The peaks and troughs of corpus-based contextual analysis

Abstract

This paper focuses upon two issues. Firstly, the question of identifying diachronic trends, and more importantly significant outliers, in corpora which permit an investigation of a feature at many sampling points over time. Secondly, we consider how best to combine more qualitatively oriented approaches to corpus data with the type of trends that can be observed in a corpus using quantitative techniques. The work uses a recently completed ESRC-funded project as a case study, the representation of Islam in the UK press, in order to demonstrate the potential of the approach taken to establishing significant peaks in diachronic frequency development, and the fruitful interface that may be created between qualitative and quantitative techniques.