Keyness

Common patterns of interactions are altered in the digital world and new patterns of communication have emerged, challenging previous notions of what communication actually is in the contemporary age. Online configurations of interaction, such as video chats, blogging, and social networking practices demand profound rethinking of the categories of linguistic analysis, given the blurring of traditional distinctions between oral and written discourse in digital texts. This volume reconsiders underlying linguistic and semiotic frameworks of analysis of spoken and written discourse in the light of the new paradigms of online communication, in keeping with a multimodal corpus linguistics theoretical framework.

Typical modes of online interaction encompass speech, writing, gesture, movement, gaze, and social distance. This is nothing new, but here Sindoni asserts that all these modes are integrated in unprecedented ways, enacting new interactional patterns and new systems of interpretation among web users. These "non verbal" modes have been sidelined by mainstream linguistics, whereas accounting for the complexity of new genres and making sense of their educational impact is high on this volume’ s agenda. Sindoni analyzes other new phenomena, ranging from the intimate sphere (i.e. video chats, personal blogs or journals on social networking websites) to the public arena (i.e. global-scale transmission of information and knowledge in public blogs or media-sharing communities), shedding light on the rapidly changing global web scenario.

Keyness analysis is perhaps the most widely used technique within corpus approaches to (critical) discourse studies. This chapter will first define the nature of keyness, and outline the research foci that keyness analysis can be usefully employed to address (e.g. not only differences, but also similarities). It will then provide a brief historical overview of the use of frequency comparisons in corpus linguistics, from manual comparisons of the frequency of particular linguistic items to the automated comparison of the frequency of all words in two corpora (or sub-corpora). The above will also act as the springboard for the discussion of the core aspects of keyness analysis, and associated common misconceptions.

1. The unit of analysis. Most studies select the word as the unit of analysis (as indicated by the common use of the term ‘keyword’). However, this need not be the default selection: frequency comparisons can focus on any lexical or morphosyntactic unit, or particular semantic or pragmatic meanings – as long as the appropriate corpus annotation has been carried out.
2. The selection of the corpora to be compared. The discussion will focus on issues relating to the nature of ‘study’ and ‘reference’ corpus, and their relative size. 
3. The issue of appropriate metrics. The discussion will focus on the difference between metrics of effect size and statistical significance, and the usefulness of combining the two types. It will also provide a comparison of the effect size metrics currently adopted by widely used corpus tools.
4. The selection of the key items that a study can usefully focus on. As an automated keyness analysis usually returns a much larger number of key items than is feasible to tackle, the approach to selection is of paramount importance, as it will determine the results and conclusions.

These issues, and the associated methodological decisions and their implications, will be discussed with reference to a case study.

Fandom is a growing phenomenon in the contemporary user-generated mediascape, inasmuch as it is capitalizing on the unprecedented possibilities of publication, distribution and interaction made available by digital technologies (Hellekson, Busse 2006; Stein, Busse 2012). In transmedia storytelling, integral elements of a story are dispersed systematically across various delivery channels with the goal of creating a networked entertainment experience (Jenkins 2006, 2007). Stories are no longer experienced through linear narratives, but they are accessed through several “points of entries” that encourage a customized reading/viewing/writing practice: the global fanfiction experience. Fanfiction is explored of this study from two distinct but complementary theoretical and methodological standpoints. Building on previous studies on the matter (Thomas 2010; Sindoni 2013) and drawing on a mono-generic corpus, LJFic, the research questions that this paper addresses deal with diatypic variation (Halliday 1991) in fanfiction from both linguistic and multimodal perspectives. Keyness analyses have been carried out using two different reference corpora (FLOB and COCA), assuming that such analyses can shed light on a range of linguistic issues, for example with regard to spoken/written variation (approximating Biber’s MF/MD analysis, 1988) and with a focus on the most prominent lexical items for the investigation of the entries’ aboutness (Scott, Tribble 2006; Bondi, Scott 2010). However, a purely computational analysis cannot account for the multimodal nature of fanfiction. To fill this gap, a sample of entries will be analysed qualitatively, by unearthing and unpacking the multimodal resources involved.

