CogSci-Talk: Prof. Dr. Soonja Choi (San Diego State University)

Language and Thought: Spatial Semantics & Spatial Cognition from Infancy to Adulthood

Jan 13, 2012
Lecture Room 2i , NIG (Universitätsstraße 7, 1010 Wien / 2nd floor)
13.30 - 15.00 

Languages differ significantly in the way they categorize spatial relations.  For example, English makes a distinction between containment (e.g. putting an apple IN a bowl) and support (e.g. putting a cup ON a table), whereas Korean makes a distinction between loose fit and tight fit regardless of containment and support.  In Korean, the verb KKITA ‘tight fit or interlock’ is used for both a tight-fit containment relation such as ‘putting a book tightly in its box-shaped cover’ and a tight-fit support relation such as ‘putting a Lego piece tightly onto another’.

The extensiveness of cross-linguistic differences in spatial semantic categorization found in recent studies on adult grammars raises questions about when and how children acquire the spatial semantic system of their native language, and more generally, about the relationship between language and cognition in children and adults.  In this talk, I present studies that examine language-specific input and spatial cognition in learners and adult speakers of English and Korean.  In particular, I examine whether and to what extent language-specific semantics can influence nonverbal spatial categorization involving tight fit, containment and support.  Overall, my studies show that there is a dynamic interaction between language and cognition from an early age and that language starts to influence spatial cognition as children use spatial words productively.  However, some perceptual aspects persist and contribute to spatial categorization in certain contexts regardless of language-specific input.  

download abstract

download short CV

First CogSci-Talk: Prof. DDr. Matthias Scheutz (Tufts University)

Steps towards embodied models of situated task-based natural language dialogues". 

Dec 14, 2011
Lecture Room 23 (HS23), Main Building of the University of Vienna map
15.30 - 17.00 

Perception, action, and natural language understanding are all tightly intertwined in the human cognitive system, giving rise to complex patterns of actions, utterances, and responses.  Different from written usages of natural language, meaningful linguistic fragments in situated spoken dialogues are typically determined by utterance and perceptual context together with prosodic, temporal, task, and goal information. Hence, one main challenge for computational models of human-like situated task-based dialogue interactions is to determine and extract meaningful linguistic fragments that are often not aligned with sentence boundaries.  Moreover, computational models need to account for the fact that real dialogues are full of incomplete, ungrammatical, and ambiguous utterances without definite truth values. In this talk, we will present results from human experiments that reveal the intricacies of human dialogue-based activity coordination in a joint remote search task.  And we will show first results from our attempts to develop a complex, integrated robotic architecture that is based on findings in psycholinguistics about the nature of lexical, syntactic, semantic, and pragmatic processing to model situated task-based natural language interactions in humans.

 download abstract

download short CV