Product data is at the heart of any e-commerce site. Product data enhances key functionalities such as search, recommendations, faceting, filtering etc. Traditionally within retail, enrichment has been done to extract features from product data using systems such as PIM. Product data consists of rich representations from images and text which contain a lot of important information about the product. With the exponential growth, and incredible developments in computing technologies have contributed to the rise of multimodal learning that utilizes NLP and different media forms like image, text, video, sound, etc.

In this talk we will explore how we can harness the power of multimodal models to learn representations from product data containing text and images to unlock a richer set of attributes that provide value to our customers.