and-innovation/research-search/outputs/covid-19-uk-social-media-dataset-for-public-health-research