Background
Our client, a medical equipment supplier, was struggling to provide personalized recommendations to their customers. They had a vast inventory of products, but their sales team found it challenging to match the right product with the customer's specific needs. Our team was tasked with developing a recommendation system that would help automate this process and improve customer satisfaction.
Problem
The main challenge we faced was the lack of data available for training the recommendation system. The existing data did not have enough information about the specifications of medical equipment, which made it difficult to create accurate matches. We needed to find a way to generate more data without relying on manual input from the sales team.
Solution
We decided to use spaCy, an open-source natural language processing library, to develop a rule-based matching engine. This engine could analyze product descriptions and extract relevant keywords and phrases that would be used for matching. To generate more data about medical equipment specifications, we performed data augmentation using techniques like synonym replacement and back-translation.
Impact
Improved accuracy of product recommendations by X%
Reduced time spent by sales team on manual matching by X%
Increase in customer satisfaction ratings by X%
Team
Amy Chen - Data Scientist
Developed rule-based matching engine using spaCy
Led data augmentation efforts using various techniques like synonym replacement and back-translation
Daniel Kim - Machine Learning Engineer
Built recommendation system using the matching engine and augmented data
Optimized system for scalability and performance
Jessica Lee - Project Manager
Oversaw project timeline and deliverables
Managed communication with client and ensured project goals were met