Voice-Sentiment-Classification

Overview

Voice-Sentiment-Classification is a project that converts Persian voice inputs to text using the Whisper speech recognition model and classifies the sentiment of the text using a fine-tuned ParsBERT model. The ParsBERT model has been pruned and converted to ONNX format for optimized performance. The entire system is deployed using FastAPI and Docker for efficient and scalable deployment.

Table of Contents

• Features
• Data
• Architecture
• Setup and Installation
• Usage

Features

• Converts persian voice input to text using Whisper speech recognition model.
• Classifies sentiment of text using a fine-tuned ParsBERT model.
• Optimized ParsBERT model in ONNX format for efficient inference.
• RESTful API using FastAPI.
• Dockerized for easy deployment.

Data

snappfood-sentiment-analysis: a Persian sentiment analysis dataset

Architecture

1. ParsBert : Transformer-based Model for Persian Language Understanding.
2. Whisper: Whisper model converts voice input to text for persian language .

Setup and Installation

Clone the Repository

git https://github.com/MohammadRoodbari/Voice-Sentiment-Classification.git

cd Voice-Sentiment-Classification

Usage

To start the FastAPI server

cd app

uvicorn main:app --host 0.0.0.0 --port 8080

Docker Deployment

To build and run the Docker container, follow these steps:

docker build -t Voice-Sentiment-Classification

docker run -d -p 8000:8000 Voice-Sentiment-Classification