Learning a region of user's preference for product recommendation

A Conversational Recommender System(CRS) aims to simulate the real world setting where a shopkeeper tries to learn the preferences of a user through conversation and the user, in turn, learns about the product base. Often the preference of the user is captured in the form of feature weights and they are updated assuming each feature is independent. These feature weights are used in product retrieval from the product base. The independence assumption of the features is not always a good choice, and we will address this concern in our work. Each product in the product base has its own prospective buyers characterized by a certain combination of preference choices. The goal of this work is to discover knowledge about the preference choices of prospective buyers for each product offline and, later use it to recommend products to the users based on their preference choices.