Intermodal Learning Architecture Sample Clauses

Intermodal Learning Architecture. Synchrony and contingency are two of the fundamental phenomena in tutoring and social learning. While there is a growing body of research on the phenomenon of synchrony, there exist only few models of synchrony on an artificial system (Xxxxxx et al., 2004; Kose-Bagci et al., 2009; Xxxx et al. 2009; Xxxx et al, submitted). Based on current results reported in literature, models have to address the following questions: • What is synchrony (in terms of a higher level and temporal structure as well as correlation measure)? (Definition) • What are the entities that synchrony works on? (Segmentation) • How can it be detected in the interaction? (Recognition) • What functions does it serve? (Model) • How does it vary in different speakers with their way of “acoustic packaging” and different situations (Analysis) • What is the role of the different modalities (e.g. does vision provide primarily spatial information whereas auditory synchrony is more related to temporal structure?) and how do they interplay? Currently, the scientific debate (Workshop on Intermodal Action Structuring, in ZiF, Bielefeld in July 2008) seems to converge towards a consensus that the important criteria for synchrony are (1) temporal co-occurrence of an event in different modalities and (2) a correlation between the characteristics of these events. In contrast, “inverse synchrony”, meaning that events in two modalities show a temporally exactly disjunct distribution – such as a sequence of speech being followed by a speech pause with a sound of noise that is deliberately being framed by the tutor’s utterance – does not constitute an instance of synchrony but rather describes the characteristics of causality or – within the context of interaction – contingency. The importance of contingency has been recognized by computer scientists and there exist already some computational models for contingency (x.x. Xxxxxxxx, 2005; Xx Xxxxx et al., 2008). However, these models tend to be focused on a single modality and rigidly limited to specific concrete applications where an “event” has been clearly defined (e.g. Xxxxxx et al., 2006). In order to xxxxxx research with respect to developmental learning on robots, the following questions need to be addressed in the near future: • What is contingency (in terms of temporal structure as well as with respect to semantic content, if any)? (Definition) • What are the entities that contingency works on? (Segmentation) • How can contingency be detected in the i...