This paper aims to conduct a linguistic analysis of discourses by three Korean politicians who ran for the Korean presidency in 2012. Data on their speech and writing are collected for this purpose from TV talk shows on which they appeared and the books they wrote. Various methods are used to analyse the data, such as network analysis, word frequency analysis and the lexical repetition measuring method. Firstly, we extract frequent words from each politician’s subcorpus and identify statistically distinctive words that each politician used more frequently than other politicians. Secondly, we construct networks of co-occurring words via which the differences in network structure are analysed. Finally, a new method for measuring lexical repetition is used to discover pragmatic differences in their discourses. By applying these methods to the discourse data, we can more effectively propose linguistic characteristics of the politicians’ discourses.