Introduction: Snapdeal is an Indian e-commerce company that offers a wide range of products, including electronics, fashion, and
household goods. Founded in 2010, it has grown to become one of the largest online marketplaces in India, with a focus on providing a
platform for small and medium-sized businesses to reach customers across the country.
Business Need: Snapdeal with a significant user base in semi-urban and rural India, faced inefficiencies in creating review summaries
and translations due to manual processes. This approach was error-prone, time-intensive, and expensive, impacting operational efficiency
and customer satisfaction.
The company needed a scalable solution to reduce purchase decision time, enhance customer engagement, and deliver accurate, concise
review summaries to drive better user experiences and optimize costs.
Implementation:
- Data Collection and Storage: Gather raw product review data and store it systematically.
- Data Chunking: Segment raw data into manageable chunks for efficient processing.
- Queuing System: Implement a queuing system to organize and manage product data flow.
- Data Processing: Extract keywords and sentiments from the product review data.
- Multilingual Summarization: Generate concise summaries of insights and translate them into ten languages.
- Regular Updates: Ensure stored summaries are regularly updated for accuracy and relevance.
Reaping Rewards:
- Scalable Processing: Customer was able to process 30,000 products and 1 million reviews in a single iteration, with each product processed in under a minute across all stages, including extraction, translation, and storage.
- Cost Reduction: Reduced operational costs by 35% by eliminating manual translation and validation activities for each product.
- Enhanced Accuracy: Achieved 87%-93% accuracy in keyword extraction and 90%-95% accuracy in sentiment extraction, ensuring high-quality data insights.
- Multilingual Efficiency: Delivered 80%-85% accuracy in summarization and translation for 9 Indian languages, improving the
accessibility and relevance of product information.
