The COVID-19 pandemic highlighted some of the limitations of traditional health surveillance systems. This 2-year research fellow post will focus on the development of non-traditional methods for disease modelling, driven by web search data and machine learning methods. It is funded by Google and is supported by a public release of Google search data. We are also active partners of other COVID-19-related interdisciplinary projects (VirusWatch, i-sense) which facilitate collaboration with and feedback from public health experts. Outcomes will be shared with established health agencies.

The purpose of this post is to investigate the spread of COVID-19 using web search activity and other related data sources (e.g. social media, news, traditional health surveillance models and metrics). The post-holder will be required to work as part of a research team, collaborate with computer science and public health experts, and will be responsible for realising the following tasks:

• Develop models for national and subnational COVID-19 prevalence using web searches including forecasting techniques that will incorporate mechanistic approaches.
• Develop subnational COVID-19 anomaly detection models to identify hotspots with diverging search behaviour compared to control locations.
• Use web searches to understand how COVID-19 might have impacted valnurable populations.

The research carried out may be both foundational and applied. The main areas of research will be machine learning and natural language processing. The results of this research are to be disseminated through peer-reviewed publications at leading scientific conferences or journals, oral presentations in seminars and project meetings, and written progress reports. We will also anticipate to open source software libraries that have been developed as part of the research.