An intention-based recommendation approach

In recent years, online ever-increasing IT offers have emerged because of the prompt development of e-commerce. Consequently, recommendations have become a prominent necessity to help internet users discover interesting new products/services. In this context, recommender systems need more specific information about the users’ needs than the traditional ones, such as user’s profiles and their historical browsing, to be able to provide better recommendations. But actual recommender systems lack a consumer intention component based on analyzing free-text requests to better provide recommendations. With the target to resolve such insufficiency, our study aims at defining an approach to assist e-commerce shopping and avoid inappropriate information by a recommender system. It seeks to identify a customer’s intention from a free text and model it as an instance of an intention ontology. This instance constitutes formal knowledge intended to be a useful input to such a system. After analyzing linguistically the extracted content from a customer request, the intention-based recommender system prototype we developed in accordance with our approach identifies the customer intention and searches the items that probably correspond to this intention. Then, it sorts these items according to their satisfaction degree of the user’s need, using a filtering process, and suggests to the customer the closest items to this need. Qualitative and quantitative experiments were performed with an illustrative case and a comparison with human judgments to evaluate the effectiveness of the proposed approach. The good results prove the performance of our proposed approach.