|
|
1.
LifeLines for Medical Patient Records (Catherine Plaisant/Ben
Shneiderman - UMD, and John Karat -IBM Research):
This project has resulted in a
refinement of LifeLines interface prototype which is currently being tested with
clinical data from Kaiser Permanente Colarado. LifeLines was included
into a recent bid by IBM to redesign the Kaiser Permanente national
health care information system.
2.
Speech-Based Information Retrieval for Digital Libraries (Doug Oard-UMD,
and Salim Roukos- IBM Research):
The initial phase of this new project
involved close interactions between the group at Maryland and several of the
developers of the speech recognition, speaker segmentation and
speaker identification software at Watson. Results of a preliminary user
study of the interface design was completed and
presented at a conference in June. In addition, Aaron Hallberg,
an undergraduate honors student, has worked on alternative
approaches for audio indexing at Watson this past summer at the
invitation of Salim Roukos. The next phase is to
start our high-volume digitization effort, given that we have developed
an understanding of the technical requirements for compatibility with
the speech recognition software through discussions at Watson.
3.
Performance Tools for Parallel Systems (Jeff Hollingsworth-UMD, and Doug
Pase-IBM Poughkeepsie):
This collaboration has focused on the
development of a standard for runtime code patching (inserting new
code into a running program). IBM is working towards a product, called
DCPL, based on the results of this collaboration. The Beta version of
the code has been released and has been installed at several major
supercomputing facilities.
4. Flexible
Distributed Query Processing (Mike Franklin-UMD, and Mike
Carey-Almaden):
This project has recently involved a study (with
Mike Carey and Markos Zhariodakis at Almaden) in which experiments with
different distributions of database function across clients and
servers were conducted. Donal Kossmann, who worked as a postdoc at Maryland,
went on to participate in the Garlic project at Almaden, and was
instrumental in developing IBM approach to distributed query optimization.
He also worked with Mike Carey on the development of a special class of
query optimizations for use in DB2. In addition, Tolga Urhan, a graduate student
from Maryland, spent a summer working with Don Chamberlin at Almaden to
support extensions to SQL in DB2.
5. Tools
for Data-Intensive Computations with Applications to Remote Sensing
(Joseph JaJa-UMD, and Harry Hulen/Bob Coyne-IBM/Houston and Chung-Sheng
Li-Watson):
This collaboration has resulted in a deployment of a pilot SP-based
archiving and processing system that includes the High Performance
Storage System (HPSS) software which has been running for a couple
of months using test data from the NASA Landsat and NOAA AVHRR sensors.
We are continuing our close coordination with the HPSS technical staff
at Houston as we move into production system. We will instrument our
system and conduct tests to help us in our modeling efforts for
the main factors affecting performance as well as for
developing a better characterization of data processing requirements of
earth system applications over a networked high performance computing
environment. Another promising direction is our collaboration through the
newly created Earth Science Information Partnership (ESIP) program at NASA in
which both IBM (Chung-Sheng Li) and UMD (JaJa and Townshend) are participating.
The goal of the UMD ESIP is to develop an integrated data processing,
retrieval, and visualization system for land cover studies, while the
IBM ESIP project is focusing on the progressive mining of remotely sensed data
for environmental and public health applications. Since the two
projects are complementary, we expect our collaboration to be mutually
beneficial and to lead to an exchange of tools and data.
|
|
