. SIMILAR TO THE NM, THEY USE THE DISCOURSE CORRECTNESS INTERPRET...

1998). Similar to the NM, they use the discourse

correctness interpretation. We find that if the NM

structure information to display a segmented inter-

was present users had fewer AsrMis and fewer

action history (SIH): an indented view of the inter-

SemMis (trend for SemMis, p<0.09).

In addition, a χ

²

dependency analysis showed

action augmented with purpose information. This

that the NM presence interacts significantly with

paper extends over their work in several areas. The

most salient difference is that here we investigate

both AsrMis (p<0.02) and SemMis (p<0.001), with

the benefits of displaying the discourse structure

fewer than expected AsrMis and SemMis in the

information for the users. In contrast, (Rich and

Sidner, 1998) never test the utility of the SIH.

3

Due to random assignment to conditions, before the

Their system uses a GUI-based interaction (no

first problem the F and S populations are similar (e.g. no

speech/text input, no speech output) while we look

difference in pretest); thus any differences in metrics

at a speech-based system. Also, their underlying

can be attributed to the NM presence/absence. However,

in the second problem, the two populations are not simi-

task (air travel domain) is much simpler than our

lar anymore as they have received different forms of

tutoring task. In addition, the SIH is not always

instruction; thus any difference has to be attributed to

available and users have to activate it manually.

the NM presence/absence in this problem as well as to

Other visual improvements for dialogue-based

the NM absence/presence in the previous problem.

computer tutors have been explored in the past

4

Due to logging issues, 2 S users are excluded from this

(e.g. talking heads (Graesser et al., 2003)). How-

analysis (13 F and 13 S users remaining). We run the

ever, implementing the NM in a new domain re-

subjective metric analysis from Section 5.1 on this sub-

quires little expertise as previous work has shown

set and the results are similar.

tions. In Proc. of Workshop on Bridging the Gap:

that nạve users can reliably annotate the informa-

Academic and Industrial Research in Dialog Technologies.

tion needed for the NM (Passonneau and Litman,

J. Allen, G. Ferguson, B. N., D. Byron, N. Chambers,