- 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
andproduct_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.
helozjisky/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
Jupyter Notebook