Navigation – Plan du site

AccueilNuméros86Carte BlancheCorpus Linguistics and Research S...

Carte Blanche

Corpus Linguistics and Research Syntheses: Respect for Data in Making Sense of Text and Content

La linguistique de corpus et les synthèses de recherche : priorité aux données pour la langue et le contenu
Alex Boulton
p. 71-82

Résumés

Cet article souligne l’importance de la rigueur et de la transparence méthodologiques pour collecter et interpréter les données de recherche, notamment en linguistique de corpus et pour les synthèses de recherche. Après un bref survol de chaque domaine, les deux sont ensuite réunis pour explorer comment une approche inspirée par les outils et techniques de la linguistique de corpus peut compléter d’autres types de synthèse à travers deux études de cas. La première vise l’anglais de spécialité avec deux HDR récentes, les deux privilégiant la fréquence de termes clés dans des articles de revues. La deuxième vise à pousser plus loin dans un domaine voisin, l’apprentissage sur corpus (ASC), avec une analyse d’un corpus d’articles de recherche exploitant en particulier des listes de fréquence de mots et de ‘n-grams’, des concordances et des mots clés et n-grams clés afin de répondre à des questions de recherche précises.

Haut de page

Texte intégral

1 Introduction

  • 1 Sir Arthur Conan Doyle. 1892. A scandal in Bohemia. In The Adventures of Sherlock Holmes.
  • 2 Douglas Adams. 1984. So long and thanks for all the fish.

1Research is an essentially human enterprise. The so-called scientific method is an attempt to reduce the foibles of human cognition by formalising the process of making sense of the world around us (see Pinker 2021). Central to this is the importance of data, with rigour and transparency in collecting and analysing it. Such is a commonplace of our everyday experience: “It is a capital mistake to theorise before one has data. Insensibly one begins to twist facts to suit theories, instead of theories to suit facts.”1 This suggests that data comes first, but data must also come from solid grounds: “Once you know what it is you want to be true, instinct is a very useful device for enabling you to know what it is.”2 We need good theory and reasoning to produce good data, and good data to provide evidence to refine theory.

2This paper explores two strands of research in applied linguistics with data at the heart: corpus linguistics and research synthesis, outlining how the former can be used to pursue the latter. Nonetheless, research remains human, and researchers are inevitably “telling stories” about their data, whether qualitative or quantitative. Indeed, all research must involve elements of both, as lucidly pointed out by Fillmore (1992, 35). Quantitative evidence (as personified by the corpus linguist) may not be interesting; it has first to be interpreted, turning data into information into knowledge, which involves an element of qualitative reasoning. Conversely, qualitative evidence (from the armchair linguist) may be interesting but not true, or at least not revealing; ‘an example’ may be true, but unless it can be quantified (is it an isolated example or frequent enough to be representative of something?), then it frustratingly tells us very little of use. In both cases, the crucial point is that researchers’ stories should be informed stories, stories based solidly on data that has been rigorously and transparently collected, reported and interpreted. Following a brief summary of the two, they are then brought together via case studies adopting a corpus approach to research synthesis, first in ESP, then in Data-Driven Learning (DDL).

2 Corpus linguistics

3Corpus linguistics has a long history, but in its modern form stems largely from work that came to the fore in the late twentieth century simultaneously in several areas around the world. For English, work in America (Kučera and Francis 1967) gave rise to the Brown family, a series of one-million-word corpora of English still used today. John Sinclair (1987) was a key driving force in the UK, creating the COBUILD corpus to provide dictionaries, grammar books and other resources aimed largely at language learners. These were initially based on a subset of about seven million words, though the ‘monitor corpus’ model meant this was ever-expanding, unlike the static British National Corpus with 100 million words (Aston and Burnard 1998). Technology had difficulty keeping up with the increasing size of corpora: the lexicographers at COBUILD worked on microfiches and printouts (Renouf 1987), while a BNC query took several minutes or a lunch break to complete (Aston 1996). The Corpus of Contemporary American English (COCA, Davies 2009) has increased from its original 385 million to over a billion words at the time of writing, made possible by the development of the internet as a source of readily available electronic texts. This is seen clearly with the WaCky corpora (Web-As-Corpus Kool Ynitiative) of a billion words each in several languages (Baroni and Bernardini 2006), and the TenTen family which derives its name from the minimum of 1010 (ten billion) words they each contain (Jakubíček et al. 2013). Today, the large language models (LLMs) behind generative artificial intelligence (GenAI) are based on hundreds of billions of words – ChatGPT version 3.5 released in November 2022 is more usually expressed as forty-five terabytes of data – and are continually expanding.

