AI & COVID-19
Scientists around the world are working hard to develop vaccines and drugs to prevent the spread of the global corona virus. And it will help to differentiate between misinformation. The question is, why is it so important? Scientists are worried about the corona virus.
Ever since the virus was discovered, people from various fields have expressed their views, but many of their published research papers have not been reviewed. That is why it is so difficult for scientists to fully understand the corona virus and to distinguish between right and wrong information. How does this tool work? It’s called “SciFact” and was designed for that purpose by the Allen Institute for Artificial Intelligence (AI2) in Seattle.
After entering any scientific claim in the system’s search bar (for example, high blood pressure can exacerbate the symptoms of corona virus sufferers), the relevant research papers will be displayed in a feed and shared with the user. It will be stated in which papers this claim has been refuted and in which it has been confirmed. In addition, a summary of each paper will be presented, highlighting the sentences based on which the accuracy or inaccuracy of the claim was determined. How is the system designed? This system uses a neural network called VeriSci. The training used a pre-existing data set to verify the facts obtained from a website, followed by a new scientific data set consisting of 4,091 scientific claims and 1,835 abstracts.
AI2 researchers took advantage of the Semantic Scholar, a public database of scientific papers, introduced in 2015 to select this second dataset. Facilitators have put these sentences in the form of claims.
These experts then analyzed the summaries of the citations and identified the sentences confirming the claims they had made. How did this tool work? As a result of testing on scientific claims about the corona virus, researchers found that 23 of the 36 tools were able to locate relevant labels and correctly label them.
This performance is not 100% accurate, but it is much better than the results of other databases that check the facts. It also proves that if systems with artificial intelligence are provided with more training data and efforts are made to increase their understanding of natural language, they can be helpful in further scientific research.
What kind of tools can this tool not be used for, but SyFact aims to assist scientists researching the corona virus in examining their theories or new claims in light of existing scientific material. Experiments with this system are still ongoing and therefore experts should do their own research instead of relying entirely on its results. In addition, researchers working on it say that the fact does not check the accuracy of the research papers shown and therefore the scientists who use this system cannot fully rely on it.