The project focuses on developing a holistic framework to analyse public opinions and social media communications using natural language processing (NLP) and deep learning techniques. This will allow a better understanding of people’s opinions, reactions, and feedback to improve 1) government policy design (e.g., crisis management), 2) public service delivery (e.g., healthcare), and 3) two-way communication (e.g., trust). The primary objective of this research project is to provide integrated semantic and linguistic analysis of opinionated text on multiple interconnected levels, including social opinion dimensions (e.g., sentiment/emotions), language style (e.g., metaphor/irony), and communication type (e.g., misinformation/hate speech).

This will be achieved by developing computational approaches, informed by a theoretical framework, that fill the gaps of the current state-of-the-art by redefining and broadening the scope of social media “language” analysis. More importantly, this project will bring the developed approaches under a unified computational framework that allows end-users (e.g., policymakers, scientists, and government officials) to analyse opinionated text automatically.

The project comprises four work packages on theoretical framework, dataset preparation, computational models development and implementation, and evaluation and application. The outcomes of this project will comprise publications, datasets, models and an open-source computational framework.