4Large recent corpora tend to use algorithms to collect texts automatically or semi-automatically, meaning that they are less ‘clean’ than smaller early corpora, and their composition less carefully controlled and often unknown – the choice is between size and rigour, each with its own advantages. Larger ones may therefore not entirely correspond to classic definitions of a corpus as being “sampled to be representative of a particular language or language variety” (McEnery et al. 2006, 5). By contrast, smaller ones with a very narrow focus, a cornerstone of much ESP, may claim only to be representative of a tiny subset of a language. The most important point may be that in both cases they represent substantial quantities of text in electronic format where regular reading can usefully be supplemented or replaced by the use of corpus tools and techniques, whatever that text may be – a single long text such as a novel is amenable to a corpus linguistics approach as it can be searched quickly, precisely and repeatedly for specific language features, which may otherwise be prohibitively time-consuming. The real advantage of corpus linguistics is its focus on actual language data, with rigour and transparency of collection and analysis, rather than intuitions of experts or serendipitous collections of examples. It allows us to see not what is possible, but what is probable and usual in different contexts. This reflects usage-based approaches to language, with the emphasis on patterns rather than rules.

  • 3 There is in fact just one occurrence of this phrase in proximity to the 465 occurrences of Figure x(...)

5The traditional tool for exploring a corpus is called a concordancer, which may include frequency lists for words or n-grams (sequences of n words), plots (to show distribution of an item within a corpus), collocates (words that co-occur frequently), and keywords (items that occur significantly more frequently in one corpus than another). An example of a standard concordance presentation (in KWIC format – key word in context) is given in Figure 1, where searching for a word or phrase (actually a string of characters) produces a list of all occurrences centred in a short context. This enables ‘vertical’ reading to see the patterns that occur left and right. Here, a semi-random selection of fifteen lines for “Figure x” from a corpus of nearly 200 research articles in Instructed Second Language Acquisition (ISLA) reveals a number of patterns: the search term is often in brackets, when it is typically preceded by “see”; in other cases, it is often followed by a verb such as “shows”, “displays”, “illustrates”, “depicts”. The absence of evidence is notoriously not evidence of absence, but if the user notes the lack of “we can see”, this can be explored further in the corpus.3 These features may help novice writers of academic English in ISLA.

Figure 1: KWIC concordance for “Figure x”.

6As far back as 1992, Rundell and Stock talked of a “corpus revolution” in linguistics, especially in fields such as lexicography and in creating resources for language learning and teaching. Many of these are “static” materials (like dictionaries), but users can also access corpora in what has come to be known as DDL, a term coined by Johns in 1990. However, corpora of various types can be used for a tremendous variety of purposes: “whether you are a student of language or literature, critical theory or history, or if you are studying for a degree in the social sciences, you will probably be working with electronic texts at some point” (Adolphs 2006, 11). In applied linguistics, this can be for the study of language per se, but also for content, in analysing any type of open-ended language data – input or output in learning, questionnaires or interview transcripts, searching for tendencies and biases, and so on. Here we go beyond language and into “aboutness” (Scott and Tribble 2006, chapter 4).

