This work focuses on the text-based recommendation challenge on the Web. In fact, with the emergence of electronic media and the explosion of news articles’ volumes on the Web, it has become difficult to suggest recommendations that best suit users’ interests and preferences. In this work, we propose a model of recommendation whose main objective is to guide Internet users in the great mass of news on the Web. Indeed, our contributions are based on three main points, namely, (1) the online semantic analysis of news articles based on their textual content, (2) the incremental segmentation of news articles into categories while taking into account the scalability problem and (3) the dynamic nature of the proposed recommendation approach that adapts its suggestions based on the users’ context and behaviors. An experimental study was conducted on a real-world use case in order to validate and evaluate the quality and the scalability of our proposal within a production environment.
Knowledge-guided Deep Reinforcement Learning for Interactive Recommendation