Sentiment Explainer: Streamlit app that makes Transformer models explainable

Hi community,

I have created the following app, deployed via Streamlit-Share (thanks a lot for this BTW)

It let’s you predict the sentiment (positive or negative) of any given input string. Additionally the attribution towards the final prediction per token can be seen.

There is even a little easter-egg in the app:) hopefully you’ll find it and it will resonate.

Repo can be found here : GitHub - sgoede/sentiment_explained

Please let me know what you think of it :slight_smile: