Graphics Reference
In-Depth Information
the system that they talked to and that talked to them during the
experiments.
Linguistic structures
First inspections of transcripts quickly revealed that there were many
variations in lexicalization but only a small number of linguistic
constructs that the subjects employed during the packing and
unpacking sub-dialogs of the LAST MINUTE experiments. For issuing
packing (or unpacking) commands, these structural options were used:
￿
full sentences with a variety of verbs or verb phrases and variations
in constituent ordering,
￿
elliptical structures without verbs in a (properly inflected) form
like <number> <item(s)>,
￿
'telegrammatic structures' in a (mostly uninflected) form with an
inverted order of head and modifier like <item> <number>.
What was the distribution of these types of linguistic constructs?
During a quantitative analysis of the 'LAST MINUTE' phase, the
absolute and relative numbers for the usage of these constructs had
already been calculated from the full set of transcripts.
Rapport
The next area explored was the question of rapport: Did the users
mirror the language of the system, e.g. on a lexical or syntactic level?
For example, this system uses the general and somewhat non-colloquial
term 'hinzufügen' (engl. to add) in its feedback for selection operations.
Similarly the system always uses 'wurde entfernt' (engl. approx. 'was
removed') when confirming unpacking requests. Did the users mirror
this usage?
Politeness
In all its utterances, the system uses the polite German 'Sie' (polite,
formal German version of 'you') when addressing the user. In requests
the system employs the politeness particle 'bitte' (engl. 'please'). How
polite were the users during their utterances?
5.2 Methods and results
The general approach for all three questions was to perform quantitative
analyses with appropriate search patterns within the complete set of N
Search WWH ::




Custom Search