. OUR NM DESIGN CHOICES SHOULD ALSO HAVE AN M. DZIKOVSKA, L. GALE...

1993). Our NM design choices should also have an

M. Dzikovska, L. Galescu and M. Swift. 2006. Ches-

equivalent in a new domain (e.g. displaying the

ter: Towards a Personal Medication Advisor. Journal

recognized user answer can be the equivalent of

of Biomedical Informatics, 39(5).

the correct answers). Other NM usages can also be

J. Allen, G. Ferguson and A. Stent. 2001. An architec-

imagined: e.g. reducing the length of the system

ture for more realistic conversational systems. In

turns by removing text information that is implic-

Proc. of Intelligent User Interfaces.

itly represented in the NM.

J. Cassell, Y. I. Nakano, T. W. Bickmore, C. L. Sidner

and C. Rich. 2001. Non-Verbal Cues for Discourse

7 Conclusions & Future work

Structure. In Proc. of ACL.

In this paper we explore the utility of the Naviga-

A. Graesser, K. Moreno, J. Marineau, A. Adcock, A.

tion Map, a graphical representation of the dis-

Olney and N. Person. 2003. AutoTutor improves deep

course structure. As our first step towards under-

learning of computer literacy: Is it the dialog or the

standing the benefits of the NM, we ran a user

talking head? In Proc. of Artificial Intelligence in

Education (AIED).

study to investigate if users perceive the NM as

useful. From the users’ perspective, the NM pres-

B. Grosz and C. L. Sidner. 1986. Attentions, intentions

ence allows them to better identify and follow the

and the structure of discourse. Computational Lin-

guistics, 12(3).

tutoring plan and to better integrate the instruction.

It was also easier for users to concentrate and to

D. Higgins, J. Burstein, D. Marcu and C. Gentile. 2004.

learn from the system if the NM was present. Our

Evaluating Multiple Aspects of Coherence in Student

Essays. In Proc. of HLT-NAACL.

preliminary analysis on objective metrics shows

that users’ preference for the NM version is re-

J. Hirschberg and C. Nakatani. 1996. A prosodic analy-

flected in more correct user answers and less

sis of discourse segments in direction-giving mono-

logues. In Proc. of ACL.

speech recognition problems in the NM version.

E. Hovy. 1993. Automated discourse generation using

These findings motivate future work in under-

discourse structure relations. Articial Intelligence,

standing the effects of the NM. We would like to

63(Special Issue on NLP).

continue our objective metrics analysis (e.g. see if

D. Litman and S. Silliman. 2004. ITSPOKE: An intelli-

users are better in the NM condition at updating

gent tutoring spoken dialogue system. In Proc. of

their essay and at answering questions that require

HLT/NAACL.

combining facts previously discussed). We also

S. Oviatt, R. Coulston and R. Lunsford. 2004. When Do

plan to run an additional user study with a be-

tween-subjects experimental design geared towards

We Interact Multimodally? Cognitive Load and Mul-

timodal Communication Patterns. In Proc. of Interna-

objective metrics. The experiment will have two

tional Conference on Multimodal Interfaces.

conditions: NM present/absent for all problems.

R. Passonneau and D. Litman. 1993. Intention-based

The conditions will then be compared in terms of

segmentation: Human reliability and correlation with

various objective metrics. We would also like to

linguistic cues. In Proc. of ACL.

know which information sources represented in the

H. Pon-Barry, K. Schultz, E. O. Bratt, B. Clark and S.

NM (e.g. discourse segment purpose, limited hori-

Peters. 2006. Responding to Student Uncertainty in

zon, correct answers) has the biggest impact.

Spoken Tutorial Dialogue Systems. International

Journal of Artificial Intelligence in Education, 16.

Acknowledgements

C. Rich and C. L. Sidner. 1998. COLLAGEN: A Col-

This work is supported by NSF Grants 0328431

laboration Manager for Software Interface Agents.

and 0428472. We would like to thank Shimei Pan,

User Modeling and User-Adapted Interaction, 8(3-4).

Pamela Jordan and the ITSPOKE group.

M. Rotaru and D. Litman. 2006. Exploiting Discourse

Structure for Spoken Dialogue Performance Analy-

sis. In Proc. of EMNLP.

References

M. Walker, D. Litman, C. Kamm and A. Abella. 2000.

K. Acomb, J. Bloom, K. Dayanidhi, P. Hunter, P.

Towards Developing General Models of Usability

Krogh, E. Levin and R. Pieraccini. 2007. Technical

with PARADISE. Natural Language Engineering.

Support Dialog Systems: Issues, Problems, and Solu-