Sentio - Sentiment & Toxicity Analysis Web App

Sentio is a sentiment analysis web application based on LSTM built using TensorFlow, Streamlit and other essential libraries. It allows users to classify text comments as Positive, Negative, or Neutral and classify Toxicity contents in the comment by simply entering text into a text field.

Features

• Classifies comments into Positive, Negative, or Neutral categories.
• Identify Toxic contents in comments
• Provides a user-friendly interface using Streamlit.
• Uses a LSTM model for sentiment classification.
• Visualizes the sentiment prediction and toxicity with color-coded progress bars.

Tech Stack

• Streamlit: For building the web interface.
• TensorFlow: For deep learning and sentiment classification.
• Keras: For implementing the neural network model.
• Pandas: For data manipulation.
• NumPy: For numerical computations.
• Pickle: For saving and loading the tokenizer model.
• sklearn: For splitting data and preprocessing

Requirements

To install the required packages, create a virtual environment and install the dependencies from requirements.txt:

pip install -r requirements.txt

The requirements.txt file should include:

streamlit
pandas
numpy
tensorflow-cpu
pickle-mixin
sklearn

How to Run the App

1. Clone the repository to your local machine:

   git clone https://github.com/RijoSLal/sentio.git

2. Install the required dependencies:

   pip install -r requirements.txt

3. Run the Streamlit app:

   streamlit run sentio.py

4. Open your browser and go to http://localhost:8501 to use the Sentio app.

Files

• sentio.py: The Streamlit app for the sentiment analysis interface.
• sentiment.h5: The trained sentiment analysis model.
• tokenizer.pkl: The tokenizer used to preprocess input text.
• train.csv: The training dataset.
• test.csv: The test dataset.
• toxic_data.csv: the dataset for toxicity detection
• sentiment_toxic.h5: The trained toxicity recognition model.
• tokenizer_toxic.pkl: The tokenizer used to preprocess toxic text.

License

This project is licensed under the MIT License. Feel free to modify, use, and distribute the code in any way you like. See the LICENSE file for details.