Hybrid Personalized English-Text Recommendation System Using Collaborative Filtering and Content Classification

Abstract

The accelerated growth of online sources of information has increased tremendously the difficulty of finding pertinent information that meets personal interest of a user. The recommendation systems have thus emerged as critical instruments to sort through vast amounts of information as well as provide customized recommendations. The proposed research presents a hybrid personalized English-text recommendation system that combines collaborative filtering and content classification approaches to enhance accuracy of the recommendation. The collaborative filtering module examines the past patterns of user interaction whereas the content classification module examines textual content characteristics through the TF-IDF representation and nave bayes. A hybrid scoring mechanism that involves a combination of the outputs of both modules with a weighted scoring mechanism is used to generate personalized recommendations. The standard measures of evaluation like precision, recall and accuracy that are conducted in an experiment also prove that the hybrid method achieves better results than the standalone methods in terms of recommendation. The findings suggest that the combination of behavioral and semantic information promotes the relevance of the recommendations and the robustness of the system when used in large-scale information setting.