The HySpirit Retrieval Platform
Thomas Rölleke1,2
roelleke@HySpirit.com
thor@dcs.qmw.ac.uk
Ralf Lübeck1
luebeck@HySpirit.com
Gabriella Kazai2
gabs@dcs.qmw.ac.uk
1
2
HySpirit GbR, Postfach 300258, D-44232 Dortmund, Germany
Queen Mary University of London, Department of Computer Science, London, E1 4NS, England
Between 1993-1999, the foundations of the retrieval platform
HySpirit (HYpermedia System with Probabilistic Inference for the
Retrieval of InformaTion) were developed at the University of
Dortmund. Since 1999, HySpirit has been used as a research tool
and as a commercial product in selected applications.
The main purposes of the HySpirit technology are:

Provide flexible abstraction layers (relational, logical and
object-oriented) for modelling hypermedia and knowledge
retrieval.

Exploit the logical structure of documents for retrieving best
entry points into documents.

Support knowledge-oriented retrieval in semi-structured and
heterogeneous data sources.

Integrate fact-oriented and content-oriented searching and
browsing.

Support retrieval strategies exploiting relationships (spatial,
temporal, semantic etc.).

Consider intrinsic features of knowledge such as uncertainty,
incompleteness and inconsistency.

Allow for easy parameter setting to reflect user preferences
(profiles) and support retrieval experiments and evaluation.

Extend database models with probability theory in order to
create a retrieval platform possessing the advantages and
expressiveness of database models, and the aggregation of
uncertain evidence as defined by probability theory.
Currently, the platform is applied in research projects on
structured document retrieval (XML) and interactive TV
(MPEG-7). Also, the system is installed in a commercial
environment, where various information needs occur in a highly
heterogeneous and complex information environment.
The HySpirit system consists of a number of modules including
the core retrieval engine, knowledge modelling languages,
indexers, analysers, statistical tools and data management
connectors. This modular structure supports distributed
development and a “plug-in” architecture.
HySpirit has four main abstraction layers:
POOL: Probabilistic Object-Oriented Logic ([3])
FVPD: Four-Valued Probabilistic Datalog ([2])
PD: Probabilistic Datalog ([4])
PRA: Probabilistic Relational Algebra ([1])
The POOL layer provides object-oriented concepts such as
classification, association, and aggregation for modelling
knowledge for the purpose of retrieval. In addition, POOL
includes the content dimension in combination with objectoriented modelling.
FVPD and PD are the logical layers providing rules and recursion,
where FVPD comes with special features for representing
incomplete and inconsistent knowledge.
The core of the HySpirit system is its PRA retrieval engine. This
layer provides the widely known relational paradigm, which
recently has been extended with an SQL interface.
All abstraction layers provide the key concept of HySpirit:
representation of knowledge and its intrinsic uncertainty with
database models and probability theory to achieve a descriptional
approach for modelling complex information retrieval tasks such
as hypermedia and knowledge retrieval.
Further information about HySpirit can
www.hyspirit.com and qmir.dcs.qmw.ac.uk.
be
found
at
REFERENCES
[1] N. Fuhr and T. Rölleke. A Probabilistic Relational Algebra
for the Integration of Information Retrieval and Database
Systems. ACM Transactions on Information Systems, 14(1):
32–66, 1997.
[2] N. Fuhr and T. Rölleke. HySpirit - a Probabilistic Inference
Engine for Hypermedia Retrieval in Large Databases.
Proceedings of the 6th International Conference on
Extending Database Technology (EDBT), Valencia, Spain,
Lecture Notes in Computer Science, 24–38, Berlin et al.,
1998. Springer.
[3] T. Rölleke. POOL: Probabilistic Object-Oriented Logical
Representation and Retrieval of Complex Objects. Shaker
Verlag, Aachen, 1999. Dissertation.
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that
copies bear this notice and the full citation on the first page. To copy
otherwise, or republish, to post on servers or to redistribute to lists,
requires prior specific permission and/or a fee.
Conference ’00, Month 1-2, 2000, City, State.
Copyright 2000 ACM 1-58113-000-0/00/0000…$5.00.
[4] T. Rölleke and N. Fuhr. Information Retrieval with
Probabilistic Datalog. In F. Crestani, M Lalmas and C.J.
Rijsbergen, eds, Uncertainty and Logics - Advanced Models
for the Representation and Retrieval of Information. Kluwer
Academic Publishers, 1998.