Chatwithrec: Toward a Real-Time Conversational Recommender System
Nowadays, the Internet helps to increase the demand for the development of commercial applications and services that can provide better shopping experiences and commercial activities for customers around the world. Research has investigated many approaches to produce optimal results in conversational recommendation systems (CRSs) that aim to analyze user-generated content (UGC) in online social networks (OSNs) landscape to suggest appropriate recommendations. However, it is hard to obtain the user’s preferences and actual requirements at the beginning of the interaction, which makes users unsatisfied with recommending items and services. In this paper, we propose a conversational recommendation system called ChatWithRec that aims to solve the cold-start problem by detecting topics and matching them with appropriate content, like ads, after analyzing the user’s contextual conversation dynamically, to increase the accuracy of recommendations. We apply the Latent Dirichlet Allocation topic model (LDA) to analyze user’s conversation and observe topics dynamically. We integrated Google AdMob and customized ads databases to deliver conversation-related tasks. We evaluated the performance of our CRS by applying several statistical metrics. The results are encouraging and indicate that the system is fast and satisfies users by getting what they seek without interrupting their conversation flow, and solve the cold-start problem.
Keywords - Conversational Recommendation System, Real-time, Advertisement, NLP, Topic Modelling, Latent Dirichlet Allocation, Cold-start Problem.