| ||||||||||||
| Themis Karaminis | ||||||||||||
Research InterestsI am currently a postdoctoral research fellow in computational linguistics at the University of Oxford. [a] What is the nature of the neurocomputational mechanisms supporting the learning and use of language? [b] How could the same mechanisms account for empirical phenomena in the acquisition of different languages? [c] What about atypical language acquisition? What type of neurocomputational deficits underlie the language-learning problems of children who follow atypical developmental trajectories? [d] How could the same type of neurocomputational deficits account for atypical linguistic profiles across languages? I approach these research questions through computational modelling, in particular through simulations with connectionist or artificial neural networks. The connectionist framework is a powerful tool for the study of cognitive development. It allows implementing artificial learning systems, inspired by the general properties of computations in the brain. Cognitive theories of development can be refined by observing how these systems are shaped by the environment to which they are exposed. Indeed, a large number of connectionist studies have proposed neural network models of the acquisition of English and/or other languages (research questions [a] and [b]). Other studies have shown how theoretically-driven manipulations of the parameter space of connectionist models may lead to conditions of constrained learning symptomatic of atypical language development (research question [c]). However, cross-linguistic variation in atypical language development (research question [d]) has been studied to a lesser extent under the connectionist framework. To address research question [d], and moreover to do so whilst also addressing research questions [a]-[c], models of language acquisition need to be general across two dimensions, at the same time: cross-linguistic variation and variation across language groups. For my Phd studies, I developed a series of such models for the acquisition of different domains of linguistic knowledge, namely inflectional morphology, syntactic comprehension and syntactic production. Cross-linguistic generality was considered by modelling the acquisition of English (relatively simple morphology, wide use of uninflected forms, fixed word-order) and the acquisition of Modern Greek (a complex fusional morphological system, absence of uninflected forms, flexible word-order). Generality across language groups was addressed by contrasting language acquisition in typically developing children and children with Specific Language Impairment.
BioPhD in Psychology, Department of Psychological Sciences, Birkbeck College, University of London. Project title: Connectionist modelling of morphosyntax in typical and atypical development for English and Modern Greek. Supervisors: Prof. M. Thomas and Dr. R. Cooper MSc (Neural Computation/Neuroinformatics, Natural Language Processing), School of Informatics, University of Edinburgh MEng (Computer Engineering and Informatics), Department of Computer Engineering and Informatics, University of Patras, Greece
PublicationsKaraminis, T. N., & Thomas, M. S. C. (in prep., invited chapter). A connectionist model for the cross-linguistic acquisition of inflectional morphology in typical development and SLI for English and Modern Greek. Karaminis, T., & Thomas, M. S. C. (in prep.). A connectionist model of the acquisition of inflectional morphology in English and Modern Greek. Manuscript in preparation. Thomas, M. S. C., Baughman, F. D., Karaminis, T., & Addyman, C. J. M. (in press). Modelling developmental disorders. In C. Marshall (Ed.), Current Issues in Developmental Disorders. Psychology Press. Click here for pdf of final draft (1.1mb) Karaminis, T. N., & Thomas, M. S. C. (2012) Connectionism. In Seel, Norbert M. (Ed.), Encyclopedia of the Sciences of Learning (Part 3, Pages 767-771). Springer, New York, NY. Click here for pdf of uncorrected proof (224k). Karaminis, T. N., & Thomas, M. S. C. (2012). Connectionist Theories of Learning. In Seel, Norbert M. (Ed.), Encyclopedia of the Sciences of Learning (Part 3, Pages 771-774). Springer, New York, NY. Click here for pdf of uncorrected proof (202k). Thomas, M. S. C., Karaminis, T. N., & Knowland, V. P. (2010). What is typical language development? Language Learning & Development, 6, 162-169. Click here for PDF version (142k). Karaminis, T. N., & Thomas, M.S. C. (2010). A cross-linguistic model of the acquisition of inflectional morphology in English and Modern Greek. In S. Ohlsson & R. Catrambone (Eds.), Proceedings of the 32nd Annual Conference of the Cognitive Science Society (pp. 730-735). Austin, TX: Cognitive Science Society. Click here for PDF version (275k)
Posters/ Presentations/ TalksA connectionist model of the acquisition of Inflectional Morphology in typical development and Specific Language Impairment in English and Modern Greek. Poster presented at the conference of COST Action A33 'Let the Children Speak: Learning of Critical Language Skills across 25 Languages'. A European-wide initiative on Language Acquisition and Language Impairment, London (UK), 22-24 January 2010. The Multiple Inflection Generator: A generalised developmental model of inflectional morphology. Poster presented at 2009 Biennial Meeting of Society for Research in Child Development, April 2-4, Denver, USA. ( *.pdf) A connectionist model of English inflectional morphology. Talk presented in the 2008 Alston Child Language Meeting, May 8-10, Alston, UK. ( *.pptx)
AcknowledgementsI am particularly grateful to the Greek State Scholarship Foundation (IKY) and the Greek Ministry of Education, Lifelong Learning and Religious Affairs for funding my MSc and doctoral studies.
LinksCognitive Science at Edinburgh
|
