Derek Sleeman
Contact details:
Research
Professor
The
Computing Science Department, Room 238A, The AB24 3FX, Email: d.sleeman@abdn.ac.uk Phone: +44
(0)1224 272288
Computing Science Department,
The
Awards
/ Honours:
Fellow of
Royal Society of Edinburgh, 1992
Fellow of The British Computing Society, 1995
Fellow of the European AI Societies, 2004
Some
Recent Professional Activities:
Program Committees for European (ESWC) &
International (ISWC) Semantic Web Conferences
Program Committee European & International
Conferences on Knowledge Acquisition / Capture
Steering Committees for EKAW & KCAP.
Conference Chair of KCAP 2007, Whistler,
Co-Chair of AAAI “Symbiotic Relationship between
Semantic Web & Knowledge Engineering”, Stanford, Spring
2008.
Publications: up to 1986 & 1986 onwards
Former
Current Research Students
Current
Research Activities
The
activities of my Research group fall approximately under 2 main headings,
namely:
Semantic
Web & Knowledge Engineering (SWKE)
Intelligent
Data Analysis (IDA) of Real-time
(Medical) Datasets
Where the SWKE activity is having a
significant impact on the IDA activities. For example, in some of the IDA systems we have implemented
recently, medical knowledge has been represented as OWL ontologies.
This, of course, makes it much easier to use the tool with further (medical)
domains.
Semantic Web & Knowledge Engineering (SWKE)
This activity can be further divided into 2
activities namely:
Ontology
Search, Consistency checking, and Ontology Creation & Evolution
The
Reuse of Knowledge Bases (& the creation of KBSs
from components)
Ontology Search, Consistency checking,
and Ontology Creation & Evolution
Here we have developed a
set of tools which allow the large number of ontologies available on the
Semantic Web to be discovered and reused for other applications (both in the
Semantic Web and by the larger knowledge engineering community). Firstly, the
ONTOSEARCH2 system addresses the problem of finding existing ontologies with
given characteristics. Secondly, we have developed tools which allow the
knowledge engineer to detect and repair errors which occur in the ontology’s
vocabulary (CleOn) or which produce logical
inconsistencies (RepairTab). Once a knowledge
engineer is satisfied that the ontology is lexically and semantically
consistent, then he/she can extend the ontology if the application requires
that. Once the extension has been done, it again would be prudent to use those
same tools to check that the resulting ontology is consistent.
References: For an overview of this
activity, see: Thomas, Sleeman et al (2008) The Aberdeen University Ontology Reuse Stack. In
Proceedings of the AAAI-08 “Semantic Web & Knowledge Engineering”
Symposium, Derek Sleeman & Mark Musen (Eds).
AAAI Press.
For more detailed references about ONTOSEARCH2, CleON & RepairTAB see
references in the above paper or my www page.
The Reuse of Knowledge Bases (& the
creation of KBSs from components)
The benefits of reuse
have long been recognized in the knowledge engineering community where the
dream of creating knowledge based systems (KBSs)
on-the-fly from libraries of reusable components is still to be fully realised.
In this paper we present a two stage methodology for creating KBSs: first reusing domain knowledge by mapping it, where
appropriate, to the requirements of a generic problem solver; and secondly
using this mapped knowledge and the requirements of the problem solver to
“drive” the acquisition of the additional knowledge it needs. For example, suppose
we have available a KBS which is composed of a propose-and-revise problem
solver linked with an appropriate knowledge base/ontology from the elevator
domain. Then to create a diagnostic KBS in the same domain, we require to map relevant information from the elevator knowledge
base/ontology, such as component information, to a diagnostic problem solver,
and then to extend it with diagnostic information such as malfunctions,
symptoms and repairs for each component.
We
have developed MAKTab, a Protege
plug-in which supports both these steps and results in a composite KBS which is
executable.
References: For an overview of this
work see: Corsar, D., and Sleeman, D.
2007. KBS Development through Ontology Mapping and Ontology
Driven Acquisition. K-CAP ’07:
Proceedings of the 4th international conference on Knowledge capture,
23–30.
See
David Corsar’s website for further details
Intelligent Data
Analysis (IDA) of Real-time (Medical) Datasets
Background:
Patient monitoring equipment can be found in many clinical settings, including
Dialysis & Intensive Care Units (ICUs). Typically, a dozen or so parameters
are recorded every hour, but in some circumstances the frequency of the data
collection can be greatly increased. Currently, this information is reviewed by
the clinical staff to determine the patient’s treatment, but very frequently
this information is not analysed for interesting trends & for that matter inconsistencies.
These datasets are major resources which should be subjected to much more
sustained and systematic analyses. Potentially the amount of information which
can be collected is huge; for example, we estimated that a 10-bed ICU is
capable of collecting up to a trillion pieces of information per year. Given
these amounts of data we realized it was essential to produce
appropriate software systems (infra-structure) to support the clinician /
analysts in preparing datasets to enable hypotheses to be investigated. We have
recently developed one such infrastructure, ACHE (Architecture for Clinical
Hypotheses Examination).