Keyness analysis is perhaps the most widely used technique within corpus approaches to (critical) discourse studies. As an automated keyness analysis usually returns a much larger number of key items than is feasible to examine manually within an appropriate co-text, the approach to selection of key items is of paramount importance, as it will determine the results and conclusions (Gabrielatos & Marchi, 2011). Currently, studies tend to adopt a methodologically naïve approach to selecting key items for manual analysis: they remove items from consideration before the automated analysis  by using frequency thresholds or stoplists and/or or select a small sub-set of items returned by the automated analysis (e.g. the top-N key items and/or key items that they deem relevant to the focus of the study) (see Pojanapunya & Watson Todd, 2016). However, the above approaches lack a principled rationale, and adopting them can remove important key items from consideration and lead to cherry-picking – consequently rendering results and conclusions questionable. Also, keyness studies predominantly focus on differences between the compared corpora, and there are very few studies using keyness analysis to examine similarities (Taylor, 2013). This paper will discuss a new approach to selecting key items in a principled fashion, and will demonstrate the relevant procedures via a case study. The approach utilises cluster analysis, and caters for a focus on both difference and similarity. However, in order to contextualise the proposed procedure, the paper will need to preface its main focus with addressing a number of relevant misconceptions regarding the nature of keyness, the selection of the corpora to be compared (usually referred to as the study and reference corpus), and appropriate metrics for establishing keyness.

""In this paper we examine the definitions of two widely-used interrelated constructs in corpus linguistics, keyness and keywords, as presented in the literature and corpus software manuals. In particular, we focus on
a. the consistency of definitions given in different sources;
b. the metrics used to calculate the level of keyness;
c. the compatibility between definitions and metrics.

Our survey of studies employing keyword analysis has indicated that the vast majority of studies examine a subset of keywords – almost always the top X number of keywords as ranked by the metric used. This renders the issue of the appropriate metric central to any study using keyword analysis.

In this study, we first argue that an appropriate, and therefore useful, metric for keyness needs to be fully consistent with the definition of keyword. We then use four sets of comparisons between corpora of different types and sizes, in order to test whether and to what extent the use of different metrics affects the ranking of keywords. More precisely, we look at the extent of overlap in the keyword rankings resulting from the adoption of different metrics, and we discuss the implications of ranking-based analysis adopting one metric or another. Finally, we propose a new metric for keyness, and demonstrate a simple way to calculate the metric, which supplements the keyword extraction in existing corpus software.""

This presentation provides a critical examination of the following: the nature of keyness, the types and foci of keyness analysis, possible units of a keyness analysis, the principled selection of corpora, appropriate metrics for establishing the level of keyness, principled procedures for selecting key items for further analysis. The above are also demonstrated via an example study.

The article reports research on the concept of key words as words statistically significant in a text or corpus. It reviews approaches to eliciting key words used in various software products for language analysis and the rationale for adopting them. Based on empirical data, a new method is proposed and tested on an exploratory corpus. The motivation and arguments for proposing the procedure are revealed, using comparisons between different languages. The adequacy of the results yielded by the different methods is tested via a mechanism developed with this research.

Length motifs (L-motifs) are defined as sequences of words whose lengths are monotonously increasing. In recent years, L-motifs have attracted well-deserved attention as they provide a new view of texts and their syntagmatic properties and nested structures. This study examines the key L-motifs, i.e. motifs that are overrepresented in texts and negative key L-motifs that are underrepresented in texts. The data reveal motifs that are typical for Czech texts, motifs that are typical for Arabic texts, and motifs that are typical for both Czech and Arabic texts – their existence suggests that there are new general language-independent patterns waiting to be explored.