Analyzing Multiple News Sites by Contrasting Articles

Today, there is access to large numbers of news sites in different countries, and there are several experimental systems, such as Newsblaster and NewsExplorer, that integrate news articles about a particular event from multiple news sites. These systems enable a good understanding of particular events by using multiple news sites, but they ignore the characteristics of the various news sites. To characterize the differences between news sites, the News Site Contrast (NSContrast) system has been proposed. This system analyzes multiple news sites using the concept of contrast- set mining. However, NSContrast has only limited analysis functions and is not mature enough for evaluation via user experiments. Therefore, in this paper, we analyze contrast set mining results for NSContrast, aiming to understand the requirements for extracting useful information that also reflects the interests of different countries. Based on this analysis, a new NSContrast is introduced and applied to a news article database constructed from multiple news sites in different countries for user experimentation.