SPONSORS:
MITRE
ACL SIGMEDIA
Automatic Content Extraction (ACE) Program
INVITED SPEAKERS:
Barbara Tversky, Stanford University, USA
Fabio Pianesi, ITC-IRST, Italy
PROGRAM COMMITTEE:
Elisabeth Andre DFKI Germany
Myriam Bras IRIT France
Rob Gaizauskas Sheffield UK
Udo Hahn Freiburg University Germany
Eduard Hovy USC-ISI USA
Grard Ligozat LIMSI-CNSRS France
Ruslan Mitkov University of Wolverhampton UK
Marc Moens University of Edinburgh UK
Dragomir Radev University of Michigan USA
Ellen Riloff University of Utah USA
Laure Vieu IRIT France
Clare Voss US Army Research Lab USA
Michael White Cogentex USA
Janyce Wiebe University of Pittsburgh USA
George Wilson MITRE USA
Cornelia Zelinsky-Wibbelt Hannover Germany
CONFERENCE WEBSITE:
http://epsilon3.georgetown.edu/ discours/spacetime.html
i
PREFACE
Temporal and spatial information is ubiquitous in natural language, yet many challenging
computational issues are relatively unexplored. While corpus-based methods have accelerated
progress in other areas of NLP, they have yet to make a substantial impact on the processing of
temporal and spatial information. Evaluations of information extraction tasks as in the Message
Understanding Conferences have revealed some of the challenges inherent in the temporal and
spatial portions of those tasks as well as some encouraging results from the evaluated systems.
Current commercial interest and funding initiatives in this area make this an excellent time to revisit
these problems. Finally, there are sizeable communities of people working in temporal databases,
knowledge representation, and theoretical computer science who are looking for real problems to
work on; spatial and temporal phenomena in natural language offer an abundant supply of such
problems.
A variety of different application areas stand to benefit from progress in spatial and temporal
information processing. These areas include:
 machine translation (e.g., translating temporal and spatial references)
 question answering (e.g., answering ”when” or ”where” questions)
 information extraction (e.g., normalizing values in spatial and temporal databases,
disambiguating place names using a gazetteer)
 summarization (e.g., producing temporally coherent summaries of multiple documents, or
generating route plans)
 information retrieval (e.g., indexing broadcast news by event time)
 information visualization (e.g., constructing event chronologies, geospatial visualization)
 multimodal interfaces (e.g., interfaces to simulations, gesture and speech input graphical
applications, navigation systems)
 interfaces to spatial and temporal databases (e.g., normalizing temporal and spatial
references)
 multimedia presentations (e.g., generating textual descriptions or captions, scene and route
descriptions, generation of spatio-temporal maps).
Addressing both temporal and spatial domains in a single workshop can be useful because of
the many issues in common across these domains. A researcher working in one of these domains
can often benefit from being aware of approaches in the other. For example, systems that process
temporal or spatial information need to deal with ’absolute’ references (”November 18, 1999”,
”Toulouse”), as well as relative references such as indexicals (”now”, ”here”), and references stated
in terms of offsets from some position in a frame of reference (”two weeks ago”, ”thirty miles north
of Paris”). Both temporal and spatial representations need to address indeterminacy and vagueness,
including non-specific references (”some time in June”, ”a town in Provence”), vague predicates
(”nearly a year ago”, ”near Dusseldorf”), and fuzzy boundaries (”Tuesday morning”, ”southern
England”). There are also many parallels between the way events are characterized in time and
objects are characterized in space. For example, events can be described relative to some point or
interval in time (e.g., ”I met John yesterday. He was crossing the street.”) while objects in space
can be described in relation to some place, object, or in terms of movement (e.g, ”The cup was on
top of that. It fell off.”).
This workshop brings together researchers representing spatial and temporal information in
natural language. The 14 papers presented here were selected by the Program Committee from 24
ii
submissions. These papers, which represent work by scientists in 9 countries, provide a variety of
interesting perspectives in areas such as:
 the semantics of temporal adverbs and spatial prepositions
 application of models of tense, aspect, modality, and discourse structure to linguistic analysis
 representation of reference frames and geometric, cognitive, and functional relationships
between objects
 automatic annotation methods and annotation schemes for representing events, times, and the
relations between them
 systems which integrate together spatial and temporal representations.
It is especially encouraging to note that these papers rely on evidence drawn from a variety of
different natural languages.
We would like to thank the authors, whose significant contributions made the workshop
possible, and the members of the Program Committee, who assisted in reviewing papers and guiding
the overall organization of the workshop. We are especially grateful to the invited speakers Barbara
Tversky and Fabio Pianesi. Thanks are also due to Anthony Davis, Keith Miller, and Leo Obrst, who
provided assistance in reviewing, and ACL representatives Rebecca Bruce, Patrick Saint- Dizier,
and Roberto Zamparelli, who helped make it all happen.
Lisa Harper, Inderjeet Mani, Beth Sundheim
Organizers
iii
