A multiturn recommender system with explanations
Luan Soares de Souza, André Levi Zanon, Lucas Padilha Modesto de Araújo, Marcelo Garcia Manzato · 2023
Recommendations engines use interactions between users and items to predict the preferences of the users and generate recommendations for them. However, because they rely on historical data, the user’s interest at the moment may not be captured. In this context, Conversational Recommender Systems (CRSs) have been proposed in order to provide recommendations that provide suggestions based on the user’s current interests by eliciting information from the user in turns in which the system can ask for the user to understand more about the current interest in the time or recommend. In that regard, we propose CRSs for cold-start users as a Knowledge Graph search. The system also employs a Natural Language Processing module to explain the recommendations, bringing transparency to the recommendation algorithm.