19-20 March 2014 - Fifth EUCogIII Members Conference, Bochum
"Embodied communication" Programme
Last updated: 19/3/2014
Organisation: Gregor Schöner (Bochum), Vincent C. Müller (ACT / Oxford)
|Wednesday 19 March 2014|
|13:00||Registration with lunch snacks|
|14:00||Opening: Gregor Schöner (Bochum) / Vincent C. Müller (ACT / Oxford)|
|14:30-16:00||Plenary lecture followed by discussion: Peter Dominey (Inserm Stem Cell and Brain Research Institute, Lyon, France)
Grounding Language Comprehension and Production in Action Read more →
In the early years of life, children come to understand goal-directed action, both as observed in others, and in their own action. In the same time frame, they also come to understand their native language, and its intimate relation with intentional action. «Meaning» is in a sense the link between physical action and language. We have developed a research program that attempts to understand how meaning is extracted from the physical world of observed and performed action, and how the correspondence between the structure of meaning and the structure of language can be learned. We take a developmental, usage-based approach that attempts to use the minimum of pre-speciﬁed or innate capabilities. I will present the framework and results with the iCub robot learning the meaning of action, and the link between language and meaning which is then expressed by the ability to understand and produce novel pertinent sentences in the context of interaction with humans.
|16:00-17:00||Coffee break with poster session|
Plenary lecture followed by discussion: Srini Narayanan (International Computer Science Institute and The University of California, Berkeley, CA, USA)
Simulation Semantics: A computational framework linking language, cognition, and action [↓PDF] Read more →
Simulation semantics hypothesizes the mind as "simulating" the external world while functioning it. The "simulation" takes the best-fitting model of the noisy linguistic input together with general knowledge and makes new inferences to figure out what the input means and to guide response. This talk will briefly describe the simulation semantics hypothesis and report on a computational realization along with recent results on applying the model to long standing problems in language understanding. In terms of computational modeling, simulation semantics integrates models and phenomena at multiple scales providing a bridging framework that combines multi-scale probabilistic dynamic inference with deep semantic analysis based on construction grammar, frame semantics, and cognitive linguistics.
|19:30 - ...||Conference dinner|
|Thursday 20 March 2014|
|9:00 - 9:30||Coffee with breakfast snacks|
Plenary lecture followed by discussion: Susan Wagner Cook (Department of Psychology, University of Iowa, Iowa City, IA, USA)
Embodied communication: Hand gesture [↓PDF] Read more →
Although spoken language is a highly effcient means of communication, speakers gesture when they talk, bringing their bodies into communication. In this talk I will discuss recent research demonstrating the variety of ways that speech and gesture are coordinated, as well as evidence that listeners are highly sensitive to temporal coordination across modalities. I will argue that speech and gesture have mutually interacting inﬂuences on how we shape our message, and how that message is understood.
|11:00 - 11:30||Coffee break|
Plenary lecture: Larissa Samuelson (Department of Psychology and DeLTA Center, University of Iowa, Iowa City, IA, USA)
Embodied Word Learning: How action brings prior knowledge to bear [↓PDF] Read more →
In this talk I will present a developmental perspective on embodied communication, focusing on early word learning. I will present data from one study showing a critical role for embodied context and active exploration in toddler’s ability to generalize novel names for nonsolid substances. Data from a second line of work with children and a computational model will demonstrate the importance of information about an object’s spatial location in children’s ability to resolve referential ambiguity. Together these data show how actions and the body work to bring children’s represented knowledge to bear on current tasks and thereby support learning.
|13:00-14:00||Lunch break with poster session|
Plenary lecture followed by discussion: Yiannis Aloimonos (Computer Vision Laboratory, University of Maryland, College Park, MD, USA)
A minimalist grammar of human action: Toward grounding of human activity [↓PDF] Read more →
I introduce a minimalist grammar for human manipulation action. It is a context free grammar involving hands, objects, tools and their movements and relationships. Using this grammar together with signal processing operators for attention, tracking, segmentation and recognition, I show how to create semantic descriptions of complex human activity from perceptual data and then how to create natural language describing the activity. The signals representing the activity are parsed in a way similar to the manner natural language is parsed. I conclude with two experimental demonstrations of this theory using robots in the real world: (a) from language to action – robots executing verbal commands and (b) from action to language – robots observing a human activity and producing a description of the activity in natural language (English).
[Work funded in part by the European Union under the FP 7 Cognitive Systems Program (Project POETICON) , by The National Science Foundation and the National Institutes of Health (USA) and Qualcomm Corporation under an Innovation Prize Grant]
|15:30 - 16:00||Coffee break|
Discussion with the panel: Discussion with the panel: What do we learn from human embodied communication for how we may enable artiﬁcial cognitive systems to communicate with us about the world in which they are embedded?
• Srini Narayanan
• Susan Wagner Cook
• Larissa Samuelson
• Yiannis Aloimonos
• Malte Schilling (CITEC Bielefeld)
• Gregor Schöner