Self-organizing neural network models of language acquisition

Self-organization is an important process of human experience. In many contexts of learning and representation, we do not rely on explicit instruction about what is correct or incorrect to learn, but instead gather information about the "input space" (i.e., the limits, constraints, and possibilities of things) and organize information in such a way that we can optimally perform a given task. In the case of language, it has been debated whether learners use corrective feedback to learn or to organize speech, grammar, or lexicon.

We would like to consider human learning as a self-organizing process, and that the activities reflected in our self-organizing network (organization, competition, and associations) can capture the nature of language acquisition, representation, and processing. In this project, we build and expand on DevLex, a self-organizing neural network model, and make it a cognitively and neuropsychologically plausible model that can account for a wide variety of phenomena in language. Our model incorporates properties of self-organization, Hebbian learning, lexical co-occurrences learning, and dynamic growth. These properties make the model well suited for the study of the mental lexicon, its structure, representation, and processing in children, normal adults, second language learners, and brain-injured patients.

Cross-linguistic studies

A large body of knowledge has accumulated especially in the last three decades on the cognitive processes and brain mechanisms underlying language use, language acquisition, and language disorders. Much of this knowledge has come from studies of Indo-European languages, in particular, English. Some researchers believe that because of the universal principles of language, theories of language and language processing should apply in the same way to all languages even if they are built on facts from specific languages. Others, however, think that language-specific variations are sufficiently strong to warrant different conceptualizations of linguistic principles and cognitive underpinnings for different languages.

Unlike generative theories of language, this second perspective itself is a mixed bag, from the strongest form of the Sapir-Whorf hypothesis that argues for linguistic determinism to modern-day psycholinguistic theories that emphasize language variation and competition. The tension between these two perspectives has yielded much debate in the cognitive and psycholinguistic studies of language, and it is against this backdrop that we do our crosslinguistic research in lexical representation and ambiguity processing, sentence interpretation, and bilingual lexical and sentence processing.