Dagmar Divjak


14.–16. 10.


Dagmar Divjak


  Russian and Slavonic Studies
The University of Sheffield

Odborné zaměření:
– Kognitivní lingvistika vycházející z užívání jazyka
– Korpusová lingvistika


14. 10. /č. 18/ 14.10


The quantitative turn in Cognitive Linguistics: why cognitive linguists (should) like data
In this first talk I look at the quantitative turn in Cognitive Linguistics. According to a recent survey (Janda 2013), the majority of conference presentations, articles, and books in our field now relies on some kind of quantitative information language data, and analyzes the data using statistical methods. I will zoom in on one type of quantitative data, i.e. information on frequency of occurrence, and explain why cognitive linguists‘ fascination with frequency is unavoidable and why working with frequencies of occurrences necessitates the use of statistics. I will then move on to considering some of the currently available ways in which to capture frequency information and argue that we need to relate our measures to what we know about human memory systems in order to be able to develop cognitively realistic ways of tracking occurrence frequencies.


15. 10. /č. 300/ 17.30
Modelling the aspect-modality interaction: pieces of a cognitive puzzle

In this talk I look into the assignment of aspect in Russian and Polish modal constructions. Unlike previous accounts, I take a strictly corpus-based, quantitative approach within which corpus data on the relationship between aspect and modality are modeled using mixed effects logistic regression.
The model suggests that, contrary to assumptions in the literature, aspect choice is not best predicted by modality type, and that the relation between modality type and aspect needs reversing: imperfective infinitives prefer to convey deontic meanings while perfective infinitives prefer expressing
dynamic modality. I will argue that, from a cognitive linguistic point of view, the use of aspect in modal contexts is a case of analogical mapping between domains. In performing this cognitive operation, speakers appear to be guided by the
deep similarities between the two domains: both perfective aspect and dynamic modality are bounded or marked in some respect, while imperfective aspect and deontic modality
remain unmarked and unbounded. This similarity may well have given rise to the very strong tendency to use the imperfective aspect to signal deonticity, and the perfective to render
dynamic modality. 

16. 10. /č. 18/ 12.30


Machine meets man: can a statistical model predict what a speaker will like best?
In this talk, I focus on the question of what language has to offer the learners in their quest for the meaning of lexemes and the concepts they give access to. I will present results of an ongoing research project that cross-validates a statistical corpus-based model designed to predict the preference for one of 6 near-synonyms that express the concept TRY in Russian by letting it compete against human beings in a series of psycholinguistic experiments. The results show that machine and human beings perform at the same level and show similar error patterns. The findings thus provide support for the Distributional Hypothesis (Harris 1954) that says that the meaning of a word is derived from the linguistic contexts in which it occurs and are relevant for concept acquisition and representation in general.