Text mining for pragmatic discourse analysis: Detecting contrast patterns in newspaper articles on Kenyan elections

Data mining aims at constructing classification models or finding interesting patterns in large data collections. This paper investigates the utility of applying data and text mining techniques to media analysis, more specifically, to the analysis of a corpus of articles covering the 2007 Kenyan elections and post-election crisis, attempting to capture the differences between local (Kenyan) and Western (British and US) newspaper articles. Methodologically this paper illustrates the effectiveness of using data and text mining techniques for pragmatic discourse analysis. Specifically, data and text mining methods which enable the analysis of contrasting patterns were used to uncover the differences between local and international news covering Kenyan elections. First, we formulated the task as an automated classification task. Next, we applied a method for semi-automated topic ontology construction and used methods for analyzing contrasting keywords which are interesting for qualitative pragmatic interpretation. Our experiments indicate that most significant differences pertain to the Western media's bias on ethnicity in their coverage of events, whereas the local media concentrate on sociopolitical aspects.

Jaar van publicatie:2011

Trefwoorden:Data mining, Kenyan elections, Text mining, Pragmatics, Media analysis, Discourse analysis