BAAL CL SIG event 2026
Corpus linguistics and AI: methodological and pedagogical opportunities and limitations
Online symposium to take place on Monday December 14, 2026.
The symposium will examine the methodological and pedagogical possibilities and limits of using AI in corpus linguistic research and teaching. In particular, we hope to discuss what LLMs can and cannot contribute to corpus-based research, how they might be integrated into existing methods and tools, and how corpus linguistics might be taught as a research method in light of these developments.
Symposium Program
All times are UK time (GMT). Expand sections to view abstracts and speaker bios
In this talk, I will propose a new approach to corpus analysis in which existing corpus tools are connected to agentic AI systems through standardized protocols. This allows researchers to operate corpus tools using natural language, with the AI system orchestrating the interactions by selecting and running appropriate analyses in response to their requests. Importantly, the analyses themselves are still carried out using established corpus methods, with AI providing an alternative way to access and combine these methods rather than replacing them.
Drawing on recent developments in AntConc, I will demonstrate how this approach can be used to carry out a range of corpus analyses and how it allows corpus tools to become more naturally integrated into broader research workflows. I will also consider the strengths and limitations of this approach, particularly in relation to transparency, reproducibility, and the role of the researcher in evaluating and interpreting results.
Bio
Laurence Anthony is Professor of Applied Linguistics at the Faculty of Science and Engineering, Waseda University, Japan. His recent work explores the role of AI in language research and education, the integration of corpus methods and generative AI, and the development of new AI models and architectures. He is best known for developing the widely used corpus analysis toolkit AntConc, for which he received the National Prize of the Japan Association for English Corpus Studies (JAECS).
Recent developments in artificial intelligence (AI) in general, and Generative AI (GenAI) in particular, have brought about changes across the academy. In applied linguistics, a growing body of work is emerging dedicated to testing and evaluating the use of AI in a range of subfields, spanning language education, sociolinguistics, translation studies, corpus linguistics, and discourse studies, inter alia. This presentation explores the impact of AI on corpus linguistics, reflecting on the alignment of contemporary AI research with the epistemological, ontological, and ethical traditions of applied linguistics. Through this critical appraisal, I identify areas of misalignment regarding perspectives on knowing, being, and evaluating research practices. The question of alignment guides the discussion as I address the potential affordances of AI and GenAI for corpus linguistics as well as some of the challenges that we face when employing AI and GenAI as part of corpus linguistics research processes. The goal of this presentation is to attempt to align perspectives in these disparate fields and forge a fruitful way ahead for further critical interrogation and integration of AI and GenAI in(to) corpus linguistics.
Bio
Niall Curry is The Carnegie 125th Anniversary Fellow and an Associate Professor of Languages and Linguistics in the Department of Linguistics and Communication at the University of Birmingham (UK). He is also a UKRI Metascience AI Fellow. His work addresses issues in applied linguistics, with specific interest in artificial intelligence, corpus linguistics, contrastive linguistics, discourse analysis, and language pedagogy. He is Co-Editor of the Applied Corpus Linguistics and Corpus Linguistics Guides book series (both Routledge), Section Editor of the International Encyclopedia of Language and Linguistics (Elsevier), and Associate Editor of Language Learning and Technology. He is also a Fellow of The Royal Society of Arts and the Global China Academy.
In early 2025, I undertook a literature review of papers in CL using the then-newly fashionable Generative AI systems based on Large Language Models (LLMs). I found two trends. First, application of CL methods to studying LLM output and to diagnosing how it differs from human language. Second, studies using Generative AI in the ways promoted by its greatest enthusiasts – that is, to do the analysis for you. This second set was subdividable according to their respective authors’ overall attitude to Generative AI. I have dubbed the two attitudes, slightly facetiously, Enthusiastic Embracers (happy to adopt LLMs to replace their own effort) and Careful Critiquers (aiming to assess utility of LLMs, often in comparison to human analysis).
I recently brought that literature review effort up to date to late 2026. This presentation will outline what I found: to what extent Generative AI work has taken over the major CL journals; to what extent the earlier trends (LLM output analysis, Enthusiastic Embrace and Careful Critique) still fairly characterise these studies. My discussion will highlight particularly use of LLMs for qualitative analysis, i.e. interpretation, this being the most controversial use, since LLMs totally lack qualia. I will argue that this commits the classic error of mistaking the map for the territory, since LLM product is not actually language, but statistically modelled predictions of language.
Bio
Andrew Hardie is Reader in Linguistics at Lancaster University. His main research interests are the methodology of corpus linguistics; the descriptive and theoretical study of grammar using corpus data; the languages of Asia and Africa; and applications of corpus methods in humanities and social sciences. He is one of the lead developers of the Corpus Workbench software for indexing and analysing corpus data, and the creator of its online interface, CQPweb. He is co-author, with Tony McEnery, of the book Corpus Linguistics: Method, Theory and Practice (2012).
11:00 Break
Ever since ChatGPT 'was born', scientists more broadly and linguists more narrowly have been discussing how useful, and how risky/deleterious, generative AI in general and LLMs in particular will be to both (more theoretical) research and (more practical) applications. With regard to corpus linguistics in particular, I actually know of colleagues who basically seem to think that corpus linguistics might have to be completely redefined or might as well be abandoned given that whatever corpus linguists did gen AI can now do (better). I am not sure I would go quite that far and that's not because I want to 'keep my job' but because I have seen gen AI do amazing stuff, but also amazingly stupid stuff -- in other words, the results are mixed. in this (more programmatic and less empirical) talk I want to discuss three ways or, more humbly, ask three questions in which gen AI can relate to corpus-linguistic work: How useful and/or reliable is, or will gen AI be, when it comes to
- lower-resourced languages? Preliminary answer: "Not very much (yet)."
