We need to clarify features of human speech to make assistive robots more human adaptive. In this study, we focused on speech produced in human-human virtual collaborative conveyer task and transcribed all utterances spoken while completing the task. We computed morphemes (words or meaningful parts of words) by word class categories. The results were that the total number of morphemes spoken in each trial decreased over ten trials. We found that a strong negative correlation between the number of morphemes by type in each trial and the number of trials. Task general verbs remain while task specific verbs decreased. Among the task specific verbs, verbs with more wide meaning tend to remain. Results suggest that in accordance with acquiring skill for the task, the entire complexity of ontology decreases and becomes more efficient.
[1]
C Snow,et al.
Child language data exchange system
,
1984,
Journal of Child Language.
[2]
Robert M. Davison,et al.
Global Applications of Collaborative Technology: Introduction
,
2001,
CACM.
[3]
Matthias Scheutz,et al.
Robust spoken instruction understanding for HRI
,
2010,
HRI 2010.
[4]
J. Gregory Trafton,et al.
Robot-directed speech: using language to assess first-time users' conceptualizations of a robot
,
2010,
HRI 2010.
[5]
Yasuhiro Katagiri,et al.
Prosodic alignment in human–computer interaction
,
2007,
Connect. Sci..