Working at the Interface of Hydrology and Corpus Linguistics

Representation of Drought Events in the United Kingdom: Contrasting 200 years of News Texts and Rainfall Records

“…If we accept that newsworthiness is closely associated with the occurrence of a drought and that while the drought lasts it will remain newsworthy, we can assume that years in which the word DROUGHT was salient are the years when the United Kingdom experienced a drought. While these assumptions have been found to be generally true (McEnery et al, 2019(McEnery et al, , 2021, the approach may still produce some false positives, meaning that qualitative analysis is still necessary to confirm or refute the trends indicated by the initial figures. Close reading of the data is therefore needed to confirm whether the reporting was, in fact, about the existence of a drought in the United Kingdom in that specific point in time.…”

“…Readers interested in a fuller description of the composition of each set and the methods and procedures required in the preparation of the texts should refer to: Baker et al (2020a) and McEnery et al (2019McEnery et al ( , 2021 for the 19th Century Corpus, Baker et al (2020b) for the 20th Century Corpus, and (Dayrell et al, 2020a(Dayrell et al, , 2020b for the Broadsheet and Tabloid Corpora. The 19th and the 20th Century Corpora were processed using CQPweb, Lancaster University's software platform for largecorpus analysis (Hardie, 2012).…”

“…Importantly for this paper, the availability of substantial corpora covering long periods of time is greatly increasing, with organizations such as the British Library providing access to machine readable text of newspapers from the past. 1 The present analysis builds on corpus analyses of the representation of drought in 19th century British newspapers in McEnery et al (2019McEnery et al ( , 2021, which contrasted evidence gathered from corpus material against the hydrological evidence provided by Cole and Marsh (2006) and Marsh et al (2007). McEnery et al (2019McEnery et al ( , 2021 demonstrate the effectiveness of the corpus linguistic method to identify drought events in the United Kingdom during the 19th century, when rainfall records were sparse and fragmentary.…”

“…1 The present analysis builds on corpus analyses of the representation of drought in 19th century British newspapers in McEnery et al (2019McEnery et al ( , 2021, which contrasted evidence gathered from corpus material against the hydrological evidence provided by Cole and Marsh (2006) and Marsh et al (2007). McEnery et al (2019McEnery et al ( , 2021 demonstrate the effectiveness of the corpus linguistic method to identify drought events in the United Kingdom during the 19th century, when rainfall records were sparse and fragmentary. While allowing us access to crucial information about past droughts, these analyses also identified droughts that had not been captured by hydrological approaches.…”

Corpus-Assisted Discourse Studies

“…• Type 1: a structured top-down process whereby the researcher applies an a priori framework or set of categories to a number of concordance lines (e.g., Culpeper and Gillings (2018) coding for politeness in the BNC1994/ 2014, and Lutzky (2021b) exploring the pragmatic functions of sorry in customer service interactions on Twitter); • Type 2: a structured bottom-up process whereby the researcher assigns categories to the concordance lines, but these come organically from the corpus rather than being imposed on it (e.g., Kopf (2019) exploring the ways that content policies are enforced on Wikipedia, and Zottola et al (2021) identifying coping strategies that patients use in autobiographical narratives while waiting for assessment at a transgender health clinic); • Type 3: an unstructured bottom-up process whereby the researcher eyeballs the concordance lines and lets that qualitative holistic judgement form the basis of analysis (e.g., McEnery, Baker and Dayrell (2019) identifying previously unrecorded droughts in nineteenth-century Britain, and Levon (2016) exploring the extent to which users on a question-and-answer forum use their replies as an opportunity for stance-taking); • Type 4: an unstructured top-down process whereby the researcher identifies concordance lines which match categories proven to be relevant in other datasets (e.g., Archer and Gillings (2020) identifying potential indicators of deception in Shakespeare's plays, and Appleton (2021) exploring how the unification of Germany is discussed in Hansard). Types 1 and 2 both call for the researcher to sift through each and every concordance line within a sample, but they differ with regard to whether categorisation is something that is applied to, or extracted from, the data.…”