Since the pandemic started every one of us has been receiving different messages that hold opinions and rumours by we don't even know who. This happened to the point that there was an actual 10-minute long video being sent around in family and friends groups and on Facebook which inculcated the idea that COVID19 could spread through 5G!!!! It was a flabbergasting moment, one where you actually stare at your screen for a while and wonder what happened to the world you live in. Hence, when Lumiata challenge came in and we saw fake news detection as one of the included topics, we knew we had to do something that could help the society we live in. This was the origin of FakeAlert and we are strong advocates of "SCAN BEFORE YOU SPAM"
What it does
FakeAlert proposes an application that provides you with an interface to input any xyz news that you hear and based on authentic resources it checks and informs you whether or not the news is true and credible.
How we built it
We are a Data Science team at a product development department of Integration Xperts (technology company). We have been working from home since March 18th and when this challenge came out we brainstormed on different ideas, checked the viability of each and decided on this one at the end.
Challenges we ran into
Communication is a challenge when you are working from home. There are times when you miss the space and would feel an urge to grab a marker and white board to explain what's in your head. Doing the whole thing from home, zoom meetings to discuss modules and approach although it was challenging but it was a learning experience in itself.
Accomplishments that we're proud of
Even though we think that we could have performed better have we been doing this sitting together in a room but still to be able to come up with an end to end solution given the daily job routines that we have we think we have pulled this quite nicely as only one of us in the team is an expert in NLP.
What we learned
Team work and deligence is the key.
What's next for FakeAlert
The solution presented is although end to end but quite amatuer and needs quite fine tuning and extra work. The ideas were numerous but we couldn't implement them because of the time constraint. Since we plan to launch this in the form of a product we will continue building and refining the modules.