Veritas is a blockchain repository of trusted news sources vetted through our machine learning algorithms and NLP techniques. Our team is committed to creating a more open information stream that is resistant to tampering and obfuscation.

Our interest in developing Veritas stemmed from the high profile cases of manipulation from seemingly innocuous fake news sources that have proliferated the Internet. With over 90 percent of Americans using the internet on a daily basis and 66 percent relying on the internet as their primary source of news, it is important to ensure that our information stream is reliable and resistant to censorship and manipulation efforts. Our team has developed a blockchain based repository that allows trusted news sources to be appended and viewed by members of the network, ensuring that previously trusted articles cannot be later modified for malicious purposes.

Veritas utilizes Taboola's native API to filter trend categories, such as law, politics, etc. and scrape news articles corresponding to distinct topics within each category of interest. Our project incorporates sentiment analysis to articles grouped under the same topic and determines a confidence score based on the proportion of related articles that agree with the sentiment of each headline. This is achieved through the use of team UCL's submission in the Fake News Challenge, which is linked below.

The main issues within the fake news space are censorship and authenticity. This problem is difficult to solve as many news sources make subtle modifications to articles for potentially malicious purposes. Furthermore, it is difficult to annotate each article manually with sites like PolitiFact due to the sheer number of articles published. Our solution is the only current model that utilizes real-time fake news detection and verifies the integrity of the article through blockchain.

Team UCL's Fake News Challenge Submission:

Taboola API:

Our Hackathon project was awarded "Best Use of Taboola Trends API" at LAHacks 2019

Share this project: