1. Protégé/2000 Advanced Tools for Building Intelligent Systems
Mark A. Musen
Stanford University
Stanford, California USA

2. Generations of Protégé systems at SMI
PROTÉGÉ LISP-Machine system for rapid knowledge acquisition for clinical-trial specifications
PROTÉGÉ-II NeXTSTEP system that allowed independent ontology editing and selection of alternative problem-solving methods
Protégé/Win Finally, a Protégé system for the masses ...
Protégé/Java The subject of this talk ...

3. Protégé/2000
Represents the latest in a series of interactive tools for knowledge-system development
Facilitates construction of knowledge bases in a principled fashion from reusable components
Allows a variety of “plug ins” to facilitate customization in various dimensions
Still needs a better name ...

4. Knowledge-base development with Protégé/2000
Build a domain ontology (a conceptual model of the application area)
Custom-tailor GUI for acquisition of content knowledge
Elicit content knowledge from application specialists
Map domain ontology to appropriate problem solvers for automation of particular tasks

5. Ontologies are cropping up everywhere!
Structured online indexing (as in Yahoo!)
Mediation among heterogeneous databases (as in UMLS)
Reference characterization of application domains (for knowledge reuse)

6. An ontology
Provides a domain of discourse for characterizing some application area
Enumerates concepts, attributes of concepts, and relationships among concepts, thus defining a structure for the application area
Defines constraints on relationships among concepts

8. Conceptual building blocks for intelligent systems
Domain ontologies
Characterization of concepts and relationships in an application area, providing a domain of discourse
Problem-solving methods
Abstract algorithms for achieving solutions to stereotypical tasks (e.g., constraint satisfaction, classification, planning, Bayesian inference)

9. Problem-solving knowledge automates specific tasks
Domain knowledge
+ Problem-solving method
Intelligent behavior

10. Heuristic classification in MYCIN (after Clancey)
Feature
Abstraction
Solution
Refinement
Compromised
host
Gram-negative
infection
Heuristic
Match
Immuno-
suppressed
Pseudo- monas
E. coli
Leukopenia
Alcoholic
WBC < 2.5

11. Problem-solving methods
Provide terms and relationships for talking about relevant problem-solving behavior (a method ontology)
Have a method ontology that defines knowledge roles, which clarify purpose of each element of domain knowledge
Assume that required problem solving can be construed as compositions of well-characterized, generic algorithms

12. Advances in Protégé/2000 Much improved
editing of ontologies
creation and customization of knowledge-acquisition tools
adaptation of system to new requirements
But still no automated support for mapping of knowledge bases to problem-solving methods—yet!
No more shuffling among different development tools!

13. Generation of usable domain-specific KA tools
Protégé/2000 system
takes as input a domain ontology
generates in real time a graphical KA tool
Developers
Tweak KA tool appearance by using direct-manipulation layout-editing facilities
Add custom user-interface widgets when complexity of domain warrants more specialized visual metaphors

15. Another way of browsing the constraints
Another way of browsing the constraints. This view encodes a lot of information.
The top level tab encodes the type of constraint we are talking about.
Left hand side-- across the top are echelons, down the left are unit types.
White indicates that it makes sense to talk about the unit/echelon pair.
White with a number indicates the number of constraints defined for that unit echelon pair.
(yellow indicates slections).
Right hand side: A “table-style” view of the instances. A few slots have their values indicated by columns; each instance defines a row in the table. The user can double click on an instance to view it more fully (and, of course, the C and X buttons work just like they do int eh rest of the application).

16. Protégé/2000 Ontology-editing tab
Add additional
constraints on classes and
attributes
Developer can see knowledge
organization clearly
Easy
to edit attributes and facets
Classification
problems become
viewable

17. A sample application tab: Authoring eligibility criteria
Developed for NCI effort to ease information management for clinical trials
Uses tab “plug in” to support custom-tailored, knowledge-driven application
Demonstrates how our ontology-oriented approach facilitates knowledge reuse and knowledge maintenance

18. The task: Authoring eligibility criteria for clinical-trial protocols
Each clinical trial protocol includes a long list of patient criteria that determine eligibility or exclusion
For given types of tumors, certain eligibility criteria can be inferred from well-characterized clinical states
Each eligibility criterion must be linked to standard patient data that can be used to determine the value of that criterion

19. The program display is divided into two parts
The program display is divided into two parts. The left hand portion of the screen helps to select eligibility criteria while the right hand portion shows criteria that have already been selected.
The initial view of the program shows a diagram of typical Clinical States for breast cancer in the upper left window. Selecting one of the clinical states (in this case State 7) by clicking on it brings up a series of eligibility criterion which are moved to the lists on the right hand side of the screen. Additional criteria that may be associated with a clinical state are shown in the lower left window. These criteria can be moved to the lists of inclusion or exclusion criteria by pressing the appropriate button (inclusion or exclusion) at the lower left hand corner.

22. Linking eligibility criteria to “Common Data Elements”
When a criterion is edited, a new window pops up. The criterion is show at the top of the window and can be edited using standard keyboard commands. At the bottom of the window are common data elements associated with the criterion. To complete the editing process and remove the popup, click the “X” in the upper right hand corner of this window.

27. EligibilityWriter knowledge base
Contains clinical state information and eligibility criteria for
Breast cancer
Prostate cancer
Non-small cell lung cancer
Demonstrates that entry of knowledge for each new type of tumor benefits from previously entered knowledge

28. The race to develop plug-ins
GUI widgets for
tables
diagrams
animation
File I/O plug ins for interoperability with databases, other knowledge-based systems
Tab plug-ins for embedded applications

29. A Protégé-2000 KA tool for entering rules for monitoring nuclear power plants

30. OKBC-compliant knowledge model
Elements of Protégé-2000
User interface
OKBC-compliant knowledge model
Customizable
instance
forms
Slots as
first-class
objects
Facets
standard and
user-defined
Means to
view
large
data sets
Easy
browsing
Instances
Classes and
class
hierarchy
Adaptability
Domain-
specific
tabs
Components
for building
knowledge-based
applications
Custom
widgets

31. Swapping components Each of the Protégé-2000 major components can be
swapped out and replaced with a different one
Knowledge
model
Storage
User
interface

32. Open Knowledge Base Connectivity (OKBC)
Standard mechanism to access knowledge bases stored as “frames” of classes and attributes
Adopted by several well-known knowledge-representation systems (Ontolingua, LOOM)
Will allow Protégé/2000 to be used as an ontology- and knowledge-editing system for any OKBC-compliant server

33. Knowledge-base development with Protégé/2000
Build a domain ontology (a conceptual model of the application area)
Custom-tailor GUI for acquisition of content knowledge
Elicit content knowledge from application specialists
Map domain ontology to appropriate problem solvers for automation of particular tasks

34. Protégé-2000
Designed to use OKBC as a universal front end to a wide variety of distributed knowledge-representation systems
Generates rapidly highly customizable, domain-specific KA and knowledge maintenance tools for use directly by application specialists
Offers access to KA tools and associated application systems as Java applets
Is a set of components that developers can easily embed within other applications

35. Current work on Protégé/2000
Enhanced GUI facilities for knowledge acquisition and knowledge browsing
Access to controlled terminologies via Lexical Technology’s Metaphrase API
Axiom language for defining constraints among instances
Plug ins for OKBC and relational-database connectivity

36. Fourth international Protégé users group meeting
July 29 through August 1
Linköping, Sweden
Hosted by the University of Linköping
See for more details