Mining scienti c data and debunking myths n Electronic and Computer Engineering team of postgraduates and so ware engineers, led by Prof. Pascale FUNG, Director of HKUST’s Center for Arti cial Intelligence Research (CAiRE), has successfully responded to a recent call to the world’s AI experts to help the medical research community during the pandemic. Competing against more than , teams globally, the School of Engineering team led by PhD student SU Dan (see P ) won one of the tasks set in the recent Kaggle COVID- Open Research Dataset Challenge (CORD- Challenge), by building a leading machine learning-based system with top natural language processing question-answering techniques, combined with summarization, for mining scienti c literature on COVID- . They have also successfully used their system for debunking virus myths, based on scienti c evidence. Kaggle is the world’s largest data science community and a subsidiary of Google. With an end-to-end neural network-based open-domain question-answering system, the CAiRE-COVID system can quickly generate ranked lists and paragraph-level summaries from CORD- ’s tens of thousands of scholarly articles. The dataset was created by the Allen Institute for AI in partnership with the Chan Zuckerberg Initiative, Georgetown University’s Center for Security and Emerging Technology, Microso Research, and the National Library of Medicine - National Institutes of Health, in coordination with the White House O ce of Science and Technology Policy. “CAiRE-COVID aims to facilitate the medical community, in the time-critical race to nd answers to various COVID- related queries in the hope of nding a cure for the virus,” said Prof. Fung, a -year expert in the challenging area of natural language processing research. The CORD- Challenge was important, Prof. Fung said, as it was the rst open-search and open-call initiative using AI in the medical eld. “With this collaboration, we hope that the power of machine learning in medical research will be unlocked,” she said. Working together with PhD student LEE Nayeon and MPhil student BANG Ye Jin, Prof. Fung has also found a way to flag misinformation by measuring how “predictable” the subject is in a statement, based on the scienti c evidence provided by the question-answering engine designed for the Kaggle contest and CORD- dataset. By June , the dataset had increased to over , articles. The HKUST team has publicly released the source code of both their question-answering system and myth debunker so that other developers can use it for further work. Meanwhile, Prof. Fung and her researchers are now collaborating with Dr. Oliver MORGAN, Director of the Health Emergency Information and Risk Assessment Department at the World Health Organization (WHO), on WHO’s Epidemic Intelligence from Open Sources initiative to co-develop question-answering and summarization technology. The goal is to nd answers from millions of online materials for early detection, veri cation, and assessment of public health risks. A Natural language processing expert Prof. Pascale Fung is seeking to unlock the power of machine learning for medical research. 21 IN FOCUS
RkJQdWJsaXNoZXIy NDk5Njg=