Semantic Analysis for Keywords Based User Segmentation from Internet Data

Nowadays, Internet has been one of the major advertising channels and behavioral targeting has become increasingly important for improving the click-through rate of online advertisements. One of the key research problems in behavioral targeting is how to group users into segments with similar interests or backgrounds. In this paper, we propose a web page-oriented and keywords-based approach to address this problem. Our approach includes two key components: keyword similarity measurement and keyword similarity based user segmentation. These two components serve as plugins and can be replaced with better algorithms or measurements, making our approach very flexible. We have implemented the first key component, and provide preliminary results to illustrate the effectiveness of this component in finding similar news pages based on their keywords.