Resume: Peter Edward Clark

Last updated: December 1998.

Personal Details:

Present Position:

Research Scientist

Knowledge Systems Group

Boeing Research and Technology

PO Box 3707, m/s 7L-66

Seattle, WA 98124-2207, USA

Education:

1. PhD in Computer Science (1991)
   • Turing Institute, Dept of Computer Science, Strathclyde University, UK
   • Research area: Knowledge-Based Systems/Artificial Intelligence
   • Thesis: A Model of Argumentation and its Application in a Co-operative Expert System (Abstract)
2. MSc in Information Technology: Knowledge-Based Systems (1985)
   • Dept of Artificial Intelligence, Edinburgh University, UK.
   • Thesis: Towards an Improved Domain Representation for Planning (Abstract)
3. BA in Physics (1984)
   • St. John's College, Oxford University, UK
   • First Class Honors (top 8% UK graduates)

Experience:

1. (Sept 1996-Present): Research Scientist at Boeing Research and Technology. I am currently working on the construction of reusable, component-based representations of knowledge, and their use by both end-users (eg. for question answering) and existing software systems. This research extends and applies the component-based approach I developed earlier with Bruce Porter at Univ Texas at Austin. This is being applied in three contexts:
   • Neutral Representation: Construction and reasoning with logic-based representations of design knowledge, specifically focussing on design rules for hydraulic tubing.
   • Knowledge-Based Training: The development of two prototypes providing knowledge-based question-answering services (one for 737 final body join, one for a space shuttle payload experiment).
   • Information Access: Use of a large ontology for improving information search, in particular exploiting the notion of ``semantic distance'' between a search query and potential targets.
2. (Mar 1994-Aug 1996): Research Fellow in the Knowledge-Based Systems Laboratory, Univ Texas at Austin. Principle Investigator for DCE Help-Desk Assistant project (funded by Digital). Research in methods for constructing knowledge-bases from reusable, modular components.
   • I developed a theory of concept composition with Prof. Bruce Porter, by which representations of complex, domain-specific concepts can be assembled by composing general representational components rather than being hand-built from scratch. (For example: a representation of a database can be built from components modeling a container, a secure resource, and a service provider). A paper on this work won a AAAI'97 Best Paper award.
   • Development of the DCE Help-Desk Assistant, for automatically answering a proportion of customer questions which would otherwise have been phoned in to a normal help-desk. These include descriptive ("What is...?"), diagnostic ("Why did...?"), and planning ("How do I...?") questions. The Assistant is an interactive, knowledge-based system, and infers answers which are customized to the user and expressed in natural language. It uses a knowledge-base constructed the above concept composition technique.
   • I implemented the current inference engine for the KM Knowledge Representation language, underlying the above application and others. The implementation includes machinery for unification, automatic classification, and reasoning in different contexts.
   • I was a consultant to the Cyc project (1 day/week June-August 1996), and performed a detailed, two-month study of the knowledge representation language Algernon for Prof. Ben Kuipers at UT Austin.
3. (1992-1994): Research Associate in Knowledge Systems Lab, National Research Council of Canada (Ottawa). Research and development of knowledge-based systems.
   • I developed Electronic Trader (ET), a prototype for automatically detecting arbitrage opportunities from a live feed of financial data. The system uses AI techniques to heuristically search for profitable transaction sequences. This was developed for Canada's Export Development Corporation, and was in commercial use during 1994 and 1995.
4. (1991-1992): Postdoctoral fellow at Dept. of Computer Science, Univ. of Ottawa, Canada. Research in using domain knowledge to guide inductive learning.
   • I developed a theory of lazy partial evaluation (LPE) with Prof. Rob Holte, an enhanced learning technique which integrates methods of explanation-based learning and partial-evaluation. LPE avoids the overhead of full partial evaluation, and the undesirable `masking effect' of explanation-based learning.
   • I developed a technique for guiding inductive learning using a domain theory (in the form of a qualitative model), with Prof. Stan Matwin. This allows domain knowledge and training data to be integrated during learning, biasing inductive learning away from rules which are ``clearly non-sensical'' (with respect to the domain knowledge). This helps reduce the (normally high) overhead of manually `sanitizing' an induction system's output for practical applications.
   • I developed CABARESS, an inductive software tool for the Canadian Center for Remote Sensing, for classifying images from attribute-value data. The system integrated case-based, rule-based and statistical classification techniques.
5. (1985-1991): Research Scientist at Turing Institute, Glasgow, UK. Research and development in knowledge-based systems and machine learning.
   • I developed a computational model of argumentation, based on Toulmin's approach, by which an expert system and user could interact (`cooperatively argue') to jointly solve a problem. The approach provides a novel way of applying expert system technology in complex domains where experts often disagree, and knowledge changes with time. Through argumentation, the system alerts users to inconsistencies with their own and other experts previous decisions, and through the users' responses acquires additional domain knowledge. This was the topic of my PhD dissertation.
   • I was Principal Investigator and developer of Optimist, a full-scale, commercial AI system which implements the above theory. Optimist assists geologists in oil exploration, and was built for Enterprise Oil plc. The system was completed in 1989, and has been in commercial use since then, with extensions added in 1991 and 1994.
   • I developed a new rule induction algorithm, CN2, with Tim Niblett (1987). CN2 learns `if...then...' rules from training examples, and unlike similar systems available at that time handles noisy data well. CN2 is still extensively used and cited in the machine learning community.
   • I devised a logic-based formalism for representing arguments `for' and `against' a statement, and using case-based reasoning to resolve arguments based on previous precidents. The formalism applied determination theory to argumentation (arguments `determine' the truth of a statement), allowing domain knowledge and case-based reasoning to be combined.
   • I worked on the ESPRIT Machine Learning Toolbox (MLT) project. I developed a database of techniques and advice for applying rule induction technology to practical problems, for inclusion in the Advisor expert system module for this project.
   • I directed and supervised three small (3-month) knowledge-based system protoypes, for commercial clients:
     • A rule-based system for seat allocation on airline flights (British Airways plc)
     • FADES, a fault diagnosis system for process control (Yard Ltd). FADES uses a simulator to simulate example faults, and then automatically induces a diagnostic rule-base from the dataset thus generated.
     • A rule-based system for interpreting neurological data (Medelec Ltd)

