LINGUIST List 11.1241

Fri Jun 2 2000

Jobs: NLP Researcher/Programmer, U of Pittsburgh

Editor for this issue: Naomi Ogasawara <naomilinguistlist.org>

Directory

1. Pamela Jordan, NLP: AI Researcher/Programmer at University of Pittsburgh

The Atlas and Why2000 projects seek a full-time AI
researcher/programmer to participate in the design, implementation and
evaluation of two advanced natural-language intensive tutoring
systems; one that supplements the existing Andes quantitative physics
problem solving coach and one that provides qualitative physics
problem solving tutoring. The primary duties for this position are
leading the design and implementation of the underlying knowledge
bases and inferencing capabilities needed for both systems.

The Atlas project's goal is to engage students in dialogue to tutor
them about applicable physics rules and concepts whenever the student
asks for assistance while attempting to solve a quantitative physics
problem. The dialogue system will attempt to have the student recall
the physics rule he needs to apply for the current problem using a
variety of techniques (e.g. leading him to derive the rule or simply
reminding him of the rule if it is one the dialogue component recently
discussed with him). The primary knowledge that needs to be developed
for this project is a physics-based ontology to support the
understanding and generation components of the system and support for
matching student responses to expected correct and incorrect
responses. Student contributions are typically short answers or
requests.

The Why2000 project's goal is to have students provide natural
language answers to qualitative physics problems that are then
followed up with a dialogue in which the tutor provides feedback and
attempts to correct recognizable misconceptions evident in the
student's answer. The primary knowledge that needs to be developed
for this project are representations for expected answers, limited
causal connections between events, and recognition of how the
student's answer matches with expected answers. In this case, the
student's initial answer is expected to contain multiple sentences and
inferencing may be necessary to determine if the student's answer is
coherent and complete.

Although the knowledge representation issues are challenging, the
problems are limited by the fact that the tutoring systems are meant
to supplement typical classroom instruction. We can assume that the
students will "talk physics" in non-creative ways. The problem is
further circumscribed by the fact that the system provides the physics
problems that the student will attempt to solve and does not have to
be able to tutor physics problems the student presents to it.

The qualifications for this position are at least an M.S. in Computer
Science or equivalent with an emphasis in either Knowledge
Representation or Natural Language Processing and experience in
designing and implementing Knowledge Bases in support of research and
development applications. Although the position does not require
prior expertise in physics, the applicant should be willing to learn
about the physics concepts involved. This position is located at the
University of Pittsburgh in the Learning Research and Development
Center. It is a full-time research staff position with full benefits.
Salary will be commensurate with experience. Send inquiries and CVs
or resumes to Pamela Jordan (pjordanpitt.edu).

The Atlas and Why2000 groups are composed of artificial intelligence
researchers, computational linguists and psychologists. Atlas is led
by Kurt VanLehn at the University of Pittsburgh
(http://www.pitt.edu/~vanlehn) while Why2000 is jointly led with Art
Graesser at the University of Memphis
(http://mnemosyne.csl.psyc.memphis.edu/home/graesser/). Both projects
are part of CIRCLE, an NSF-funded research center that studies human
and computer tutoring (see http://www.pitt.edu/~circle).