On the Measurement and Prediction of Web Content Utility

Abstract

Nowadays, various types and large amount of content are available on the Web. Characterizing the Web content and predicting its inherent usefulness become important problems that may benefit many applications such as information filtering and content recommendation. In this article, we present a brief review of the existing measurements and the corresponding prediction methods for Web content utility. Specially, we focus on three close and widely studied tasks, i.e., content popularity prediction, content quality prediction, and scientific article impact prediction. While reviewing the existing work in each of the above three tasks, we mainly aim to answer the following two fundamental questions: how to measure the Web content utility, and how to make the predictions under the measurement. We find that while the three tasks are closely related, they bear subtle differences in terms of prediction urgency, feature extraction, and algorithm design. After that, we discuss some future directions in measuring and predicting Web content utility