Objective
Enhancing Event, Temporal, and Location Reasoning on the Web:
WG A1 investigates common-sense forms of reasoning that are relevant
to most advanced and adaptive Web systems, i.e. reasoning with
events, time, and locations.
Back to Top
Description of Work
With geotemporal information we mean any kind of temporal
information that is based on some established human calendar system
plus real-life temporal reference points such as events ("last
Sunday", "begin of easter holidays", "WC-finals", "next
elections", etc.) seasons ("early spring", "late summer" etc.)
as well as epochs in history ("middle ages", "Roman empire",
"second world war" etc.). The kind of temporal information that we
want to address includes time points, intervals, durations and
periodicities ("every second year"). We speak of "geo"temporal
information since we do not focus on the temporal behaviour of
microsystems, algorithmic processes, etc.
With geospatial information we mean any kind of spatial information
that is based on an established system of geographic coordinates and
on real-life locations such as countries, cities, places, rivers,
etc. The kind of spatial information that we address includes
geographic positions, distances, directions, and relations such as
neighbourhood, inclusion etc.
With topical information we mean any orientation scheme that helps
to position information in a suitable hierarchy of thematic fields,
areas, domains and topics. Examples of general topics are "arts",
"sports", "entertainment", examples of specific topics are
"generation of electricity", "photographic equipment", "personal
security". Topical information is ordered and represented in
taxonomies and classification schemes such as, e.g., the UDC
(universal decimal classification).
Already in the actual Web we find an enormous amount of information
systems and services that deal with suitable combinations of
temporal, geographic and topical data of the above form. One
category of Web pages supports hotel booking offering suitable menus
for selecting towns, places within towns, travel dates, price
categories etc. Other Web pages offer informations about locations,
dates and thematic areas of trade fairs or scientific conferences in
various countries of the world. A third group of Web pages describe
transport enterprises and services in terms of the kind of items that
may be transported and the temporal and geographic restrictions that
must be respected. Yet another kind of services offers actual news,
following a systematics for topical and geographic filtering. In the
present form, all these systems are designed or the human user. In
most cases, automated access to the temporal, spatial and topical
information is difficult or impossible, and for specifying temporal,
geospatial and thematic conditions, only a small number of predefined
choices ordered in menus exist.
A substantial progress for automated processing of Web contents -
such as envisaged in many scenarios for the Semantic Web - can be
expected from special reasoning mechanisms that are able to deal with
temporal, geospatial and topical data. Most importantly, these
mechanisms should satisfy two requirements: First, they should
support a flexible matching and translation between temporal, spatial
and topical conditions that are expressed in different ways. This
task includes the interpretation of semi-formal and natural language
expressions. Second, they should offer algorithms for solving these
constraints. Matching formalisms provide an interface between
conditions or queries formulated by a human or robot client on one
side and properties/constraints/offers that are found in Web pages of
enterprises or inquiry offices on the other side. Constraint solvers
may then be used to establish compatibility of the requirements and
to enumerate possible solutions
. Furthermore,
matching mechanisms and special constraint solvers for geotemporal,
geospatial and topical information may be used to enhance
"syntactic" mechanisms of general query languages for the Web - as
they will be developed in REWERSE - with flexible "semantic"
reasoning facilities for a broad range of applications.
Formal ontologies that collect and order distinct temporal and
spatial notions and relate concepts represent an important
prerequisite for common-sense geotemporal and geospatial reasoning,
in particular they are indispensable for the above mentioned matching
and translation techniques. Obviously, conventional temporal and
spatial coordinate systems represent a backbone for these ontologies.
But geotemporal and geospatial references found in Web pages and
inquiries are often not expressed in explicit coordinates. Most
references to locations are based on "named geo-entities" such as
countries, cities,places, streets, oceans, rivers etc. Similarly
temporal reference is often based on named singular or periodical
events. Examples are calendar events (easter, Christmas, ramadan),
holidays, political events (elections), sport events (world
championships), scientific events and business events ("immediately
after CADE", "two weeks before the IAA"). Events of the latter
type also carry important topical information. Hence formal
ontologies are needed that yield a bridge between standardised
calendar systems and geo-coordinate systems on the one hand and
temporal or locational expressions that are found in natural
language, anchor texts or element names on the other hand. For
topical classification of Web pages, besides events also other named
entities can be important indicators (the ACM, the UN, BMW,
etc.).
In Workpackage A1, suitable ontologies for geotemporal and
geospatial notions and concepts will be built. On this basis,
matching and constraint solving mechanisms for temporal and spatial
reasoning will be realized. A thesaurus for events will be
constructed that describes real-world events and positions them in a
temporal-spatial-topical hierarchy. The construction of the hierarchy
is a part of this task. A Web-based appointment scheduling system
will be realised for proof-of-concept of the ontologies and reasoning
mechanisms.
Back to Top