3 Research synthesis

7If the development of corpus linguistics reflects a growing awareness of the importance of data rather than expert intuition, the same applies to synthesising research in the field of applied linguistics. Huge strides have been made here too, notably since the work of Norris and Ortega (e.g. 2000). Chong and Plonsky (2024) provide a useful overview of the many different types of research synthesis, making two main distinctions. The first is between the relatively ad hoc, subjective narrative synthesis, akin to the literature review typically found in the introduction to research articles. They call this “traditional (non-systematic)” as compared to the “systematic (research synthesis)” (Chong and Plonsky 2024, 1571), where the field is clearly defined, the collection and conditions for inclusion are transparent, and the methodology for analysis of the collection spelt out in advance to respond to specific research questions.

8The second distinction is between qualitative and quantitative syntheses, each having its role to play. The former aims at a deep understanding of the research designs and findings in order to identify and interpret overall trends, similarities and differences in results. This is often based on a coding sheet which does, inevitably, entail an element of quantification. On the other hand, meta-analyses combine quantitative data from the studies included, pooled together as a single effect size (see Boulton and Cobb 2017, for an example in DDL). Effect size is different from statistical significance since, as the name suggests, it is not interested in significance but in the size of an effect. In second language acquisition, for example, this typically answers questions such as how big is the effect of teaching between a pre- and post-test following a particularly form of instruction, or between a control and experimental class (within- and between-groups designs respectively). The pooled data can also be broken down into moderator variables, i.e. categories decided by the synthesist to compare, for example, speaking face-to-face or online. All of this also necessarily involves an element of qualitative interpretation. In addition to exploring research to date, both qualitative and quantitative syntheses enable suggestions for future directions in research topics and practices.

  • 4 Glenn Stockwell, editor: personal communication.

9Such work in applied linguistics is indicative of a greater “synthetic-mindedness” overall, as Plonsky (2023, 9) calls it, though one unfortunate side effect has been to encourage some to see syntheses as the application of a set formula. Journal editors have been inundated with such mechanical syntheses, with the same authors serially submitting near-identical papers but on different topics, often within months of each other. The obvious inference is that they cannot be truly expert in the fields covered, lacking an understanding of the very thing they are writing about. For this reason, journals such as Computer Assisted Language Learning no longer accept surveys or syntheses.4

10One obvious question today is: why not use GenAI to conduct syntheses? This will no doubt feature more and more prominently as the technology improves and as programs appear with this specific aim in mind. At the present time, however, GenAI should be used carefully, treated with caution, and the results always checked. GenAI is susceptible to many kinds of errors, of which hallucination is probably the most widely cited, but transparency is a key issue at this stage of its development. To take one example, Udaya and Ramamuni Reddy (2024) recently published “a machine-generated literature review” of “vocabulary, corpus and language teaching” (the two halves of their title, not in that order). The book is not a synthesis as such, but a summary of published research papers, divided into four sections. The problems begin with the writing, with very short introductions (there are two) and an even shorter conclusion, which give no definitions or research questions, no criteria for selecting the studies or devising the groupings, and so on. This lack of transparency continues with the algorithms used to summarise the studies, dividing each into sections and identifying “the most important sentences” (Udaya and Ramamuni Reddy 2024, 7) which are presented verbatim, with no actual summarising by machine or authors. One of the crucial points of research synthesis is transparency, which depends first upon the authors in framing the study, and second on the procedures used; with little or no information on either, we should be sceptical of the results.

  • 5 Plonsky’s homepage lists 844 at the time of writing (July 2024): https://lukeplonsky.wordpress.com/ (...)

11Chong and Plonsky’s (2024) typology lays out thirteen different categories of research synthesis, most of which, as seen above, involve both qualitative and quantitative elements. Their analysis was empirically motivated since it was based on their own collection of synthesis-related papers in applied linguistics, clearly reported.5 Intriguingly, no overt mention is made of corpus-based syntheses, which are a relatively recent development. The following two sections look at some recent corpus-based research syntheses, first in relation to English for specific purposes (ESP), then expanding to the related field of DDL.

