日時:7月5日(火) 13:00-16:15
場所:北海道大学 学術交流会館1F 第4講義室
13:00-14:30
Speaker: Peter Ford Dominey PhD (CNRS Research Director, Robot Cognition Laboratory)
Title: “Neural basis of semantic representations: Grounding Meaning in Sensory-Motor Experience and Narrative with the iCub”
Based on data from human and non-human primate neuroscience we have developed neural network models of grammatical construction
processing for sentence comprehension and production. These models are embedded in a cognitive system that allows the iCub
humanoid to interact and cooperate with people. Language semantics was initially linked directly to robot sensory and motor
systems via symbolic interfaces. The semantics is extended in two dimensions: Neuroscience indicates that meaning is grounded
in experience and encoded in convergence-divergence zones (CDZ) centred on highly interconnected areas such as the temporal-parietal
junction. We thus developed self-organizing maps that encode and recall the robot’s sensorimotor experience based on this CDZ framework.
In parallel, we ground semantics over extended time in narrative constructions that allow meaning representations to be enriched by
establishing causal links between events and intentional states over time. The grammatical construction models are extended to
narrative constructions by including narrative context. This is part of a long term effort to build towards narrative construction of meaning.
14:45-16:15
Speaker: Jocelyne Ventre-Dominey PhD (Senior Researcher, INSERM-Unity 1208, University of Lyon)
Title: “Neural basis of semantic representation: I- Human neuroscience”
This presentation will address in human the conceptualization processes and their neural substrates during events analysis.
The main questions are: 1) how event comprehension unfolds when we process pictures and auditory or visually presented sentences
describing events and 2) Is there a common neural network and processes in semantic analysis of events independently of the sensory
modality? We address these issues by different behavioural and neuroimaging approaches including the localisation of the neural
substrates and their connections respectively by fMRI activation and DTI tractography as well as the spatio-temporal interactions
in the network by evoked potentials technique. In this talk I will present the results of these successive studies showing
(1) different levels of an imageability score corresponding to the capability of forming a mental image during sentence reading,
(2) a shared neural network activated during event comprehension independently of the visual or language modality,
(3) significant structural connections between the majority of the nodes in the network associated to correlations between their
neural activities, including the anterior temporal pole and parietal, para-hippocampal, prefrontal cortex,
(4) significant correlations between the imageability of sentences for individual subjects and the relative pathway density
issued from these parieto-temporal and temporo-frontal cortical tracts suggesting a potential functional link between comprehension
and the temporo-parieto-frontal connectivity strength and finally (5) late positivity in the evoked potentials activity recorded in
common for sentences and images. These data help to define a “meaning” network that includes components of recently characterized
systems for semantic memory, embodied simulation, and visuo-spatial scene representation.