Sentiment Review Detection

Overview

This application utilizes Artificial Neural Networks (ANN), Recurrent Neural Networks (RNN), and Transformer models to detect and analyze sentiment in user-provided reviews. Users can input reviews manually or upload a CSV file for bulk analysis. The application is built using Streamlit, making it user-friendly for interactive prediction tasks.

Features

• Manual Review Input: Users can type or paste a single review and get the sentiment analysis in real-time.
• Bulk Review Processing: Users can upload a CSV file containing multiple reviews to get batch sentiment predictions.
• Model Selection: Choose between ANN, RNN, and Transformer models for sentiment prediction.
• Visual Analytics: Generates bar plots showing the distribution of sentiments across the reviews.

Prerequisites

Ensure you have the following installed:

• Python 3.8 or newer
• Streamlit
• TensorFlow
• Scikit-learn
• Pandas
• Numpy
• Matplotlib
• Joblib

Usage

To run the application, navigate to the project directory in the terminal and run:

streamlit run main.py

Interactive Components

• Choose Model: Select the prediction model from a dropdown.
• Choose Mode: Choose either 'Manual Input' for single review predictions or 'Upload CSV' for bulk predictions.
• Detect Sentiment: After entering a review or uploading a file, click this button to generate predictions.

Data Format

For CSV uploads, ensure your data is formatted with a column named 'review' containing the text entries for analysis.

Example:

review
"I love this product!"
"Terrible customer service."

Models

The models used in this application are trained using separate notebooks:

• LSTMANDANN.ipynb for the ANN and RNN models.
• transformer.ipynb for the Transformer model.

Ensure these models are correctly loaded from the models directory.