Inspiration
With the recent outbreak of COVID-19, there was a major online epidemic of fake news causing mass hysteria. We wanted to offer a fast and simple way for any user to check if a news article is real or fake, then read a summary and have options to view related links.
What it does
It is an extension, that when opened on a valid news article, displays a summary of the article, related links to the article and whether the article is real or fake.
How we built it
We built an algorithm that retrieves the URL of the tab open when Credable is used. This URL is then evaluated to be a news article (by seeing if it's on Google News), and if it is, is then summarized using newspaper3k, and related articles are provided by requesting for the title of the website and printing the top 3 results when searched into Google News. Then the algorithm uses an ML model, trained with a database of known credible and incredible articles, to predict the credibility of the given article. All of this is displayed using Google's Chrome Extension API and JavaScript.
Challenges we ran into
Originally we planned to write a compare function that compares the credibility of one article with the known credibility of another related article, then with each inputted article we planned to calculate the comparison between the article and all its related articles that we have defined credible. This would return an index from 0 to 1 which would be a more useful answer than either 'Real' or 'Fake' but we were unable to think of a way to compare the semantic value of two texts reliably. Hence we switched our idea to our current one.
Accomplishments that we're proud of
Other than our previously stated challenge, we didn't have too many major problems regarding the way our project functioned. Most of our initial planning was thought out and the dividing of roles was done so efficiently.
What we learned
Natural Language Processing is more complicated than it seems. It may seem easy to look at two texts and understand if they are talking about the same topic, but teaching a computer to compute that answer is very complex and requires a lot of deep thinking.
What's next for Credable
Credable only currently outputs whether an article is real or fake. Our initial idea of a float from 0 to 1 representing the credibility of the article is much more useful towards users, if implemented correctly. Our next step would be finding a way to implement this feature, working around our problems previously stated.
Log in or sign up for Devpost to join the conversation.