Modeling language acquisition in humans and robots

What are the mechanisms needed to learn language? How is it possible that children learn language so easily and with apparently poor linguistic input? How can a robot guess the meaning of a new word? How are the social regulation mechanisms involved in language learning? How can one draw the attention of a robot towards particular aspects of their environment? What are the interactions between acquisition mechanisms and language evolution?

We investigate the mechanisms that enable humans and robots to learn new words and to use them in appropriate situations. We have built a number of robotic and computational experiments studying the mechanisms of concept formation, joint attention, social coordination and language games, and articulating the roles of learning, physical and environmental biases in language acquisition. The unifying theme of all these experiments is development: we explore the hypothesis that language can only be acquired through the progressive structuring of the sensorimotor and social experience.

More specifically, we have been focusing on language learnability and on the problem of the poverty of the stimulus. The poverty of the stimulus argument, central in linguistics theory, posits that only positive examples of correct sentences are present in children's linguistic input, and that it is not enough to account for language learning without assuming innate linguistic knowledge. This argument has nevertheless been strongly challenged by field studies of language acquisition as well as conceptual advances in learning theory and computational modeling.

Computational models have played a central role in the debate over language learnability, from generative views to more recently introduced explanatory frameworks based on embodiment, cognitive development and cultural evolution. We have developped a body of models and computational approaches which has allowed us to organize, transform and rephrase defining questions about what makes language learning possible for children. In particular, we propose a tentative synthesis to recast the debate, based on the notions of development, learning biases and self-organization (Kaplan, Oudeyer and Bergen, 2008), and integration language acquisition issues within the larger question of language evolution conceptualized as a cultural Darwinian process (Oudeyer and Kaplan, 2007; Oudeyer, 2005).

This research has been conducted in collaboration with my colleague Frédéric Kaplan.

Keywords: Language acquisition, social interaction, robots, computational modeling, joint attention, intrinsic motivation, learning biases, physical biases, social biases, self-organization, selection for learnability, poverty of the stimulus.

Selected papers:

On the role of intrinsic motivation in language development:

Oudeyer P-Y., Kaplan F. (2006) Discovering Communication, Connection Science, 18(2), pp. 189--206. Bibtex

On language learnability and the poverty of the stimulus problem:

Kaplan, F., Oudeyer, P-Y., Bergen B. (2008) Computational Models in the Debate over Language Learnability, Infant and Child Development, 17(1), pp. 55--80. Bibtex

Oudeyer, P-Y (2005) How phonological structures can be culturally selected for learnability, Adaptive Behavior, 13(4), pp. 269--280. Bibtex

On language evolution conceptualized as a cultural Darwinian process:

Oudeyer P-Y., Kaplan F. (2007) Language Evolution as a Darwinian Process: Computational Studies, Cognitive Processing, , 8(1), pp. 21--35. DOI: 10.1007/s10339-006-0158-3. Bibtex

On concept formation in language acquisition:

Kaplan F., Oudeyer P-Y. (2006) Comment les robots construisent leur monde, in proceedings of Sony CSL 10th anniversary symposium, Paris.

Oudeyer, P-Y. (2006) Self-Organization in the Evolution of Speech, Studies in the Evolution of Language, Oxford University Press. Bibtex