This chapter describes the process of analyzing a corpus. It discusses the general notions of quantitative and qualitative research methodologies within the context of various types of corpus analyses, including Biber’s work on multi-dimensional analysis as well as a more qualitative analysis of Donald Trump’s style of speaking termed Trump Speak. These analyses are used to illustrate the various stages of doing research in general, and corpus-based research specifically. For instance, the analysis of Trump Speak is linked to a discussion of “how to frame a research question"; “how to select relevant corpora that can be analyzed to pursue the research question"; “how to ‘extract’ grammatical and lexical information from a corpus,” using such tools as concordancing programs; and, finally, how to “integrate relevant linguistic theories into the analysis of corpus results,” so that the research does more than simply list, for instance, frequency counts. There is also a discussion of “mixed methods”: how to introduce quantitative and qualitative research methods into a corpus analysis, followed by an example corpus analysis employing descriptive statistics such as the chi square statistic, and additionally, a description of multi-dimensional analysis in the work of Douglas Biber.

The chapter addresses a problem of contrastive pragmatics: How can we study correspondences between pragmatic markers in two languages if one language has a class of elements that the other language lacks? Specifically, the contribution deals with modal particles of German (ja and doch) and their reflexes in English translations. As there is no predetermined set of potential English correspondences, traditional distributional analyses are not feasible, and methods from Natural Language Processing are explored instead. Using 32 types of n-grams, differing in length and type of annotation, three classification tasks are carried out, in order to identify cues in the English translations that reflect the presence (or absence) of a particle in the German original. The results show that lemma-unigrams and -bigrams are often most informative (i.e. most accurate), while trigrams and 1-skip-2-grams provide important information about concomitants of modal particles that unigrams and bigrams miss. The results show that linguistic observables (n-grams) as the basis of quantitative analyses need to be carefully selected and explored in terms of their contribution to linguistic analysis.

Chapter 3 provides a brief overview of the particular approach to corpus linguistics adopted in this book: namely, corpus linguistics as a method (as opposed to corpus linguistics as a theory) (McEnery & Hardy, 2012; Tognini-Bonelli, 2001). The specific corpus linguistic analytical methods used, such as collocation and keyness analysis, and the statistical tests and cut-offs associated with each analysis type are detailed. Using data from the MI 1984–2014 Corpus (specifically the data collected during a pilot study and an illness-specific sample of the data), each analytical method used is exemplified. The utility of each analysis type for analysing ideology in texts is also discussed.

While written language development involves reducing erroneous expressions, traditional error-based measures are problematic for several reasons, including low inter-coder reliability for lexical errors, limited sensitivity for capturing development within a short time period, and the questionable separation of lexical and grammatical errors. Given these problems, we explore automated accuracy measures rooted in a usage-based theory of Second Language Acquisition, which views language as a set of constructions or chunks. For this study, we examined 139 essays in terms of using traditional measures of complexity, accuracy, lexical sophistication, and fluency, as well as novel corpus-based n-gram measures. A factor analysis was conducted to explore how traditional measures grouped with corpus-based measures, and regression analyses were used to examine how corpus-based measures predicted error counts and holistic accuracy scores. With the results of these analyses, we suggest that automated n-gram based measures are a viable alternative to traditional accuracy measures.