|
Abstract Recurrent Temporal Network(ATRN) Model (Dominey, Hoen,Blanc,Lelekov) (Dominey et al.)
|
The model developed during the research and tested against real life neurophysiological
data is focused on the benefits of the particular features of sequential cognition for
human language skills.A model which is based on the functional neuroanatomy
of primate cortex and basal ganglia is constructed with the purpose of cognitive sequence
processing.Then the capabilities of this model in human infant simulation for drawing out
3 important structures of serial, temporal and abstract concepts from language-like sound
sequence input. Current research presented how this model can perform adult level syntactic
understanding by training based on open and closed class words.Two main predictions are made in relevance
to model and their validity has been tested against existent data in literature. Those predictions are
that impaired syntactic processing is associated with non-linguistic cognitive sequencing task relevant
damages and syntactic processing should be involved in non-linguistic cognitive sequence tasks as well as
it is involved in neurophysiological processes. As a result of this research it is concluded that sequential
cognition will provide a valuable pattern in neurophysiological language studies.
|
|
Abstract Temporal Recurrent Network ( Dominey - Hoen - Blanc - Lelkov-Boissard) (Dominey et al.)
|
A model that describes the cognitive sequence processing based on the functional neuroanatomy of the primate cortex and basal ganglia. This model illustrates neurologically plausible system for extracting serial, temporal, and abstract structure in both linguistic and non-linguistic cognitive sequencing tasks.
|
|
Syntactic Comprehension and Artificial Grammar Learning (Dominey,Inui,Hoen) (Dominey - Inui - Hoen)
|
The organization of brain in terms of distributed neural networks and how this structure can be used in nonlinguistic sequence cognition as well as being used in language acquisition is a valuable area to study profoundly. Important researches have been made about important aspects of this issue such as sentence comprehension, nonlinguistic sequences and artificial grammar learning (AGL) in order to examine the brain mechanisms and related processes involved in these tasks. Current research accommodates data obtained from those studies and several other neurophysiological models of sentence processing and specifies a new architecture using known human language cortico-striato-thalamo-cortical (CSTC) constraints within a neural network model. One important aspect of evaluating the proposed system in respect to human performance is developing simulations that display the system is able to learn and carry out similar language and artificial syntax tasks while considering the actual neuroanatomical connectivity boundaries and real world brain imaging data. In current model a recurrent cortical network in Broadmann Area 47(BA47) encodes structural cues which later activate above mentioned CSTC circuit. This circuit modulates lexical semantic information flow between Broadmann Area 45 (BA45) and Broadmann Areas 44 and 6 (BA44/6). The latter one is the locus of sentence level meaning representation. In language learning process closed class words control thematic role assignment (TRA) trough corticostriatal plasticity. For Artificial Grammar Learning (AGL) tasks, repetitive internal structures are encoded in BA47 and activate CSTC circuit to anticipate the next element in the sequence. The model is a syntactic comprehension architecture that depicts interaction of BA44/45/47/6, Basal Ganglia (Caudate and SNr) and Thalamus. Two main tasks performed for performance evaluation of this developed model are Caplan's Test for thematic role assignment and syntactic comprehension (Caplan, D., Baker, C., & Dehaut, F., 1985) and Artificial Grammar Learning (AGL) study of Gomez and Schvaneveldt. (Gomez and Schvaneveldt, 1994)
|
|
