I. Introduction
Recommender systems, which extract user preferences and recommended content characteristics, aim to provide users with content that matches users’ needs. To encourage users to share their opinions, several feedback mechanisms are designed to guide users to share their views more concisely and clearly [1]. The most widely used explicit feedback mechanism in the recommender systems is the binary and numerical ratings and text reviews. The simple rating mechanism offers users a convenient approach to express their preferences, while many details may be left out. In contrast, writing comprehensive and high-quality text reviews is a time-consuming and unpopular work.