To evaluate the initial version of ACHE, a study to
detect Acute Myocardial Infarctions, was conducted with data from Glasgow Royal
Infirmary’s Intensive Care Unit (ICU). Initial results from the study are very
encouraging and ACHE substantially reduced the time required to perform the
study.
Reference:
For an
overview of this work see: Laura Moss, Derek Sleeman,
To date, we have been working with primarily 2
groups of clinicians. Namely, the Renal Unit at Aberdeen Royal Infirmary
(Consultant; Dr Nick Fluck) & the ICU at Glasgow Royal Infirmary (Professor
Analysis of ICU Datasets (
Current projects with the Glasgow ICU include:
Cardiac
Study
Can minor cardiac “events”
be detected?
Do these occur in only a
subset of patients and can they be predicted?
Laura
Moss, Derek Sleeman,
Dialysis
Study
How do ICU patients
respond to dialysis?
Does dialysis cause
patients to become haemo-dynamically unstable?
Improving
ICU Scoring Systems
Enhanced traditional
scoring system (APACHE) to take into account clinical interventions such as
fluid infusions & drugs; we have introduced a 5-point scale (A-E) where “A”
is normal physiology & “E” is
significantly deranged.
Conducted an empirical
study to determine to what extend clinicians from a single ICU classify their
patients (on the A-E scale) consistently. (Developed some novel technology, the
INSIGHT system, to support this investigation.)
Derek Sleeman, Andy
Aitken, Laura Moss, John Kinsella, Malcolm Sim. (2009) A System to
Detect Inconsistencies between a Domain Expert’s Different Perspectives on
(Classification) Tasks Janusz Kacprzyk
(ed), Advances in Machine Learning II, Studies in
Computational Intelligence: pages 293-314. Springer
Derek Sleeman,
Andy Aiken, Laura Moss, Martin Hughes, John Kinsella, & Malcolm Sim (under
review) Detecting & resolving inconsistencies between domain experts’
different perspectives on (classification) tasks
Study
with Consultants to detect time-periods when patients are reacting anomalously
to treatment; collect experts’ explanations for these sequences. Build a
knowledge-based system to replicate these explanations.
Laura Moss, Derek Sleeman, Malcolm Booth, Malcolm Daniel, Lyndsay
Donaldson, Charlotte Gilhooly. Explaining Anomalous Responses to Treatment
in the Intensive Care Unit C Combi, Y Shahar, A Abu-Hanna (ed),
Proceedings of the AIME-2009 Conference (
Laura Moss, Derek Sleeman, Malcolm Sim, Malcolm Booth, Malcolm Daniel, Lyndsay Donaldson. Ontology-Driven Hypothesis Generation
to Explain Anomalous Patient Responses to Treatment Max Bramer,
Richard Ellis, Miltos Petridis (ed), Proceedings of AI-2009 (R&D in
Intelligent Systems XXVI) (
CBR
system: Uses actual patient datasets collected in this ICU; & gives advice
about a new patient (typically one which is behaving atypically)
Analysis of
Renal Datasets (
The Research agenda set by the clinician /
analysts includes:
Are there distinct groups of patients with specific physiological
characteristics?
Can major intra-dialytic complications such as
hypotension be predicted?
What are the major differences between patients who are stable on
dialysis and those that are not?
Do certain modes of dialysis suit some patients better than others?
Is it possible to create
patient specific optimum dialysis strategies?
We
have carried out a pilot study to investigate the first issue using primarily
data collected from the dialysis machines. Clustering analysis techniques
detected the same 3 groups of patients (young, elderly, elderly
with renal & cardio-vascular problems) as the clinician had identified. We
used 2 different techniques to confirm our clusters.
Secondly,
we have investigated the second issue (i.e., predicting the most commonly
occurring problem with dialysis patients namely hypotension.) The number of
patients was small & so this is only an encouraging but tentative result.
The
datasets for these studies were prepared “manually”. Now that ACHE is available
we plan to rerun these studies with much larger data-sets.
Reference: For more details of these dialysis studies see: Derek Sleeman, Nick
Fluck,
Earlier Activities (arranged in alphabetical & not chronological
order)
Co-operative
Knowledge Acquisition & Knowledge Refinement Systems.
We built a variety of systems which were able to
acquire & refine knowledge expressed in a number of different knowledge
representation schema, including:
KRUST / STALKER
(“classical rules”)
REFINER (cases)
ReTax (Taxonomies)
CRIMSON (Constraints)
TIGON (Qualitative
models)
R4 / R5 (cases)
Cognitive
Studies of Student’s / Pupils misunderstanding
This activity underpinned my work on building
Intelligent Tutoring Systems (ITSs). In fact I felt
strongly that before embarking on building an ITS in a
subject domain it was very important to understand the nature of the students’
misunderstandings in this domain. Our main studies were carried out in Algebra,
Arithmetic & Programming.
Intelligent
Tutoring Systems
The culmination of this early career activity was
a book, co-edited with John Seely Brown, Intelligent Tutoring Systems,
published in 1982 by Academic Press. (See my references for further
details).