- how reliable or useful would gen AI be as a 'helper'/'research assistant', e.g. when it comes to annotating corpus data (such as concordances)? Preliminary answer: "The results are mixed."
I will report results from two alternation applications:
one regarding adjective comparison where gen AI was used for adjective lengths
another regarding the dative alternation where gen AI was used to annotate half a dozen characteristics of recipients and patients.
- determinants/predictors of alternation choices when the predictors are more 'exotic' and/or ones that go above and beyond things easily approached by letter sequences or sub-word tokens. As examples I will discuss predictors involving articulatory factors (e.g., rhythmic or segment alternation) and predictors related to alpha- and beta-persistence (i.e. priming) and predictors they interact with (e.g., surprisal and cumulativity).
Bio
Stefan Th. Gries is Distinguished Professor of Linguistics in the Department of Linguistics at the UC Santa Barbara (since 2005) and Chair of English Linguistics in the Department of English at the JLU Giessen (since 2018). Gries earned his M.A. and Ph.D. at the University of Hamburg, Germany in 1998 and 2000 and his Habilitation at the University of Marburg in 2024. He was a visiting professor at six LSA Linguistic Institutes, a Visiting Chair of the Centre for Corpus Approaches to Social Science at Lancaster University, and the Leibniz Professor at Leipzig University.
The development of language technologies for minoritised languages presents unique challenges, but also opportunities to create resources that are closely aligned with community needs and language revitalisation goals. This presentation explores how corpus-informed approaches have supported the development of Welsh-language technologies, tracing a pathway from the creation of large-scale linguistic resources to emerging AI applications.
Central to this work is CorCenCC (the National Corpus of Contemporary Welsh), a 12-million-word balanced corpus of written, spoken and digital language that provides core infrastructure for Welsh-language research and technology development. The paper draws on a series of interdisciplinary projects involving corpus linguists, applied linguists and natural language processing (NLP) researchers, including the development of Thesawrws, FreeTxt, and other openly available Welsh digital resources. These have supported work on sentiment analysis, readability assessment, CEFR (Common European Framework for Reference) level prediction, and the creation of a corpus-informed small Welsh language model.
Beyond their research value, these resources have generated new opportunities for language learning, teaching, digital literacy and community engagement. The paper demonstrated how corpus-led, context-aware approaches offer a sustainable methodological framework for language technology development and provide transferable lessons for other minority and minoritised language communities seeking to harness NLP and AI for both research and educational purposes.
Bio
Professor Dawn Knight is University Dean of Research Environment and Culture at Cardiff University and a corpus and applied linguist. Her research spans language, society, digital humanities and multimodality, with a particular focus on corpus development and analysis. She has helped pioneer Welsh-language corpora and language technologies, expanding research opportunities for minoritised languages. A former Chair of BAAL, she has led numerous UKRI- and Welsh Government-funded interdisciplinary projects and collaborations.
12:30 Lunch
Large language models (LLMs) are slowly entering the corpus linguist’s toolkit. Used critically, they can help researchers, among other things, clean, annotate, structure, query and interpret corpus data on a scale that manual analysis rarely allows. The evaluation of this work is still in the making. Much of it remains exploratory, and we surely need systematic evidence of when LLM output can be trusted and when it cannot. In language teaching, I have proposed (Pérez-Paredes, 2026) two plausible scenarios in which corpora and GenAI work together. In the first, GenAI extends established Data-Driven Learning (DDL) practice. In the second, corpus literacy becomes the core of a critical AI literacy that language teaches learners to test machine output against attested usage. There is already some evidence that a well-thought-of integration of both can be useful in language classrooms. The challenges are, however, many. LLMs generate synthetic language whose use cannot be tracked down to evidence, whereas corpora record attested, mainly human use. Bias, reproducibility, learner overreliance and the licensing of corpora for GenAI use unfortunately remain unresolved. In this talk, I argue, with McEnery (2025), for a critical adoption of GenAI in both corpus research and DDL, in which corpus specialists assess these tools and adopt only those that offer reliable value in the context of epistemological practices in corpus linguistics.
References
McEnery, T. (2025). Twenty years of Corpora. Corpora, 20(1), 1–2. https://doi.org/10.3366/cor.2025.0325
Pérez-Paredes, P. (2026). Corpus linguistics and AI in the reconfiguration of language learning ecologies. Language Teaching. https://doi.org/10.1017/S0261444825101109
Bio
Pascual Pérez-Paredes is Professor of Applied Linguistics and Linguistics at the University of Murcia, and former Lecturer in Research in Second Language Education at the University of Cambridge. His main research interests are the use of corpus linguistics methods in applied linguistics, corpora and digital resources in language education and corpus-assisted discourse analysis. He is Co-Editor in Chief of Cambridge University Press ReCALL journal, Associate Editor of TESOL Quarterly and Co-Editor or the Applied Linguistics Press book series Corpus Linguistics and Technology-Mediated Language Education in the AI Era.
This discussion opens with documented trends in recent corpus analysis research comparing human and AI-generated written English, particularly in academic genres. Next, I note implications for these patterns and close with how we might explore them in research and teaching.
Bio
Laura Aull is Professor of English and Linguistics at the University of Michigan. Her work focuses on institutional histories of English and lexicogrammatical patterns in student writing, published writing, and in human versus AI-generated written English. She is the author, most recently, of two forthcoming books: AI English: Humans, Machines, and the Future of Writing (Cambridge University Press), and Grammar for Good: How Language Knowledge Makes the World Better (Bloomsbury).
15:45 Break