Social media sites are now used by more than two-thirds of the global internet users. In Singapore, 7 out of 10 Singaporeans use social media platforms on mobile to connect with one another, consume news content, and share information. With the vast amount of information exchanged over these platforms, they have become the primary drivers for the spread of false information.

False information, such as news or stories, are information that may be exploited by entities with ill intent such as misleading individuals for personal gain. The World Economic Forum lists mis- and disinformation as one of the biggest threats to society as it has the potential to unduly influence attitudes and behaviour, leading people to think and act differently than they would if they were correctly informed. As observed in the 2016 United States Presidential Election and voluntary vaccination programmes and mask mandates for COVID-19 pandemic, rumours and mis- or disinformation could cause confusion and alarm to the public, and eroding the trust society’s fundamental institutions.

To combat the threat of mis- or disinformation, a multi-pronged approach — integrating social and behavioural science research, digital technologies, data-driven methods and policy studies — is needed to holistically examine the Internet and its societal impact.

To this end, NUS-CTIC, in collaboration with Prof Wynne Hsu from Institute of Data Science, has developed a three-stage automated fact-checking framework using evidence retrieved from a dense passage representation. Led by Prof Lee Mong Li, an expert in data management and analysis of semi-structured spatio-temporal data, the team has designed a model to retrieve relevant passages containing semantically close words from independent and authoritative sources. Users can use these passages to discern information that they read online. This approach also facilitates the verification of claims that are time sensitive since the evidence is gathered in real-time.

In the recent COVID-19 pandemic, NUS-CTIC has deployed this evidence retrieval technology on the public LetsCheck platform for users to verify claims related to COVID-19. Users can explore circulating claims in Twitter and use the semantic search engine to retrieve relevant evidence from news articles and scientific publications, such as the COVID Open Research Dataset, to assess the truthfulness of a claim.