Being freshmen at the University of Chicago, we always hear from upperclassman about the chilly, bone-biting winter quarter. More than the physical ramifications of this weather, it's believed that the winter affects the students mentally and psychologically. The student body apparently has a more negative/depressed attitude compared to other quarters. Therefore, we, 3 curious freshmen, decided to investigate more on this topic by diving into realms of NLP and web scrapping.
What it does
We decided to carry out a sentiment analysis on the "Uchicago Secrets" Facebook page. This is a student-run Facebook page wherein students throw their opinions, reflections, secret confessions about their daily lives on campus. Analysing the sentiments in these Facebook posts gave us a general sense of the mood on campus. We compared the mood of the winter quarter with other quarters and saw many interesting features.
How I built it
First, we used Selenium and BeautifulSoup to web scrap the Facebook page and clean up the data. Next, we set up a Google Cloud Platform on Google Colab and set up a Sentiment Analysis Model. We fed the cleaned data to the model and used Pandas and Numpy to make a data frame of the average sentiment and the corresponding time of the year. Lastly, we used MatPlotLib to graph our data and make valid conclusions. Next, we scaled this to other colleges and their seasonal sentiments.
Challenges I ran into
Web scraping from Facebook without using Graph API, Setting up the Google Colab Platform
Accomplishments that I'm proud of
We finally got the code to run and produced a corresponding graph for it. The data did show that the general mood during winter is significantly lower than in other quarters.
What I learned
Google Colab, Web scraping, Lot of patience, How to thrive in sleep deprivation
What's next for Secret Sentiment Analysis
Analyse sentiments not only seasonally but weekly. Find other means to collect free expression/sentiment for example memes page etc.