Meeting Notes November 11

From Semantic Portal Wiki

Contents 1 Action Items 1.1 Updates 1.2 Todos 2 Notes 3 Presentation 3.1 Post-presentation Questions

Action Items

Updates

• Peter : There will be a group meeting on Nov 25 (Thanksgiving week). Please come if you are going to be on campus, but it is not required if you have travel plans during this time.

Todos

Patrick requests for volunteers :

• script (PHP / PERL) for populating individuals for VICODEMA Knowledge Base - D2RQ server - Josh
• install 'drupal instance' with sparql end point - on tier 2 machines - Jin and Evan
• C++ JAVA integration - Shankar
• classify 460 'parameter's in CEDAR ontology (build hierarchy on parameter classes - temperature, velocity) - Eric

Li for Alvaro's work :

• Excel to CSV for crime data - Dominic

Notes

Presentation

• Presenter : James Michaelis
• Topics
  • Workflow Provenance mining
  • provenance representation strategies

• Provenance Mining
  • represent data derivation provenance from logs using SW technology, in a way understandable to the scientists.
  • capture processes, agencies, operations and data (input/ output)
  • IMP - we'll be mining provenance - done post execution - i.e. no control over log creation.
  • Challenges
    • too much data
    • RDF structure not intuitive first up.
  • How
    • Map provenance to domain independent model - PML - for consistency
    • abstract representations for conciseness
  • Workflow Mapping Example
    • GILA flight scheduling, so as to remove any conflicts
    • system generates OWL based logs
    • Step 1 - find obscure relationships between nodes (?)
    • Step 2 - concisely - logs contain links from Problem nodes to Solution nodes. In PML we have direct mapping of these nodes to Nodesets, and they are linked via inference step nodeset.
  • Provenance Abstraction
    • still working on automating the strategies for generating abstractions.
  • Provenance Representations
    • How OPM relates to PML
      • Process - InferenceStep
      • Used - hasAntecedentList
      • WasGeneratedBy - isConsequentOf

...

Post-presentation Questions

1. Peter: What is the goal for abstraction?
A. provide a concise view, for better browsing - present provenance relevant to user
2. Peter: How do you determine what is relevant
A. decided from the artifacts whose provenance user is interested in (?) But we need to know
3. Ankesh : Provenance mining relevant in context of TAMI. Question, why would it be more intuitive to have standard but non-domain specific basis for representation (ex. PML)?
A. Showed one of the slides that the PML nodes are linked back to the original nodes in log. so user can refer to that for better understanding
4. Peter : In GILA we want to check if someone ran de-confliction(?) algorithm to verify a solution. If that step is missing we want to know that. Has James looked into that?
A. Not encountered such problems yet.
5. Josh : What is the level of granularity you have in mind?
A. same as that in the log (?), at-least bound by it.
6. Greg : May be a linguistic issue but how could an OPM:Actor map to PML:InferenceRule?
A. that's the most sensible mapping but its contentious issue, being discussed with Prof McGuinness and other researchers.
7. Peter : Many mappings are missing (between OPM and PML), and existing ones do not make sense linguistically. How do you think of overcoming those?
A.
8. Peter : Do you know if anyone has tried mining provenance from syslog - UNIX syslog.
A. Not to my knowledge.