Friend Flow

Inspiration

Communication is enabled through language, and every culture has its own distinctive characteristics stemming from it! I wanted to build something focused on language, to facilitate conversation between people who may not speak the same language. This would be a super useful tool when travelling to countries where you don't speak the local language.

What it does

Friend Flow offers literal translations services, along with a custom sentiment analysis classifier. I trained an xgboost model with 6 classes ('sadness', 'joy', 'love', 'anger', 'fear', 'surprise'), and an F1 score of 0.94. I used this Kaggle dataset.

Friend Flow can also detect threats and recommend the user to call EMS if they are in a dangerous situation.

With OpenAI and some prompt engineering magic, I also added a feature to generate recommended ways to respond (you can refresh the tab to get new responses!), translated in English and in the foreign language. To hear how to pronounce the recommended response, simply click "Speak".

You can input either text or audio files, and record new audio on the web app for convenience.

Challenges I ran into

It was a challenge to synthesize all of these features together, and learn how to use Flask, of which I am relatively new to using!

Accomplishments that I'm proud of

I'm proud that I was able to get everything working together smoothly. I think this is a pretty practical project, and I'd like to add more features in the future like detecting languages without explicitly inputting them.