Sentiment Analysis API for E-Commerce Fashion Reviews

Overview

This project provides a backend API for sentiment analysis, designed to classify reviews into positive, neutral, and negative sentiments. It also includes additional features such as keyword extraction, word cloud generation, and summary interpretation.

Features

• Sentiment Classification:
  A pretrained model classifies sentiments as Positive, Neutral, and Negative.

• Keyword Extraction:
  Identifies key phrases and terms from the review data.

• Visualizations:
  Generates word clouds for quick identification of common terms.

• Summaries:
  Uses an LLM to provide concise summaries of key themes and issues.

Model Training

• Dataset:
  Amazon fashion review data: https://www.kaggle.com/datasets/haoboxu/amazon-reviews-for-sentiment-analysis

• Performance Metrics:

  • Precision: 73.2%
  • Recall: 73.2%

Getting Started

Prerequisites

• Python 3.8+
• FastAPI
• Pydantic
• FastText
• spaCy
• pandas
• scikit-learn
• wordcloud
• matplotlib

Installation

1. Clone this repository:

   git clone https://github.com/yourusername/sentiment-analysis-api.git
   cd sentiment-analysis-api