/Realtime-Sentiment-Analysis

Web-based business intelligence tool that analyses customer feedback on an e-commerce store to derive useful metrics about products or services and make data-driven business decisions. Built using Kafka, Flink, Druid to handle data stream, Bag-of- Words model with TextBlob library for sentiment analysis and HTML/CSS/JS for visualization dashboard

Primary LanguageJupyter Notebook

Sentiment Analysis for E-Commerce Platform

  • A big data analytics solution to analyze customer reviews using Kafka, Flink and Druid for businesses to make data driven decisions based on customer feedback on e-commerce websites.
  • The system tracks product_viewed, product_added_to_cart and product_review events and have a sentiment analysis model in the background that provides the sentiment of the text review.
  • Raw input events are converted into processed events in the Flink job and are stored back into Kafka. Druid's Kafka indexer directly picks the processed events from the Kafka queue.
  • The paper is attached as a pdf that talks in detail about the design in building this system.
  • In short, there is a Flink job, analytics service to query Druid, a Flask API to do sentiment analysis and a dashboard to track the events generated in the system.