Recent Advances in Refinement Recommendations
Akshay Jagatap, Sachin Farfade · 2024
Navigating vast e-commerce websites with extensive product cata- logs can be a daunting challenge for shoppers. To assist customers in finding the products they desire, e-commerce platforms provide product attribute filters, commonly referred to as "refinements." These refinements serve as a vital navigational aid, enabling cus- tomers to refine their search results based on specific product at- tributes such as material, color, size, brand, etc. However, on mobile devices refinements are not easily discoverable due to lack of screen space. To improve discoverability, contextually relevant refinements are suggested in-line on search page by refinement recommenda- tion systems. In the work, we discuss the evolution of refinement recommendations strategies i.e, a) search query-based classifica- tion approach, for a given search query we train a classification model, with the refinements as labels b) session-based classification approach, for the given sequence of session interactions we train a sequence classification model, with the refinements as labels and c) session-based generation approach, with the sequence of session interactions as input and output as the refinement name.