florianebhz / sentimentanalysisapp

End-To-End Sentiment Analysis web application with Twint, Rubrix, BERT, Streamlit, FastApi, Docker

End-to-end sentiment analysis Application (from scrapping to App deployment)

Context :

As a data scientist, my favorite tasks were machine learning/deep learning models creation and training and data analysis and vizualisations. After my first professional experience, I realised that training models locally is not meaningful in the real-world. Indeed, in order to resolve business problems with machine learning models, we have to deploy them into production.

I decided to get more skills in machine learning engineering and data engineering related tasks through personnal projects. This project's finality is a web application that mainly predict the sentiment (NEGATIVE-NEUTRAL-POSITIVE) in an input sentence of the user.

Future functionalities : On the dashboard page, you can see the performances of the model. On the Database page, you can explore the test dataset and interact with it.

The projet consists of multiple steps:

• Data collection and engineering
• Model training
• Model serving through APi
• Web Application creation
• App containerization
• App deployment on the cloud
• Next : Database connection

1. Data collection and engineering with (Twint, Rubrix, Pandas)

2. Model training (Bert)

3. Model serving through APi (FastApi)

4. Web Application creation (Streamlit)

5. App containerization (Docker , Docker-compose)

6. App deployment on the cloud (AWS EC2)