Dr Dagmar Divjak
Externally funded research
Leverhulme Research Leadership Award [2017-2022]
Over five years, the Research Leadership Award from the Leverhulme Trust will allow us to develop new, accurate ways of describing speakers’ linguistic knowledge, by using machine-learning techniques that mimic the way in which humans learn.
The patterns we find will be verified in laboratory settings and then tested on adult foreign language learners to see if such patterns can help them learn a foreign language in a way that resembles how they learned their mother tongue.
The aim is to lead a step-change in research on language and language learning by capturing the linguistic knowledge adult speakers build up when they are exposed to a language in natural settings. These insights will help with the development of strategic language teaching materials to transform the way in which we teach foreign languages.
British Academy Mid Career Fellowship [2015-2016]
Frequency of exposure is among the most robust predictors of human behaviour, including linguistic behaviour: words that are encountered often are identified more quickly and processed more easily and more accurately. Frequency effects have been found in virtually every subdomain of language, and linguists have proposed usage-based theories that consider language as emerging from input frequencies.
The nature of frequency and the mechanics by which frequency achieves its effects are, however, not well understood, and currently no method is specified for tracking frequencies. This project aims to tackle these issues and put usage-based work on a sounder footing: to this end, it combines insights from cognitive scientific research on memory and attention with a deeper understanding of the mathematical properties of the measures used to track and model frequency.
A better understanding of how frequency of exposure impacts language knowledge will facilitate the development of cognitively realistic theories of language and of more effective language teaching materials.
A monograph, entitled "Frequency in language: context, memory and attention" is under contract with Cambridge University Press.
If you are looking for a 5-minute summary of this research and its implicatons for language learning, then our video with "Top tips for language learning" is for you!
BA Skills Acquisition Award [2014-2015]
To compensate for noise in communication language encodes bits of information redundantly: we repeat things, spell out important words, say the same thing in different ways, or add gestures and exaggerate our intonation. Redundancy is implicitly built into language structure as well. Thanks to the redundancy of language, yxx cxn xndxrstxnd whxt x xm wrxtxng xvxn xf x rxplxcx xll thx vxwxls wxth xn ‘x’ (tgts lttl hrdr f y dn’t vn kn whr th vwls r) (Pinker 1994: 181).
The statistical techniques that are now standardly used by corpus linguists to model linguistic data (i.e. regression techniques) require this redundancy to be removed from the dataset prior to modelling it. As a consequence, all properties carry an equal weight, making the models of language that are obtained cognitively unrealistic and therefore unsatisfactory to usage-based linguists.
To remedy this problem, I wanted to explore alternative ways to model their data, using techniques that learn from data in cognitively realistic ways. The Naïve Discriminative Learner (Baayen et al 2011) offers great potential in this regard as a computational implementation of the inductive learning protocol promoted in usage-based approaches to language.
Thanks to a BA Quantitative Skills Acquisition Award I was able to spent 6 weeks in the Quantitative Linguistics Lab in Tuebingen (Germany) during the 2014-2015 academic year. These stays gave the impetus to a special issue of the journal Cognitive Linguistics Looking back, looking forward.