... sound?¹

Our system does not support this exact sample dialogue. For example, at this point in a conversation it displays the information about the restaurant on the screen and asks ``How does this one sound?'' or a similar question.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... place.²

This response shows that the Inquirer will have learned how to use the system more efficiently as well.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... components.³

As further discussed in Section 5.2, our approach to destination advice draws on an earlier analysis of the task by [30,31].

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... attribute.⁴

Because other constraints can later be modified, the system lets the user later specify any value, even the one that caused the over-constrained situation.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... user.⁵

In the user study described in Section 4, no users specified a disjunctive query.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... exactly.⁶

For attributes where the user has selected more than one value, we assume that any supplied value would be acceptable.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... reverse.⁷

CBR systems do not necessarily use the same weighting factors for each of similarity computation and question ordering. However, for our application area, it is correct to make the assumption that an attribute's importance is the same as its impact on the similarity computation.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... constraints.⁸

When we discuss the number of items matching the constraints, we refer to those items that remain after similarity filtering as discussed in Section 3.2.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... attribute⁹

Recall from Section 3.2 that this is actually the lowest ranking attribute in the user model.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... systems.¹⁰

The work on adaptation of speech recognition grammars (e.g., Stolcke et al., 2000), while related, addresses a different problem and uses different learning techniques, so we do not discuss it here.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.