Computational model for generating interactions in conversational recommender system based on product functional requirements

摘要

Conversational recommender system is a tool to help customer in deciding products they are going to buy, by conversational mechanism. By this mechanism, the system is able to imitate natural conversation between customer and professional sales support, for eliciting customer preference. However, many customers are not familiar with the technical features of multi-function and multi-feature products. A more natural way to explore customer preferences is by asking what they want to use with the product they are looking for (product functional requirements). Therefore, this paper proposes a computational model incorporating product functional requirements for interaction. The proposed model covers ontology and its structure as well as algorithms for generating interaction that comprises asking question, recommending products and presenting explanation of why a product is recommended. Based on our user studies, both expert users (familiar with product technical features) and novice users (not familiar with product technical feature) prefer our proposed interaction model than that of the flat interaction model (interaction model based on technical features). Meanwhile, functional requirements-based explanation is able to improve user trust in recommended products by 30% for novice users and 17% for expert users.