Detecting Contrast Patterns in Newspaper Articles by Combining Discourse Analysis and Text Mining

Text mining aims at constructing classification models and finding interesting patterns in large text collections. This paper investigates the utility of applying these techniques to media analysis, more specifically to support discourse analysis of news reports about the 2007 Kenyan elections and postelection crisis in local (Kenyan) and Western (British and US) newspapers. It illustrates how text mining methods can assist discourse analysis by finding contrast patterns which provide evidence for ideological differences between local and international press coverage. Our experiments indicate that most significant differences pertain to the interpretive frame of the news events: whereas the newspapers from the UK and the US focus on ethnicity in their coverage, the Kenyan press concentrates on sociopolitical aspects.