Zhao, Yichun

Project title: Dealing with Fake News: Automated Fact-checking AI Algorithms and Trust Database to Monitor, Summarize, and Cross-reference Information, and to Identify Claims

Department: Computer Science

Faculty supervisor: Dr. Jens Weber

"Fake news is defined by the Ethical Journalism Network as “deliberately fabricated and published” information intended “to deceive and mislead others.” It manipulates the ignorant into false beliefs and causes negative societal impacts. Although there are various organizations such as Snopes and Full Fact dedicated to check facts and verify news articles, automation must be introduced to assist humans to monitor, summarize, and cross-reference user-generated information, news articles or reports, and to identify claims from them due to the speed and scalability of the internet.

Such automated fact-checking algorithms need known facts to compare. We do this by building a trust database sourced and summarized from reputable news sites such as Reuters and PolitiFact. This should be achieved in an automated way to constantly update facts and to avoid repeated information. The database should also contain claims previously verified by the algorithm for comparison.

To design the automated fact-checking algorithm, I will evaluate the suitability, performance and accuracy of using machine learning and text analysis algorithms for the purpose of monitoring wide-spread information on social media, summarizing information, identifying claims, and cross-referencing with the trust database. Other details include spotting facts which were already checked in the database, and analyzing questions in a text to check for facts as answers. Training data must be designed carefully to avoid any human bias for machine learning models to accurately identify fake information. There are already available test data in the public domain. Lastly, identified limitations for this research will also be addressed."