Next: Interactive Constraint-Satisfaction Search
Up: Personalized Conversational Recommendation Systems
Previous: Conversational Recommendation
Conversation via Dialogue Management
Dialogue systems carry out conversations with users in natural
language, whether spoken or typed. The main tasks performed by
dialogue systems are language interpretation, language generation,
and dialogue management. Natural language interpretation and
generation are topics onto themselves and we will not discuss them
here; for two introductory texts, see
[9] and
[42].
To enable a focus on user modeling, our system
allows moderately complex user utterances but has a pre-coded set of
system utterances, as discussed further in Section 3.3.
The simplest dialogue managers are based on finite-state automata in
which the states correspond to questions and arcs correspond to
actions that depend on a user-provided response
[75,81].
These systems support what are called
fixed- or system-initiative conversations,
in which only one of the participants controls the actions, whether it be the
system helping the user or the user asking questions of the system.
Next in complexity are frame- or template-based
systems in which questions can be asked and answered in any order
(Bobrow et al., 1977). Next,
true mixed-initiative
systems
allow either dialogue participant to contribute to the interaction as
their knowledge permits (Allen, 1999; Haller & McRoy, 1998;
Pieraccini, Levin, & Eckert, 1997).
Thus, the conversational focus can change at any time due to
the user's (or system's) initiative of that change.
Finally, some different approaches that support sophisticated
dialogues include
plan-based systems (Allen et al., 1995; Cohen & Perrault, 1979)
and systems
using models of rational interaction [66].
To allow reasonably complex
conversations while keeping the system design straightforward,
we chose a frame-based approach to dialogue management.
Thus, the ADAPTIVE PLACE ADVISOR allows
more conversational flexibility than a fully system-initiative
paradigm would allow. Users can fill in
attributes other than or in addition to those suggested by the system.
However, they cannot force the system to transition to
new subtasks, nor can the system negotiate with users to determine which
participant should take the initiative.
Next: Interactive Constraint-Satisfaction Search
Up: Personalized Conversational Recommendation Systems
Previous: Conversational Recommendation
Cindi Thompson
2004-03-29