==============================
REPORT FOR JULY-SEPTEMBER 2007
==============================

TECHNICAL ACCOMPLISHMENTS

() Completed the work on an initial mechanism for the learning of
"common-sense" assumptions about resources and scheduling constraints,
and began work on a more advanced mechanism.

() Completed the work on an initial mechanism for identifying clusters
of similar rooms, which improves the learning of default room properties.

() Developed an advanced search mechanism for the integration of ST
Module with CM Radar, based on the analysis of possible schedule
improvements after specific room requests.


IMPROVEMENTS TO PROTOTYPE

() Began work on a procedure for the automated placement of vendor
orders through the vendor portal, which will support the selection and
placement of vendor orders with minimal user participation.

() Completed the work on a new GUI for editing of control parameters;
also, implemented an initial version of a new GUI for the construction
of STP world models, and began work on its improved version.

() Thoroughly tested the overall Space-Time Module and Space-Time GUI,
and fixed the remaining minor bugs.


PROJECT PLANS

() Continue work on advanced learning mechanisms, which include the
learning of (1) common-sense assumptions, (2) elicitation strategies,
and (3) correspondence between conference needs and available goods
and services.

() Begin work on the task-coordination mechanism, which is part of the
work headed by Stephen Smith; specifically, develop procedures for
ranking of advice, conflict resolution, and elicitation of additional
related data.

() Prepare publications on the main Year-Three results.


PROGRESS AGAINST PLANNED OBJECTIVES

We have accomplished most of the planned objectives, which included
(1) implementing an initial mechanism for the learning of common-sense
assumptions; (2) developing an initial mechanism for identifying
clusters of similar rooms; (3) completing the work on the advanced
algorithm for integrating ST Module with CM Radar; (4) completing the
work on the new GUI for editing control parameters; and (5) evaluating
the elicitor performance under various conditions. We have not yet
integrated the newly developed mechanisms with the rest of the system.


OBJECTIVES FOR NEXT PERIOD

() Develop an advanced mechanism for the learning of common-sense
assumptions about resources and scheduling constraints.

() Integrate the mechanisms for the learning of common-sense
assumptions, clustering of similar rooms, and analysis of schedule
improvements after specific room requests with the rest of the system.

() Develop an initial mechanism for the automated placement of vendor
orders, and evaluate its impact on the schedule quality.

() Begin work on a mechanism for the learning of the correspondence
between conference needs and available goods and services, which will
improve the effectiveness of the automated vendor-order selection.

() Begin work on the learning of elicitation strategies.

() Begin work on the task-coordination mechanism, including the
procedures for ranking of advice, conflict resolution, and elicitation
of additional data related to task-coordination strategies.

() Complete the work on an extended GUI for constructing STP world models.


PRESENTATIONS

We did not give any external presentations. On July 17, Eugene Fink
gave an internal presentation on the status of Space-Time Module and
Year-Four plans. On August 28, Eugene has given an internal
presentation about Space-Time work to incoming graduate students.


CONCERNS

The reduction of the Space-Time group has required to transfer code
and tasks from the outgoing programmers to the remaining programmers,
and the main related concern is smooth transfer of the code.