Progress Report

Internet Reasoning Service:
Progress Report
Wenjin Lu and Enrico Motta
Knowledge Media Institute
Monica Crubézy
Stanford Medical Informatics
IRS: What it is?
• Web-based tool to support reuse of reasoning
services
• Different levels of support
– Manual browsing/configuration
– Intelligent Assistant
• In the long term: broker-mediated service
Generic Classification Task
• Input roles
– Candidate Solutions, Match Criterion, Solution Criterion,
Observables
• Precondition
– Both observables and candidate solutions have to be
provided
• Goal
– To find a solution from the candidate solutions which is
admissible with respect to the given observables, solution
criterion and match criterion
3. Internet Reasoning Service
Task Selection
 IRS provides
 a graphical and browsable description of each generic
task
 examples of pre-existing instantiated task models.
 Can we do more?
Task Configuration (application inputs)
• Application inputs = case-independent ones
• Instantiate by
– Mapping to domain model
• Solution Space -> Hierarchy of apple types
– Directly filling task roles
• Defining a new match criterion encoding constraint
according to the relevant task ontology
– Selecting from available options
• choosing existing match criterion
Task Configuration (Case inputs)
• No need to fill case inputs at this stage
• Still, mappings may be required
– Observables features -> apple properties
Task Model Verification
 Task Model Verification = Checking task
assumptions (only if they do not rely on casespecific inputs).
 Can task assumptions rely on case-specific inputs?
PSM Selection
• Through a direct link between a PSM and a task.
– e.g., in OCML PSMs are linked to the tasks that they can
solve by a special slot “tackles-task”.
• Through an existing PSM-Task bridge
• As the result of users’ choice among available PSMs.
– IRS will need to support the creation of relevant PSM-Task
Bridge
• As the result of a competence matching process
between the task and available PSMs.
– Competence matching should generate appropriate PSMTask bridge
PSM Configuration
• Same as task configuration
• Roles inherited from relevant task
• PSM may define additional roles
– e.g., heuristic classifier introduces abstractors and
refiners
PSM Verification
• Checking PSM Assumptions
– again, only if no case-specific roles are involved
PSM Execution
 Acquiring case-specific input from user.
 Checking precondition/assumptions
 Calling the PSM code with the mapped inputs.
 Interpreter may be local or remote
 Displaying the progress of the PSM execution, at least in a
console window (that assumes that the code interpreter or
the PSM code sends trace messages to the console).
 Filling-in the domain outputs with the results of PSM
execution
(through
mapping
relations)
and
presenting those results to users.
Possible Platforms for IRS
• Specialized WebOnto Configuration
– Unlikely
– Nobody working on it
• Protégé
– Based on pre-existing PSM Librarian plug-in
– Monica working on it
• New Java/Lisp Tool
– Java Applets interfaced with library sitting on Lisp
server
– Wenjin working on it.
IRS in Protege
Additional Developments
• Classification library to be tried out in 2 domains
– E-commerce
• user classification, product selection
• configuration of ‘user basket’
– will use parametric design library
– Paleontology
• Classification is everything in Paleontology
• Complicated problem
• No agreed hierarchy/classification rules
– gaps in the models