IW Meeting 2012-10-04

From Inference Web

Jump to: navigation, search


Meeting Information



  • Tim
  • Cynthia
  • Jim
  • James (regrets)
  • Deborah probable regrets due to transit to big data meeting - i would like to see you all use the time to make progress on pml 3.
  • Patrice
  • Paulo

Meeting Preparation

Around the room

 * Add a section for yourself 2 hours before meeting.  * Mark any discussion point that you would like to raise during meeting (with DURING MEETING).   * Otherwise, assume that others will read the rest before meeting.   * Also, please be considerate and read others' discussion points before the meeting starts.


  • Replaced PML2 with prov-o in csv2rdf4lod's pvload.sh, which loads RDF URLs into named graphs of a Virtuoso triple store.
    • Paulo: Is this intended to be plain >PROV-O< or PML3? -- Why not PML3? Because we do no have one? 
    • Tim: b/c I didn't need anything beyond PROV-O for my purposes. We could press on it to squeeze some class extensions.
    • Tim will start a wiki page for how we're using PROV-O at each step.
  • I'm on the hook for a PML 3 schedule - but how to?
    • suggest: choose a couple of terms each week, define and exemplar before meeting, get review during meeting?
  • sees Cynthia and Paulo's responses to http://inference-web.org/wiki/Pml2-concept-discussion
    • will review these, thanks!



  • Working on twc-healthdata, not much provenance specific work this week.
  • Finishing up provenance and data cube lit review


  • Investigating older identity relations: QUA-relation from KL-ONE
  • Will be working on pubmed dataset for aggregation, provenance, identity soon.
  • Help James out with FuSE.


  • Visualization service for FUSE.
  • Feature data analysis.


  • Fuse ontology
  • Fuse explanation data


Tim: I think that we should be developing PML 3 a an accumulation of best practices, not an abstract ontology that doens't have actual uses.

Paulo: hasLanguage, hasMimeType, hasFormat. They need to be discussed together. Paulo suggests clusters of terms, we need to discuss together.

Paulo: wants PML 3 examples. Tim: can we have PML 3 problems to solve?

Paulo will be selling PMLP and PROV-O to some colleagues soon, needs PR materials.

Information extraction problem "has it all" - it exercises all of the features of provenance. Ramaz example, original text and annotated with entity recognition, coreference, etc.

Mississippi - single document

  • take original word document
  • we want to learn a conclusion


Consider the followingg text: http://inference-web.org/proofs/MississippiAutomatedSystem/IBMSampleDocument.html

From this text, a machine can use NLP (e.g., information extraction) and learn that Major Julian Allen manages a project called Mississipi Automated System.

Jeopardy: Who is the manager of the Mississipi Automated System? Watson: Who is Major Julian Allen?

But now we can ask questions such as why is Major Julian Allen such manager?

NLP and provenance, Let Me Google That For You:

Patrice =


request that you spend time on PML 3 in today's meeting.

i will report on big data conference and research data alliance meeting provenance discussions next time.

Outstanding Items

Facts about IW Meeting 2012-10-04RDF feed
Date4 October 2012  +
Personal tools