4 Corpus-based syntheses in English for specific purposes (ESP)

12ESP today relies heavily on corpus linguistics tools and techniques in exploring small, specialised corpora, but not often to synthesise the field itself. For this, we turn to two recent HDRs. The HDR (habilitation à diriger des recherches) is a postdoctoral diploma allowing candidates in France to supervise PhDs and apply for full professorship, among other things. The written texts include an extensive CV, a volume of the candidate’s research publications accompanied by a lengthy synthesis of the same (100-plus pages), contextualising it in relation to the specific research area, and finally the manuscript for a complete book. It is perhaps not coincidental that my own HDR (Boulton 2012) featured a corpus-based review of my previous publications, furthering my interest in research synthesis.

13The first HDR (Millot 2023) looked at discourse analysis, the second (Sarré 2024) at pedagogical aspects of ESP. Both authors had previously compiled and analysed corpora in their respective fields, presented in detail in their respective HDRs, for linguistic analysis and to inform their teaching practices. As researchers in English for specific purposes in France, they both work between two cultures – English (ESP) and French (ASP – anglais de spécialité). Millot (in press, Chapter 1) was interested in the use of corpora in the two languages, and simply searched for corpus and corpora in research articles in three journals: one (ASp) which publishes in French and in English, and two which are English only (ESP and JEAP – the Journal of English for Academic Purposes). He divided the collection into three decades, finding, in particular, that in the final period (2010-19), 47% of all papers contained one of the target words in the title, abstract or keywords: 47% in ASp, 48% in ESP, and 46% in JEAP. This apparently simple procedure provides several insights, among which: (a) in these journals at least, there is very little difference, quantitatively speaking, between English and French research cultures in this domain; (b) corpus linguistics holds a key position throughout, though the definition of a “corpus” may vary substantially. It is of note that this analysis is almost an afterthought (“anécdotique”, Millot 2023, 37) in the conclusion to the first chapter, but the basic idea is taken up and pursued further in the next HDR.

  • 6 The book (Sarré, in preperation) is structured as a series of idées reçues in the field; this is #1 (...)
  • 7La volonté de se démarquer de l’approche internationale anglophone de l’ESP.” “L’approche dite fra (...)
  • 8 RPPLS: Recherche et Pratiques Pédagogiques en Langue de Spécialité; RDLC: Recherches en Didactique (...)

14Sarré makes more extensive use of such comparisons in several parts of his HDR, exploring in particular differences between ESP and ASP (2024a, section 2.3; in preparation, section 1.3) which, despite the similarity in their names, are often claimed to cover rather different ground. Typically, work in French cultures is seen as more linguistically oriented, while internationally it may have a more overt pedagogical aim, though with some overlap. To test this “idée reçue6, Sarré adopts a corpus approach to bibliometric analysis for two journals, ASp and ESP Journal – again, it is the research culture (France or Francophonie vs the rest of the world) rather than the language of publication which is important. The analysis is taken to support his initial position that the differences have been exaggerated, reflecting “a desire to stand apart from the international ESP approach (in preparation, 23), a “deliberate scientific positioning rather than reality” (p. 25).7 This respect for data can also be found in a separate analysis comparing several French journals in ESP (ASp and RPPLSP) against language teaching and learning (RDLC and Lidil).8 A number of key terms listed by other authors are searched for in these journals to compare relative frequencies. ESP seems to feature more papers mentioning “needs analysis” and “corrective feedback”, while pedagogical journals favour “identity” and “autonomy”, among other things. Taken together, such findings are suggestive of certain differences but not of clearly opposing preoccupations.

15In both these studies, the corpus analysis is limited to the number of papers in different journals that feature the search terms. This is certainly a step towards a corpus approach, but the authors do not exploit the full possibilities of corpus linguistics. For this, we turn to DDL.

