A team led by a researcher of Indian origin in Canada has developed an innovative system that significantly enhances the accuracy and reliability of online health-related searches, resulting in an 80 percent improvement. This advancement aims to assist individuals in making better-informed decisions regarding topics such as COVID-19.
The research team, based at the University of Waterloo in Canada, emphasized the widespread use of internet search engines by the public to gather information about COVID-19 and its impact on health. Given the proliferation of misinformation, the team set out to create a more dependable method for conducting these searches.
Ronak Pradeep, a PhD student at the Cheriton School of Computer Science at Waterloo and the lead author of the study, highlighted the challenge posed by the constant influx of new information. He noted that misinformation can have serious consequences, such as individuals buying ineffective medications or employing harmful home remedies.
The researchers pointed out that even major search engines, which handle billions of searches daily, struggle to keep up with the rapid accumulation of scientific data and research on COVID-19. Many existing systems are trained on well-vetted data and can’t always distinguish between legitimate health information and articles promoting dubious remedies.
Mr. Pradeep emphasized the project’s objective of refining internet search algorithms to promote reliable health information for users. The team developed a two-stage neural reranking architecture for search and incorporated a label prediction system trained to differentiate between accurate and questionable information. This system integrates with a search protocol that relies on data from the World Health Organization (WHO) and verified information as the basis for ranking, promoting, or occasionally excluding online articles.
The authors of the study explained that their design has the potential to enhance consumer health searches and combat misinformation, a challenge that has been exacerbated by the COVID-19 pandemic.
Mr. Pradeep and his co-authors, Xueguang Ma, Rodrigo Nogueira, and Jimmy Lin from the University of Waterloo, presented their preliminary findings at SIGIR ’21, a conference on research and development in information retrieval held online from July 11-15.