Graphics Reference
In-Depth Information
For instance, if the user asks to “edit a process”, the system could
suppose that the user has a wrong belief, that is that a process can be
edited. When a discrepancy is found, a feedback utterance is generated.
Nakano et al. (1999) developed a spoken dialogue system, called
WIT, that enables robust utterance understanding and real-time
listener's responses. The content of the responses generated by this
system is highly appropriate, but since it lacks control on the timing
of response emission, users can feel uncomfortable with the resulting
delayed interaction.
Heylen et al. (2004) introduced affective feedback in the tutoring
system INES. Such a system was used to let students practice on the
computer the task of giving an injection in the arm of a virtual patient.
INES, graphically represented by a talking head, provides emotional
feedback according to the level of correctness of the student's actions.
Four emotions have been considered: joy, distress, happy-for and
sorry-for; for example, INES shows joy when the student succeeds. The
aim of this type of feedback consists in making the learning process
more effective by taking into account the user's emotional state. The
system is not able to recognize the student's emotional state, but it
makes some assumptions looking at the student's achievements and
failures in doing injections.
According to Kopp et al. (2008), a conversational agent should be
able to respond in a pertinent and reasonable way to the statements
and the questions asked by a user. Backchannels provided by this
type of agent should derive from its internal state that describes how
it feels and what it thinks about the speaker's speech. The model
proposed by Kopp reflects this idea and it has been tested with Max,
a virtual human developed at the A.I. Group at Bielefeld University.
Max appears in an information desk in the Heinz-Nixdorf-Museums
Forum and provides visitors with information about the museum and
the exhibitions. To communicate with Max, visitors type what they
want to say on a keyboard and Max responds through both verbal
and non-verbal behaviors. The backchannel model implemented for
Max is based on a reasoning and deliberative processing that plans
how and when the agent must react according to its intentions,
beliefs and desires. Max is able to display multimodal backchannels
(like head nod, shake, tilt and protrusion with various repetitions
and different movement quality) that are triggered solely according
to the written input that the user types on a keyboard. The system
evaluates the input in an incremental fashion, parsing the words typed
by the user to update constantly its knowledge about the topic of the
conversation. To determine backchannel timing, the system applies the
Search WWH ::
Custom Search