Our team was inspired by the movie "The Social Dilemma." A 2020 Netflix documentary examining the inner workings of social media's addictive design. A design created to maximize profit and brainwash people's views and actions through subtle, but extremally effective manipulation.
What it does
Our project analyzes twitter user data mentioning: political figureheads, politicians, presidents, senators, congress, news channels, and specific keywords through social network scraping tools and programs. With this data we collected we trained a machine learning model to pick up on the negative, positive, and neutral sentiments of these tweets. With our system, we collected even more data from the past 2015-2016 elections. Using this as a comparison with our 2020 COVID-19 political data set, we were able to evaluate the differences in sentiment. When comparing these two times of political tension we saw a larger percentage of negative sentiment proving an increase in discourse and divide in our nation as a product of COVID-19.
How we built it
A majority of our project took place on google colab - a cloud based analyzing platform optimized for users to tackle big sets of data. Using the language python, we wrote multiple systems to search for specific users / keywords and sort those data sets, create data sets using specific time periods and names, and search the surface of the twitter database to access public tweets. With this data we fed large amounts to machine learning model that was then trained and honed to mark tweets for different sentiments. With this we were able to see COVID-19 was a large factor in agitating already tense political relations - furthering them into intense polarization.
Challenges we ran into
The main challenge we ran into was our data not being large enough for the model to take in. Being that we wanted this model to be as accurate as possible when taking in our date we had to very large data sets to originally train this model.
Accomplishments that we're proud of
One of our accomplishments that we are proud of is how we handled these large amounts of data. Some of our members barely dealing with python had to take on the task of creating large sorters with specific variables. We are proud of how it came out and believe it is something that could be used again.
What we learned
We learned the importance of large and well put together data sets for machine learning. If given more time we would create out own large data sets over time and use that for our model instead. Collecting data takes time and thus we learned that spending more time on this project in the future is worthwhile.
What's next for Divided We Fall
With Divided We Fall our system is ready to take on more data for a variety of topics. Continuing on our model we would get data from events across history to show the change in time. Not only this but using the scraping tools we can expand into other social networking sights and tackle other topics beyond our main topic of political division.
Log in or sign up for Devpost to join the conversation.