After doing some research on JetBlue's mission, we found that there was a major emphasis on inspiring humanity and customer-centric service. Although there are many tools for Sentiment Analysis available online, many of them did not pick up on the nuances found within discussions that take place online. Thus, we developed a social media analysis tool that takes common factors into account, like emojis and sarcasm, in our algorithm.

What it does

We collected the data using the Twitter API and stored our tweets. From there, we applied text pre-processing techniques, created a model using a variety of factors, and developed a dashboard for data visualization purposes.

How we built it

For prototyping/testing purposes, we used Jupyter Notebooks. The dashboard was created with React.

Challenges we ran into

Cleaning the data was a challenge because there were many instances where variables such as sarcasm, verb tense, context, etc. would impact the sentiment score, but various sentiment analyzers would give us incorrect results.

Accomplishments that we're proud of

Connecting our React app to our MongoDB that is stored on an instance using Google Cloud's Compute Engine!

What we learned

We learned that finding meaning insights from real, untrained data can be a challenge in a short time span. The data we worked with was messy, and finding a consistent format was certainly a challenge, but decent progress was made!

What's next for for JetBlue

We can further improve on this model by incorporating more human-like features that can be found within social media posts, such as slang words/acronyms. We would also like to see if this would work with data from other social medial platforms, such as Instagram or Facebook.

Share this project: