I have been doing research mostly in the areas of language acquisition and development, cognitive semantics, autonomous knowledge construction, and qualitative reasoning. My main research method consists in building computational models.

Linguistic development

Some characteristics of a child's active vocabulary at some developmental time point can be a predictor of later problems in linguistic development. My research (with S. Stokes and J. Hay from Uni. of Canterbury, and A. Knott from Uni. of Otago, New Zealand) included building connectionist models of early lexicon and phonology acquisition trying to answer the questions: What characterizes words that are acquired early in development and why? What is the underlying mechanism of lexicon acquisition and what is the difference between healthy and delayed language learners?


Takac, M. and Knott, A. and Stokes, S.: What can Neighbourhood Density effects tell us about word learning? Insights from a connectionist model of vocabulary development. Journal of Child Language 44(2). 346-379. 2017.

Language acquisition

Besides meaning acquisition that I discuss below, I have been interested in models relating acquisition of syntax to language production. During my postdoc research at the Univ. of Otago, New Zealand, I worked with A. Knott and L. Benuskova on a connectionist model of language production, based on the assumption that high-level semantic representations of episodes are delivered by sensorimotor routines that have inherently sequential nature. This novel idea tries to reconcile Chomskian-flavoured nativist theories with empiricist connectionist linguistics, giving a new interpretation of linguistic universals as a consequence of common sensorimotor apparatus (rather that a genetically specified universal grammar). Our model combines acquisition of surface language conventions such as collocations and idioms with parameter setting mechanism of acquiring abstract grammar rules (word-order conventions in our case). Both surface and abstract knowledge is acquired from noisy input data (simulating child-directed speech).

Selected publications:

  • Takac, M., Knott. A.: A neural network model of episode representations in working memory. Cognitive Computation 7. 509-525. 2015.
  • Takac, M., Benuskova, L., Knott, A.: Mapping sensorimotor sequences to word sequences: A connectionist model of language acquisition and sentence generation. Cognition 125(2). 288-308. 2012.pdfbib
  • Takac, M., Benuskova, L., Knott, A.: A Sentence Generation Network that Learns Surface and Abstract Syntactic Structures. In Honkela, T.; Duch, W.; Girolami, M.; Kaski, S. (eds.), Artificial Neural Networks and Machine Learning - ICANN 2011, Springer, Heidelberg. 342-348. 2011.pdfbib
... more

Cognitive semantics and theory of meaning

I believe that making sense of one's own environment is not exclusively human ability. Building on works of C. S. Pierce, J. von Uexküll, and J. Šefránek, in the theoretical part of my PhD research I have developed a theory of meaning and understanding that extends these terms to animals, pre-verbal children and artificial agents. Recently, J. Šefránek and I have substantially revised our theory (Takáč and Šefránek, 2012).

Selected publications:

  • Takáč, M. and Šefránek, J.: Semantics of distinguishing criteria: from subjective to intersubjective. Interdisciplinary Description of Complex Systems 10(3). 248-269. 2012.pdfbib
  • Takáč, M.: Construction of Meanings in Biological and Artificial Agents. In Trajkovski, G., Collins, S. G. (eds.), Agent-Based Societies: Social and Cultural Interactions, IGI Global, Hershey, PA. 139-157. 2009.bib
  • Takáč, M.: Construction of Meanings in Living and Artificial Agents (PhD thesis). Comenius University, Bratislava. 2007.pdfbib
... more

Adaptive knowledge representation

A good theory of semantics has to explain how do we acquire the meanings we have. I build on Šefránek's concept of "cognitive semantics of distinguishing criteria". A distinguishing criterion is an abstraction of ability to distinguish, recognize, understand and is a building block of our cognitive semantics. I have developed the cognitive semantics of distinguishing criteria into a formalized and implemented framework, where each distinguishing criterion can be autonomously constructed in interactions with the environment and can be continuously adaptively updated/refined. My current research interests in this area include episodic organization of knowledge, affective and pragmatic aspects of meaning and communication, speech acts, cognitive motivation and reinforcement learning.

Selected publications:

  • Takáč, M.: Autonomous Construction of Ecologically and Socially Relevant Semantics. Cognitive Systems Research 9(4). 293-311. . October 2008.pdfbib
  • Takáč, M.: Developing Episodic Semantics. In Honkela, T., Pöllä, M., Paukkeri, M., Simula, O. (eds.), Proceedings of AKRR '08, the 2nd International and Interdisciplinary Conference on Adaptive Knowledge Representation and Reasoning, Helsinki University of Technology, Helsinki. 90-96. 2008.pdfbib
... more

Qualitative modeling

Qualitative modeling is a method of reasoning about systems around us (from physical/mechanical to social and economical) allowing to draw some conclusions about predicted behaviour of the studied system without the necessity to deal with numeric/quantitative details. I mainly specialize in signed digraph models. I have developed a method for classifying equilibria in qualitative models and I supervised a master research on application of signed digraph models in economy.

Selected publications:

  • Takáč, M.: Kvalitatívne modelovanie a simulácia. Comenius University Press, Bratislava. 2003.bib
  • Takáč, M.: Fixed Point Classification Method for Qualitative Simulation. In Costa, E., Cardoso, A. (eds.), Progress in Artificial Intelligence: Proceedings of the Eighth Portuguese Conference on Artificial Intelligence (EPIA '97), Springer, Berlin. 255-266. 1997.pdfbib
... more