Covid-19 and its variants have created a hitherto un-matched challenge to public health in Hong Kong where only about 60% of its citizens have been vaccinated voluntarily. There is a lot of incentive for HK health authorities to automatically understand the sentiment of the local community towards the government’s health recommendations according to information available in online forums. We propose a project that aims to produce an appraisal-based language modeling method to produce better results than conventional lexis-based models for the purpose of public sentiment assessments. More specifically, the project will construct a large collection of texts gathered from online forums and segment them according to different sentiment categories according to Appraisal Theory such that sophisticated big data technologies can be applied to identify classes of expressions and constructs that are characteristic of specific sentiment types. The project is expected to advance the previous research by the PC and the Co-PI paving way towards a collaborative grant application.