IW Meeting 2011-02-01

From Inference Web

Jump to: navigation, search

Contents

Meeting info

Attendees

  • Tim
  • Nick
  • Jitin

Discussion

This is a continuation of yesterday's notes.


http://inference-web.org/wiki/PML_2.1_-_a_Layered_Approach


http://inference-web.org/wiki/September_2010_plan_for_PML_2.1

(bottom) First level - not a lot of constraints Second level - some constraints Later levels - more constraints

(Jitin notes that this corresponds with the semweb layer cake - RDF has no constraints, then RDFS, then OWL)

(Have they decided to have a WG?)

PNNL - other depts: Seismic models in Geology


Tim: Trying to establish scope for what need/should be done. Then sketch in who is covering the different aspects. Then identify the missing gaps. (What does a Layered PML design look like?)

P -> P_cp + P_r + P_r1 + P_r2 (P J -> j1 + j2 T

Why?

  • understanding the tools (what they need)
  • easier third-party adoption of PML
  • extensibility - aligning with other ontologies
  • incremental development (adding things to PML)


Multiple purposes:

  • Documentation (examples for people to know what to)
  • Test cases

Version-control the test cases and make it available for those trying to understand the tool.

Probe-It loads bookmarks - things they know it loads.

Definitely can assist in compiling a list of assumptions that probeit/visko make about the PML they are visualizing

What Tim is doing:

anyhoo, the LOGD takes csv, converts to rdf using parameterized structural interpretations. the process is PML captured using some tools I ginned up. [1/31/11 4:03:08 PM] Tim L: prov-xg mappings - I want to tweak the table they have to make it parsable and cast it to SKOS encoded in RDF.

UTEP SPARQL Queries: http://trust.utep.edu:8080/sparql-pml/

crawls CI Server: http://rio.cs.utep.edu/ciserver/CI-Projects


Where else was Jitin crawling PML instances?

  • Leo's SAWs
  • RPI

e.g. hasFormat on Information (not)

  • essential for visualizing
  • but people focused on just querying.

Tuesday 3pm

  • ProbeIt! discussion - what PML does it use
  • VisKO a service that provides suggestions for how to visualize
  • Might be nice if validator could tell you what actions are supported by the pml you are validating*


Li's "quad" a tools requires a particular instance construct to do a perticular action to support a particular user need. <T, I, A, N> (frequency of occurrence \alpha importance of elements of quads)


instance data weighting - find out what is important. (frequency of occurrence \alpha importance)

TODO: Next meeting Jitin will present fairly detailed discussion of what crawling he has done.


Nick and Jitin are on tools and instance data side. Less so on use cases and fundamental provenance concept (first -principles)

Nick: Just got them PML, they aren't sure yet what they can do with it; ideas for waht they will be able to do shoiuld be coming soon.


SPIN - http://www.spinrdf.org/sp.html

One benefit for codifying the "what tools use" instead of "just documenting it". being able to recognize that some instance data is missing a construct that a tool needs to perform an action that a user would need to have.

TODO: Nick to pick a few simple instance constructs (e.g., on Info, need EITHER a hasURL or hasRawString (xor)) start with narrative, attempt at quad, so we can

Jitin: when looking at frequency of different constructs, do we separate (modules based on function) willing to split? e.g. P vs. J - one metric P,J X Projects (more J in TPTP and 50/50 in LOGD, some might have JUST P)

Nick: a layer should be able to tell a developer what actions and use case supports.

e.g. : Stephan would have two layers (one with antecedents as a list, one with it just as triples)


LAYER \cross {RDF, RDFS, OWL1 , OWL 2} lower layers should have more of the lower sem cake layer constructs. (is something is OWL 2, then it is likely to be in a higher PML layer)

Personal tools
Navigation