5 Corpus-based syntheses of data-driven learning

16DDL can be defined as the use of corpus tools and techniques for learning or using a foreign or second language (L2), first attracting attention over 30 years ago (Johns 1990). At the time, the classroom was virtually the only way to learn foreign (as opposed to second) languages, complemented occasionally by language resource centres, newspapers or books, penfriends or the occasional trip abroad, and other individual initiatives. Since then, the arrival of the internet and, more recently, GenAI, have opened up new opportunities for learning by contact with language, but the underlying principles still hold: a shift of emphasis away from “being taught” and towards allowing learners to discover the language for themselves, especially using software to examine large quantities of text, just as with a concordancer and corpus. As the field has expanded, over twenty syntheses have appeared to help make sense of more than 800 empirical research studies, from the narrative (qualitative) to the meta-analytic (quantitative) to mixed methods, along with various other types such as bibliometric or scientometric reviews (Boulton et al., in press). Because this area intimately involves corpus linguistics, it is perhaps not surprising that it has given rise to several corpus-based syntheses which go further than the two HDRs outlined above. The rest of this section focuses on two such studies I have co-authored, though there are others, notably the multiword keyword analysis in Pérez-Paredes (2022).

17The two studies by Boulton and Vyatkina (2021; 2024a) both make a point of transparency in defining the target dataset, formulating research questions, and detailing collection procedures and inclusion criteria, as well as the analyses to be conducted. The first is wider-ranging and includes empirical DDL publications of different types (mainly journals, chapters, conference proceedings; n=489), while the latter is limited first to studies with English as the L2, and second to research articles (RAs) published in journals ranked by the Web of Science (n=148). In general, a wider spread is desirable; however, prestige journals tend to have more rigorous selection procedures, typically double-blind review. While this is no simple proxy for “quality” (Plonsky 2024), especially with many inspiring papers appearing in other sources, it does have the merit of making the dataset more manageable, promising a near-exhaustive collection within its parameters and, by definition, reflecting the most visible research available. Once collected, the papers were all read and coded, which constitutes the main part of the subsequent synthesis for different time periods. For example, in Boulton and Vyatkina (2024a), this enabled the identification of trends towards an increase in studies in English for Academic Purposes (EAP), from 25% to 45% of empirical studies in the last five years (2018-2022) compared to earlier periods, contrasting with a drop in ESP studies from 10% to just 5%. The focal point for present purposes though is that, in addition to coding, both studies also involved a corpus analysis to answer some of the research questions.

18The corpus element involved the same procedures in both cases: the original documents were automatically converted to .txt and the results cleaned of conversion errors (for ligatures, hyphenation, diacritics, etc.), though errors in the originals were not touched. To concentrate on the text proper, all extraneous elements were set aside – metadata (affiliations, contact details, acknowledgements, etc.), headers and footers, figures and tables, primary and secondary data extracts (with the exception of academic quotations), ethical statements, references and appendices. The corpora were analysed using AntConc (Anthony 2024).

19The aim of the corpus analysis in Boulton and Vyatkina (2021) was to identify the researchers’ recommendations for future directions, for which the Conclusions sections (253,569 tokens) were compared against the rest of the corpus (2,563,589). Analysis of keywords and key 3-4-grams identified items that were statistically significantly more frequent in the Conclusions sections of at least three publications. Some were not meaningful for our purposes (e.g. “conclusion”, “limitations”), the others were analysed using the KWIC/Concordance tool to explore their uses in context. The resulting relevant key items (11 keywords, 41 key n-grams) were grouped thematically and, also, compared over time to see whether the recommendations had in fact been taken up in later periods. For example, “theory” was a Conclusions section keyword, with authors recommending greater consideration of theoretical underpinnings, prompting the authors to explore the different theories over time.

