Measuring the Influence from User-Generated Content to News via Cross-Dependence Topic Modeling

Abstract. Online news has become increasingly prevalent as it helps the public access timely information conveniently. Meanwhile, the rapid pro-liferation of Web 2.0 applications has enabled the public to freely express opinions and comments over news (user-generated content, or UGC for short), making the current Web a highly interactive platform. Gener-ally, a particular event often brings forth two correlated streams from news agencies and the public, and previous work mainly focuses on the topic evolution in single or multiple streams. Studying the inter-stream influence poses a new research challenge. In this paper, we study the mutual influence between news and UGC streams (especially the UGC-to-news direction) through a novel three-phase framework. In particular, we first propose a cross-dependence temporal topic model (CDTTM) for topic extraction, then employ a hybrid method to discover short and long term influence links across streams, and finally introduce four measures to quantify how the unique topics from one stream affect or influence the generation of the other stream (e.g. UGC to news). Extensive exper-iments are conducted on five actual news datasets from Sina, New York Times and Twitter, and the results demonstrate the effectiveness of the proposed methods. Furthermore, we observe that not only news triggers the generation of UGC, but also UGC conversely drives the news reports