Euphorus

Inspiration

We had an existing idea on how we could increase customer satisfaction through not just feedback but other activities as well. The prompt gave us an idea, when it asked us to document the trends in customer reviews, that we could include different statistics, such ad past few days trends, and how critical fixing each problem is. We could cluster alike reviews together and rate them in their importance.

What it does

It collects data from public websites like reddit and classifies them into different levels 1 through 5 or red to mild yellow, with red being critical. It groups reviews with similar keywords and creates trends for them like how a stock market does. It has a voice input and output feature for the disabled and normal people depending on whether they want to read the reviews or listen to them. We also experimented with creating an AI song with the top hot reviews as lyrics. The negative reviews have a 4 step solution and the positive reviews have a 4 step on how to improve even more or why that feature should stay.

How we built it

We used Python, and many of its libraries like pandas,, pynum, torch, transformers, Streamlit, scikit-learn, gTTS, SpeechRecognition, and matplotlib for backend functionality and also its web framework, flask and its api to combine front end and back end for an almost full stack product, with AI voice integration.

Challenges we ran into

One ofd our files forecast.py was unable to read the .csv files we created using a preceding program, which we are still working on. we also ran in the issue of using api for collecting data from reddit or other social media websites but the websites for it are not free, so we chose a substitute praw which is a python wrapper for reddit to achieve the same functionality as an api. we also ran into problems with the pyAudio when integrsating voice Input/Output for the project so we decided to go with sox, which was a better option. The flask framework also gave us trouble but we were able to overcome the challenge to design a siml yet decent UI.

Track: T-mobile challenge prompt