20A similar approach was adopted in the second paper (Boulton and Vyatkina 2024a), here focusing on methodological aspects of visible DDL research with English as the target language. This involved looking in particular at the Methodology sections (252,326 tokens) of the 148 RAs included, lemmatising the corpus (so that e.g. “student” and “students” would be grouped together). In the keyword analysis, a stoplist was applied to eliminate grammar-function words (e.g. “the”, “of”, “and”) which are highly frequent and, while of course essential in language studies, do not carry lexical meaning helpful for our type of analysis. As before, lists of key n-grams were also generated. Taken together, these highlight not just the obvious (e.g. changes in tools and technologies) but also a greater focus on research design and methodology, with the appearance of many key items such as “the comparison” (group), “by the researchers”, “of the target” (word, etc.), “before and after the” (instruction, etc.), “in the pre- and/post-” (tests). This does not mean that such items have appeared from nowhere, but they are substantially more frequent in the last five years compared to earlier periods. There are signs then that research methodologies are indeed evolving in DDL studies, a fact that may not have arisen from more traditional analysis.

21The same authors (Boulton and Vyatkina 2024b) are currently pursuing this corpus-based approach to synthesising DDL studies by compiling a comparison corpus of similar studies taken from the same journals and years and even the same issue, as long as they represent empirical ISLA studies other than DDL. The key words for DDL serve to confirm previous syntheses (i.e. what DDL is), but more revealing are the keywords in the other papers, highlighting what DDL is not. The top twenty-five keywords in the comparison corpus include “interaction”, “social”, “communication”, “face (-to-face)”, “collaborative”, “peer”, “negotiation”, etc. The question then is whether to push the (language) strengths of DDL, or to attempt a reconciliation – and what would “communicative DDL” look like? (Hirata and Thompson 2022).

22The objective here has been to outline the methodologies and opportunities afforded by corpus tools and techniques rather than the specific results, for which the reader is referred to the original publications. Though the complete corpus cannot be shared for copyright reasons, the full coding sheets are available as supplementary materials to the respective papers. In the studies outlined, the corpus analysis was not the sole methodology employed, but served rather to confirm and extend results obtained from the coding, clarifying them and adding depth and nuance.

6 Conclusions

23This paper has aimed to show the importance of respect for empirical data, involving a systematic and transparent approach upstream (in defining the field, deciding the research questions) and in collecting, analysing and interpreting the data. These features come together in corpus linguistics, whether for the study of language itself or, through language, to content and information, thus permeating virtually all fields of linguistics and text studies. In the particular area of research synthesis, corpus tools and techniques can enable the researcher to nuance and add depth to purely quantitative analysis, as well as adding a quantitative aspect to qualitative analysis. As such, it is typically used to complement other types of synthesis, though it can also enable the analyst to spot items that would otherwise have gone unnoticed.

Haut de page

Bibliographie

Adolphs, Svenja. 2006. Introducing Electronic Text Analysis: A Practical Guide for Language and Literary Studies. London: Routledge.

Anthony, Laurence. 2024. AntConc (Version 4.3.1). Tokyo, Japan: Waseda University. https://www.laurenceanthony.net/software

Aston, Guy. 1996. “The British National Corpus as a language learner resource.” In Proceedings of TALC 1996, edited by Simon Botley, Julia Glass, Anthony McEnery and Andrew Wilson, 178–191. UCREL Technical Papers, 9.

Aston, Guy, and Lou Burnard. 1998. The BNC Handbook: Exploring the British National Corpus. Edinburgh: Edinburgh University Press.

Baroni, Marco, and Sylvia Bernardini, eds. 2006. Wacky! Working Papers on the Web as Corpus. Bologna: Gedit.

Boulton, Alex. 2012. De la didactique aux corpus. Le Havre : Université du Havre.

Boulton, Alex, and Tom Cobb. 2017. “Corpus Use in Language Learning: A Meta-Analysis.” Language Learning 67 (2): 348–393. https://doi.org/10.1111/lang.12224

Boulton, Alex, and Nina Vyatkina. 2021. “Thirty Years of Data-Driven Learning: Taking Stock and Charting New Directions.” Language Learning & Technology 25 (3): 66–89. https://doi.org/10125/73450

Boulton, Alex, and Nina Vyatkina. 2024a. “Expanding Methodological Approaches in DDL Research.” TESOL Quarterly 58 (3): 1193–1204. https://doi.org/10.1002/tesq.3269

Boulton, Alex, and Nina Vyatkina. 2024b. “Exploring Data-Driven Learning with a Corpus of Research Articles: What it Is… and What it Isn’t.” 16th Teaching & Language Corpora International Conference. Manchester: University of Manchester, 7–10th July.

