Meeting Agenda Meeting Agenda I will be present some guidelines on giving talks, giving "Do's" as well as "Don'ts". I will illustrate my points by giving two (very!) short versions of a talk on Computer-Aided Violin Design. The points I cover in my talk are discussed in this WWW page.
This talk provides an overview of database research in the Exploratory Database Systems Department at the Almaden Research Center. We are developing database systems with features that will dramatically improve the functionality and performance of tomorrow's DBMSs in a number of ways. Sophisticated type systems (inspired by object-oriented programming languages) and better semantic models for data will incorporate more meaning into the database without sacrificing the advantages of relational technology. DBMSs won't just store large data objects, such as images and voice, but will also model their semantics and relate their contents. Techniques for integrating heterogeneous data will allow DBMSs to become powerful "information brokers" that organize the wealth of machine-readable data that bombards us daily. Triggers and stored procedures will consolidate within the database system more behavior, control, and reasoning capabilities. Applications such as work-flow management, data mining, and data warehousing will exploit these features and require further extensions to DBMS access paths, transaction models, etc. Many of these exciting enhancements have already been incorporated into IBM products such as DB2, or are being prototyped; others are still in the early stages of design.
Date: 22 September 1995
Time: 1:30 p.m.
Place: Davis Centre Room 1304
Currently, documents are typically represented in one of two forms :
In our research, we studied how to exploit a hybrid text and relational model to support document management. We describe database design trade-offs involving the partitioning of information between the text and relational database components. With an appropriate design, the advantages of both models can be exploited, while the shortcomings of using them individually is diminished.
By investigating the various design trade-offs, we propose a set of guidelines for partitioning data, based on their contents, structure, and anticipated usage. Some of these partitioning criteria have been verified in our prototyping work.
DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES
ARTIFICIAL INTELLIGENCE SEMINAR
-Wednesday, September 20, 1995
Eduard Hovy, Information Sciences Institute of the
University of Southern California, will speak on "The
Multiplexity of Discourse".
TIME: 2:00-3:00 p.m.
ROOM: DC 1302 *NOTE ROOM CHANGE*
ABSTRACT
Two-year-olds learn to make sentences automatically,
and master complex constructions even before they are
six. But producing well-written and coherent
multiparagraph texts is something we have to learn
consciously, and something that many people have not
mastered even by adulthood.
One reason is that discourse is a multiplex phenomenon:
we communicate at several levels simultaneously. We
have to learn how the levels work, how to decide which
level to address at any time, how to signal to the
audience the articulation of the discourse structure,
and so on. However, despite the ubiquitousness of
discourse, we know surprisingly little about its actual
structure or the processes of planning and generating
it.
This talk briefly outlines some recent work in the
study of discourse structure and then focuses on two
questions: What levels of discourse does one need to
represent? How does one signal each level to the
audience? The talk explores how the semantic and
interpersonal information give rise to such notions as
Theme, Topic, Focus, Given, and New from Text
Linguistics and Rhetorical Relation, Discourse
Intention, and Rhetorical Structure from Computational
Linguistics.
The Grand Valley Section of the Canadian Information
Processing Society is pleased to present its first dinner
meeting of the 1994-95 programme:
Topic: Toxic Waste in your Backyard: can Computer
Simulation Help?
Speaker: Dr. P. A. Forsyth, Professor, Department of
Computer Science, UW and Director, Institute
for Computer Research
Date: Wednesday, 20 September 1995
Time: 4:30 registration and light lunch
5:30 presentation
Place: Davis Center 1301 and 1302, University of Waterloo
Cost: CIPS members: free
non-members: $5.00
In the US it is estimated that the cost of cleaning up toxic
waste dumps is between 300 billion and one trillion dollars.
More locally, Kitchener-Waterloo is the largest municipal
region in Canada which depends on groundwater. There have been
several examples in recent years of industrial contamination of
our water supply.
Computer simulation can be used to design processes to remediate
hazardous waste sites in a safe and cost-effective way. This
talk will explain the basic principles used to simulate toxic
waste site remediation, and describe how engineers and scientists
use simulation to help clean up groundwater contamination
problems. Since simulations typically produce hundreds of megabytes
of data, visualization is a powerful tool used to extract useful
information from all this data.
An animated visualization of a simulation will be presented, and
the limitations of current visualization technology will be
discussed.
If possible, please confirm attendance with John Porr,
Nichols-Radtke Limited, by voice mail at (519)740-7479 x363,
by Fax at (519)621-8437, or by email at JohnP@nrl.com as soon
as possible.