105) elaborates on previous definitions by adding that, in corpora, the texts are ‘stored on a computer, sometimes analysed automatically or semi-automatically’. The digitalisation of texts makes it easier for researchers to share their material and gain access to larger amounts of data, thus fostering quantitative significance of the results – the more occurrences of a norm may be observed, the more likely that norm is to be representative of the language explored.
Traditionally, working on Corpus Linguistics (CL) has meant to resort to already existing large corpora, of both written and spoken language. However, it may be argued that CL has evolved into some kind of methological tool or research methodology for many areas of knowledge within Linguistics, and even for other disciplines (Olohan 2004; Walsh et al. 2011; Laviosa 2013). In this paradigm, researchers either use the large corpora available or compile ad hoc–designed, often smaller, corpora that better fit their specific research purposes and interests. In this respect, the selection of texts in accordance with research criteria is an essential feature that distinguishes corpora from ‘more random collections of texts held in texts archives’ (Barnbrook 1996: 23).
Within the field of Linguistics, a paradigmatic shift in relation to corpora perception took place in the 2000s, when some researchers started to argue for the need to undertake corpus-driven studies to the detriment of corpus-based studies, which had been the norm until then. In this spirit, Tognini-Bonelli (2001: 84–85) claims that:
Corpus-driven linguistics rejects the characterisation of corpus linguistics as a method and claims instead that the corpus itself should be the sole source of our hypotheses about language. It is thus claimed that the corpus itself embodies a theory of language.
Corpas-Pastor (2008: 53, my translation) offers a table that succinctly compares the main differences between both approaches (Table 1).
| Corpus-based approaches | Corpus-driven approaches |
| A priori theoretical assumptions | A posteriori theoretical assumptions |
| Inductive method | Deductive method |
| Intuition and introspection | Statistical methods |
| Lemmatised, annotated corpora | Non-codified, large and representative corpora |
| Grammar patterns | Lexicogrammatical and phraseological patterns |
Corpus-driven studies confer more power to data, to the text in hand, than corpus-based studies do. The latter use corpora to answer a preconcibed research question (e.g. ‘How often does nominalisation occur in this text?’), while in corpus-driven studies emphasis is placed on the text rather than on the research questions, that is, it is the text that leads the researcher to specific questions. If results are corpus-driven, it means that ←14 | 15→researchers have first approached the text trying to detach themselves from any assumptions derived from any previous specialised literature. Given their non-prejudicial nature, corpus-driven studies have been compared with a ‘tabula rasa’ (Corpas-Pastor 2008: 52).
Hanks’s (2004, 2013a) ongoing lexicographic project, the Pattern Dictionary of English Verbs (PDEV), adopts a corpus-driven approach to lexicon to tease out linguistic norms and the entries correspond to the different semantic patterns of each word, manually annotated from corpora. The scholar has proposed the Theory of Norms and Exploitations (TNE) together with a specific work methodology in CL, namely Corpus Pattern Analysis (CPA), which have served as a framework for the qualitative study of lexicon in this book (Chapter 8). As specified in Jezek and Hanks (2010: 8), this type of lexicographic task consists of ‘us[ing] corpus evidence to tease out the different patterns of use associated with each word in a language and to discover the relationship between meaning and patterns of usage’. Thus, the perception of norms in TNE intersects with the use given to norm in quantitative linguistic research, to mean prototypical, average use of the language, from which creative realisations may deviate to a greater or lesser extent (Muller 1973, 1992).
Corpas-Pastor (2008: 218, my translation) explains that undertaking corpus-driven research does not mean to undertake objective research: ‘Former paradigms, reference frameworks, ideological positions, cultural traditions and key concepts, among others, are bound to influence ←15 | 16→the researcher’s perspective’. Manual annotation, as in the case of the PDEV, necessarily involves the annotator’s subjectivity (Church and Hanks 1990; Renau 2012). The point is, however, that in corpus-driven approaches, ‘intuitions are used to interpret data, but not to create it’ (Hanks 2004: 250). In other words, as posited by Sinclair (2004a: 6), ‘intuition can help in many ways in language research, in conjunction with other criteria of a more examinable nature’.
2.3. Norms in Translation Studies
The difficulty and subjectivity entailed in applying labels to corpora in accordance with semantic, syntactic or pragmatic criteria is strong evidence that languages are not fully measurable, mathematical entities, as emphasised by Pedersen (1997: 111): ‘[I];t seems to me an illusion that any student of the arts can limit himself to objective statements about his object of study’.
The author was probably referring here to scholars who, already from the 1960s but crucially in the 1990s, had advocated the implementation of scientific methods to the study of translations, to the detriment of a prescriptive approach to translation ‘geared primarily to formulating rules, norms or guidelines for the practice or evaluation of translation or to developing didactic instruments for translation training’ (Hermans 1999: 7). As summarised by Chaume (2013: 292), ‘DTS is also interested in how systems influence each other, or the relationship between the source and the target cultures’.
In this new paradigm, Toury (1995) advocated DTS as a means to provide the discipline of Translation with rigorous approaches to its object of study, that is, translations. In the framework of DTS, research accounts for existing translated material without emitting value judgements about it and in a systematic manner, supported by statistical measures. Because it is based on observable facts, it has also been regarded an empirical approach to translation, as opposed to theory of translation (theoretical) or ←16 | 17→applied translation studies (aprioristical) (Rabadán 1991; Hermans