Boulton, Alex, Nina Vyatkina, and Tom Cobb. In press. “Classroom Applications of Corpus Analysis.” In The Cambridge Handbook of English Corpus Linguistics (2nd edition), edited by Douglas Biber and Randi Reppen. Cambridge: Cambridge University Press.

Chong, Sin Wang, and Luke Plonsky. 2024. “A Typology of Secondary Research in Applied Linguistics.” Applied Linguistics Review 15 (4): 1569–1594. https://doi.org/10.1515/applirev-2022-0189

Davies, Mark. 2009. “The 385+ Million Word Corpus of Contemporary American English (1990-2008+): Design, Architecture, and Linguistic Insights.” International Journal of Corpus Linguistics 14 (2): 159–188. https://doi.org/10.1075/ijcl.14.2.02dav

Fillmore, Charles. 1992. “Corpus Linguistics or Computer-Aided Armchair Linguistics.” In Directions in Corpus Linguistics, edited by Jan Svartik, 35–60. Berlin: Mouton de Gruyter.

Hirata, Yoko, and Paul Thompson. 2022. “Communicative Data-Driven Learning: A Two-Year Pilot Study.” ELT Journal 76 (3): 356–366. https://doi.org/10.1093/elt/ccab066

Jakubíček, Miloš, Adam Kilgarriff, Vojtěch Kovář, Pavel Rychly, and Vit Suchomel. 2013. “The TenTen Corpus Family.” In Proceedings of the 7th International Corpus Linguistics Conference, 125–127. Lancaster: UCREL.

Johns, Tim. 1990. “From Printout to Handout: Grammar and Vocabulary Teaching in the Context of Data-Driven Learning.” CALL Austria 10: 14–34.

Kučera, Henry, and W. Nelson Francis. 1967. Computational Analysis of Present-Day American English. Providence, RI: Brown University Press.

McEnery, Tony, Richard Xiao, and Yukio Tono. 2006. Corpus-Based Language Studies: An Advanced Resource Book. London: Routledge.

Millot, Philippe. 2023. Approche empirique des discours spécialisés en contexte anglophone. Lyon: Université Lumière Lyon 2. (Manuscrit d’HDR non publié)

Millot, Philippe. In press. Anglais de spécialité, discipline empirique. Limoges : Lambert-Lucas.

Norris, John M., and Lourdes Ortega. 2000. “Effectiveness of L2 Instruction: A Research Synthesis and Quantitative Meta-Analysis.” Language Learning 50: 417–528. https://doi.org/10.1111/0023-8333.00136

Pérez-Paredes, Pascual. 2022. “A Systematic Review of the Uses and Spread of Corpora and Data-Driven Learning in CALL Research during 2011–2015.” Computer Assisted Language Learning 35 (1–2): 36–61. https://doi.org/10.1080/09588221.2019.1667832

Pinker, Steven. 2021. Rationality: What It Is, Why It Seems Scarce, Why It Matters. Penguin.

Plonsky, Luke. 2023. “Sampling and Generalizability in Lx Research: A Second-Order Synthesis.” Languages 8 (1): 1-13. https://doi.org/10.3390/languages8010075

Plonsky Luke. 2024. “Study Quality as an Intellectual and Ethical Imperative: A Proposed Framework.” Annual Review of Applied Linguistics. Advance Open Access. https://doi.org/10.1017/S0267190524000059

Renouf, Antoinette. 1987. “Corpus Development.” In Looking Up: An Account of the COBUILD Project in Lexical Computing, edited by In John McHardy Sinclair, 1–40. London: Collins.

