Instant, Personalized Search Recommendation

Personalization has proven to be a useful method in helping users find relevant items and documents. More and more digital stores implement some form of personalization to help users find products. Even though there are a lot of work in the field of personalization and recommender systems, search engines are only personalized to a limited degree. In this work, we investigate how to develop a personalized search engine, which gives instant product suggestions to the user based on user preferences and queries. As part of this, we implement and test our methods based on Bayesian classification and Kullback-Leibler divergence. In addition, we compare these methods with the baseline methods used today. Our experiments and user-based evaluation show promising results with respect to the relevance of the personalized suggestions and interactiveness. Overall, this research found that our approach is able to retrieve personalized suggestions instantly, even for products not directly containing typed terms. We expect our method to be useful for digital stores, in general