MobileMoods

Inspiration

We were inspired by TMobile's dedication to customer happiness, and the possibilities to create an innovative product when listening to real feedback. Keeping human sentiment at the heart of our application, we let the customer's inspire us to formulate ways to address their concerns.

What it does

MobileMoods creates summaries based on information collected after scraping DownDetector (https://downdetector.com/status/t-mobile) comments on TMobile, and then utilizes Gemini to create steps for moving forward and addressing customer concerns.

How we built it

We built the project by splitting it into three parts - writing its backend, writing its frontend, and writing the scraping script. Python libraries such as BeautifulSoup were used to create the script, and data collected from DownDetector was formed into a JSON file. The backend processes this day by parsing through keys defined by comment info and organizing it into a 2D list, and then traversing that list and calling Gemini API to do sentiment analysis and assign each comment as positive/negative/neutral. Duplicate comments are also appended. The frontend was created with Streamlit, where a minimal color palette uses TMobile's brand identity for a sleek and minimal UI. Three pages on the website display the data of customer sentiment, the analysis, and decisions from the metrics.

Challenges we ran into and what we learned

Challenges we faced included figuring out how to interface between the collected data streams, work with new libraries and dependencies that team members were previously unfamiliar with, learning how to scrape a website, and determining the best way to integrate AI use and navigate Gemini's API quote limit.