Rundell, Michael, and Penny Stock. 1992. “The Corpus Revolution.” English Today 8 (3): 21–32. https://doi.org/10.1017/S0266078400006520

Sarré, Cédric. 2024. Contribution à l’épistémologie de la didactique de l’anglais de spécialité : l’interaction comme élément structurant d’une dynamique didactique. Paris : Sorbonne Université. (Manuscrit d’HDR non publié).

Sarré, Cédric. In preparation. Approche didactique de l’anglais de spécialité : de la didactique à la pédagogie, ou comment faire dialoguer recherche et pratique.

Scott, Mike, and Chris Tribble. 2006. Textual Patterns: Key Words and Corpus Analysis in Language Education. Amsterdam: John Benjamins. https://doi.org/10.1075/scl.22

Sinclair, John McHardy, ed. 1987. Looking Up: An Account of the COBUILD Project in Lexical Computing. London: Collins.

Udaya, Muthlaya, and Chada Ramamuni Reddy. 2024. Vocabulary, Corpus and Language Teaching: A Machine-Generated Literature Overview. Cham: Springer. https://doi.org/10.1007/978-3-031-45986-3

Haut de page

Notes

1 Sir Arthur Conan Doyle. 1892. A scandal in Bohemia. In The Adventures of Sherlock Holmes.

2 Douglas Adams. 1984. So long and thanks for all the fish.

3 There is in fact just one occurrence of this phrase in proximity to the 465 occurrences of Figure x in this 1.2-million-word corpus.

4 Glenn Stockwell, editor: personal communication.

5 Plonsky’s homepage lists 844 at the time of writing (July 2024): https://lukeplonsky.wordpress.com/bibliographies/meta-analysis/

6 The book (Sarré, in preperation) is structured as a series of idées reçues in the field; this is #1.3.

7La volonté de se démarquer de l’approche internationale anglophone de l’ESP.” “L’approche dite française de l’ASP nous semble ainsi plus relever d’une posture scientifique stratégique que d’une réalité tangible.

8 RPPLS: Recherche et Pratiques Pédagogiques en Langue de Spécialité; RDLC: Recherches en Didactique des Langues et des Cultures; Lidil: Revue de Linguistique et de Didactique des Langues.

Haut de page

Table des illustrations

Légende Figure 1: KWIC concordance for “Figure x”.
URL http://journals.openedition.org/asp/docannexe/image/9052/img-1.png
Fichier image/png, 123k
Haut de page

Pour citer cet article

Référence papier

Alex Boulton, « Corpus Linguistics and Research Syntheses: Respect for Data in Making Sense of Text and Content »ASp, 86 | 2024, 71-82.

Référence électronique

Alex Boulton, « Corpus Linguistics and Research Syntheses: Respect for Data in Making Sense of Text and Content »ASp [En ligne], 86 | 2024, mis en ligne le 18 novembre 2024, consulté le 05 décembre 2025. URL : http://journals.openedition.org/asp/9052 ; DOI : https://doi.org/10.4000/12ry1

Haut de page

Auteur

Alex Boulton

Alex Boulton is Professor of English and Applied Linguistics and former director of the ATILF research group (CNRS & Université de Lorraine). He is editor of ReCALL, and is on committees for several other scientific journals including Alsic, ASp, CALL-EJ, IJCALLT, and Language Learning & Technology and associations (EUROCALL, TaLC and AFLA). Particular research interests centre on corpus linguistics and potential uses for ‘ordinary’ teachers and learners (aka data-driven learning), with numerous publications in this area including several syntheses of DDL in recent years.

Articles du même auteur

Haut de page

Droits d’auteur

CC-BY-NC-ND-4.0

Le texte seul est utilisable sous licence CC BY-NC-ND 4.0. Les autres éléments (illustrations, fichiers annexes importés) sont susceptibles d’être soumis à des autorisations d’usage spécifiques.

Haut de page
Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search