Through the Spotify lecture, we had access to a language neural net. Knowing we wanted to work with Chicago data, it gave us the idea to use the neural net on Twitter posts written about Chicago communities.
What it does
Each neighborhood of Chicago is highlighted by a color ranging from green to red that indicates the average happiness confidence of the neural net. When you hover over the neighborhood, you can then see a pie chart of the proportion of positive tweets to negative tweets. Note that these 2 metrics are distinct. For example, O'Hare is green on the map, but there is a greater ration of negative tweets. This is because the negative tweets have low confidence scores (meaning the text is closer to neutral than extreme negative), and the positive tweets about this region have higher scores.
How we built it
We used a Twitter API to request tweets referencing the communities in Chicago. We then ran them through a language neural net to assess the positivity of each tweet (accompanied by a confidence score). Using the confidence score we were able to assign a "happiness score" to each neighborhood and put them on a map of Chicago with borders of the neighborhood, allowing our data to be interactive. We also used Chart.js to create pie charts of the ratio of positive to negative tweets.
Challenges we ran into
We didn't have access to the full archive of the Twitter API, which limits our data. Also due to lack of time, we couldn't have the neural net adjust for non-English tweets.
Accomplishments that we're proud of
We used the Google Maps API and Chart.js to make an interactive map that represents our data in an aesthetic way.
What we learned
We learned that it's difficult to access Twitter data. We also improved our web development skills and learned how to make an interactive page.
What's next for Chicago SatisFraction
Expanding our dataset and be able to account for non-English Tweets.