---
title: Sentiment classifier
emoji: 🎭
colorFrom: blue
colorTo: red
sdk: streamlit
sdk_version: 1.25.0
pinned: false
app_file: sentiment_analysis.py
---

# bert-sentiment-analysis

Prototype that classifies text into positive or negative sentiments using a fine tuned bert model

## Installation of dependencies

`pip install -r requirements.txt`

## Usage

1. Download the [trained model](https://huggingface.co/rootstrap-org/bert-sentiment-classifier/blob/main/sentiments_bert_model.h5) and move it to the *models* directory
2. Create a `.env` file and set a MODEL_NAME property with the name of the trained model file and optionally a MODEL_REPOSITORY_NAME property with the name of the huggingface repository of the model.
3. Use the tool:
* To use it as a **streamlit web app** run:

  `streamlit run sentiment_analysis.py`

  It will open a web app on `http://localhost:8501`

* To use it from **command line** run

  `python sentiment_classificator.py <TEXT_TO_CLASSIFY>`

## Training

1. Download the [all_sentiment_dataset.csv](https://drive.google.com/file/d/175Ccd3B6kLWMBvr1WAUzQJT4TwgzXF6N/view?usp=sharing)
2. Execute the *classify_sentiment_with_bert* notebook which is in the *notebooks* directory
3. The model should be saved under *models* directory as **sentiments_bert_model.h5**