Sentiment Analysis with BERT + CNN

Project description

This project was developed during the study of natural language processing using the BERT architecture.


The code includes a series of comments about the project's development and any decisions made. You can find the code from this project on my Github, thought this link.


The goal of this project is to predict if a phrase represents a positive or a negative sentiment.


After training phase, the model presents 82% of accuracy.

Example

Live app

You can acess and test the app visiting a Streamlit App. Click here.

Nathalia Castelo Branco

I'm a Civil Engineer who has a master's degree in Civil Engineering with a focus on Structural Dynamics and Numerical Modeling from State University of Rio de Janeiro (UERJ). I also have an MIT degree in Artificial Intelligence from Infnet Institut.

Follow me