Research Interests:

1. Knowledge-Based Systems:
   • Building and using representations of domain knowledge to assist with real-world problems such as tutoring, help-desk support, and natural language processing.
   • Component-based architectures for knowledge bases, so as to ease the problem of building and reusing representations.
2. Machine learning: Use of domain knowledge to guide inductive learning.

Awards

• AAAI'97 Best Paper award (Building Concept Representations from Reusable Components, co-authored with Bruce Porter)

Representative Publications:

• P. Clark and B. Porter. Building Concept Representations from Reusable Components. In AAAI'97, pages 369-376, CA, 1997. (Abstract and compressed postscript).

• P. Clark. Representing knowledge as arguments: Applying expert system technology to judgemental problem-solving. In T. R. Addis and R. M. Muir, editors, Research and Development in Expert Systems VII, pages 147-159. Cambridge Univ. Press, 1990. (Abstract and compressed postscript).

• P. Clark and S. Matwin. Using qualitative models to guide inductive learning. In P. Utgoff, editor, Proc. Tenth Int. Machine Learning Conference (ML-93), pages 49-56, CA, 1993. Kaufmann. (Abstract and compressed postscript).

Selected Additional Publications:

Knowledge Representation and Knowledge-based Systems

• P. Clark and B. Porter. Using Access Paths to Guide Inference with Conceptual Graphs. In Proc Int Conf on Conceptual Structures (ICCS'97), pages 521-535, Ed: Dickson Lukose et al, Seattle, WA, 1997. (Abstract and compressed postscript).
• Mike Uschold, Mike Healy, Keith Williamson, Peter Clark, Steven Woods. Ontology Reuse and Application. In Proceedings of the International Conference on Formal Ontology and Information Systems (FOIS'98), 1998. ( Abstract and compressed postscript).
• Mike Barley, Peter Clark, Keith Williamson, Steve Woods. The Neutral Representation Project. In Proc AAAI Spring Symposium on Ontological Engineering, Stanford, CA, 1997. (Abstract and compressed postscript).
• P. Clark and B. Porter. A Compositional Approach to Representing Planning Operators. Tech Report AI96-242, Dept CS, Univ Texas at Austin. (Abstract and compressed postscript).
• P. Clark. A Model of Argumentation and its Application in a Cooperative Expert System. PhD thesis, Strathclyde University, Glasgow, UK, 1991. (Abstract and compressed postscript [750k, 190pp]).
• C. Wood and P. Clark. FADES: An Expert System for Fault Analysis and Diagnosis. TIRM 87-024, Turing Institute, 1987. (Abstract).
• P. Clark. The Syntax vs. Semantics Debate Revisited?. In N. Cercone and G. McCalla, editors, Computational Intelligence, 9(4):366-367, 1993. (compressed postscript).
• P. Clark. Towards an improved domain representation for planning. Master's thesis, Edinburgh Univ., Edinburgh, UK, 1985. (Abstract).

Knowledge-Guided Machine Learning

• P. Clark and S. Matwin. Learning domain theories using abstract background knowledge. In P. Brazdil, editor, Proc. Sixth European Conference on Machine Learning (ECML-93), pages 360-365, 1993. Springer-Verlag. (Abstract and compressed postscript).
• P. Clark and R. Holte. Lazy partial evaluation: An integration of explanation-based generalisation and partial evaluation. In D. Sleeman and P. Edwards, editors, Proc. Ninth Int. Machine Learning Conference (ML-92), pages 82-91, CA, 1992. Kaufmann. (Abstract and compressed postscript).
• P. Clark. Knowledge representation in machine learning. In Y. Kodratoff and A. Hutchinson, editors, Machine and Human Learning. pages 35-49, London, 1989. Kogan Page. (Abstract and compressed postscript).
• P. Clark. Nonmonotonic reasoning, argumentation and machine learning. TIMLG-38, Turing Institute, Glasgow, UK, June 1990. (Abstract and compressed postscript).

Inductive Learning

• P. Clark and R. Boswell. Rule induction with CN2: Some recent improvements. In Y. Kodratoff, editor, Machine Learning - EWSL-91, pages 151-163, Berlin, 1991. Springer-Verlag. (Abstract and compressed postscript).
• P. Clark and T. Niblett. The CN2 Induction Algorithm. Machine Learning, 3(4):261-283, 1989. (Abstract and compressed postscript).
• P. Clark and T. Niblett. Induction in Noisy Domains. In I. Bratko and N. Lavrac, editors, Progress in Machine Learning: Proc. 2nd European ML Conference (EWSL-87). pages 11-30, Sigma, Wilmslow, UK, 1987. (Abstract and compressed postscript).
• P. Clark and T. Niblett. Learning if-then rules in noisy domains. In B. Phelps, editor, Interactions in Artificial Intelligence and Statistical Methods, pages 154-166. Gower, Hants, UK, 1987. (Abstract).
• P. Clark, C. Feng, S. Matwin, K. Fung. Improving Image Classification by Combining Statistical, Case-Based and Model-Based Prediction Methods. Technical report, Dept. of Computer Science, Univ. Ottawa, 1993. (Abstract and compressed postscript).
• P. Clark, B. Cestnik, C. Sammut, and J. Stender. Applications of Machine Learning. In Y. Kodratoff, editor, Machine Learning - EWSL-91, pages 457-462, Berlin, 1991. Springer-Verlag. (Abstract and compressed postscript).
• P. Clark. Machine learning: Techniques and recent developments. In A. R. Mirzai, editor, Artificial Intelligence: Concepts and Applications in Engineering. pages 65-93, Chapman and Hall, London, 1990. (Abstract and compressed postscript).
• P. Brazdil and P. Clark. Learning from Imperfect Data. In P. B. Brazdil and K. Konolige, editors, Machine Learning, Meta-reasoning and Logics, pages 207-232, Boston, 1990. Kluwer. (Abstract and compressed postscript).

Case-Based Reasoning

• P. Clark. A Comparison of Rule and Exemplar-based Learning Systems. In P. B. Brazdil and K. Konolige, editors, Machine Learning, Meta-reasoning and Logics, pages 159-186, Boston, 1990. Kluwer. (Abstract and compressed postscript).
• P. Clark. Exemplar-based reasoning in geological prospect appraisal. TIRM-89-034, Turing Institute, Glasgow, UK, 1989. (Abstract and compressed postscript).
• P. Clark. Representing arguments as background knowledge for constraining generalisation. In D. Sleeman, editor, Proc. Third European Working Session on Learning (EWSL-88), pages 37-44, London, October 1988. Pitman. (Abstract and compressed postscript).
• P. Clark. Representing arguments as background knowledge for the justification of case-based inferences. In E. L. Rissland and J. A. King, editors, Proc. AAAI-88 Workshop on Case-Based Reasoning, pages 24-29. August 1988. (Abstract and compressed postscript).
• P. Clark. PROTOS: A Rational Reconstruction, Turing Institute Tech Report, 1987. (Abstract and compressed postscript).

References:

• Professor Bruce Porter (porter@cs.utexas.edu).
  Dept Computer Science, Univ Texas at Austin, Austin, TX 78712. Tel: (512) 471 9565. Fax: (512) 471 8885.
• Dr. Martin Brooks (brooks@iit.nrc.ca) (Lab Head).
  Interactive Information Laboratory, Institute for Information Technology, National Research Council, M-50 Montreal Rd, Ottawa, Ontario, K1A 0R6, Canada. Tel: (613) 990 7661. Fax: (613) 952 7151.
• Professor Rob Holte (holte@site.uottawa.ca).
  School of Information Technology and Engineering, McDonald Hall, University of Ottawa, Ottawa, Ontario, K1N 6N5, Canada. Tel: (613) 562 5800 x 6678. Fax: (613) 562 5187.