IW Meeting 2011-02-10

From Inference Web

Jump to: navigation, search

Contents

Meeting info

Attendees

  • Tim, Li
  • Stephan
  • Geoff
  • Cynthia
  • Paulo, Nick, Jitin
  • Deborah, Jim (regrets)

Agenda

Discussion

Proof combining

AB example simple test case combined proofs are easy to understand Geoff: We are not operational; ready to go to realistic example.

http://www.inference-web.org/proofs/tptpwork/tptpextract/test020811/TST/TST001+1/gcc.html shows two tables, one starting from A one starting from B (But they are symmetric)

ready to do multi step combo 3 metrics, small is good. (small := more different than other proofs).

TODO: run single step combo on set theory example THEN: do multi step hill climbing combining with AB THEN multi step on set theory THEN: on suite of examples THEN write paper

Tim/Nick/Jitin scheduling

Tentative: Tue 3pm like last time (became Wed 3pm)

Nick's Probe-It Tool Analysis

Nick broke down into different views.

  • Required: will fail if not there
  • Optional: recognize and display them (make richer vis)
    • has defaults to roll back to.
  • Ignored

Required/Ignored/Optional vs. Error (+critical => fails), warning, informational


Stephan: tool requires a predicate that the model doesn't enforce.

Stephan: validator infrastructure right now. test for pmlpLhasFormat, output informational response. pluggable test infrastructure like JUnit e.g. Spring - config file to add tests. first round will be hard coded. option: caller specify subsets of tests to run (by tool, warning only)

Tim: Implicit action: "we are showing it to the user". e.g. (two properties optional - action is to make inference)

todo: Nick to think about other actions the tool provides.

DONE: Cynthia to expand her list of IW Browser to other columns that Nick shows. (Keep in mind what actions the tool could be providing beyond "showing to user" (which is valid))

We need to consider other ontologies that depend on PML. VisKo is a tool but also an ontology.

Nick may need to have a conversation with Leo since the WDO ontology depends on PML and WDO-It uses PML.

todo: Make sure WDoIt is part of the analysis.

Four Layered PML components

Paulo mentioning 4 components:

  • Use cases need more work
  • Provenance concepts.

VisKO is an OWL ontology and Tool. (VisKO should be considered in Tool Analysis and PRov oncepts.)

VisKO and WDoIt import PML ontology.

todo: Note that any ontology related to PML is part of the provenance concepts component.

WDoIT is a dependency for other ontologies (?) - Paulo

todo: Tools that write PML versus tools that read or visualize PML great distinction ^^

Tool: Data annotator (Antonio Garza) - generates PML from abstract workflows (creates PML)

Tool: Derivator (Antonio Garza) - interactive Drupal/CI-Server app that creates PML when documents are added/updated in Drupal (creates PML)

Tool: CI-Server (Aida, Patricia) - It has some minor PML-specific functionalities such as the capability of exposing Drupal users and pmlp:Person instances. It manages documents in general including documents containing PML instances (creates and reads PML)

Tool: PML-Sparql Front-End (Jitin) - A sparql-based query interface for PML instances (reads PML)

Jitin Arora

Tim: UTEP's endpoint doesn't provide the URL of the query results, as in http://logd.tw.rpi.edu/sparql b/c of post? yes.

Tim: can I query the endpoint programmatically?

Jitin will adapt Lehigh benchmark for PML

Cynthia wants check on loops in the new validator -- added to Stephan's table.

Stephan: Nick's table helped Stephan with the validator. Cynthia doing it for IW Browser. Which are required, which are optional? e.g. no conclusion required, but expected by probe it tool => test.


Paulo regarding walking documents and validating.

  • PML with nodeset and infstep.
  • infstep needs to point to infengine and infrule.
  • validator checks for valid URL/URI, but does NOT check if alive and if it is of the type it should be.
  • UTEP has one PML validator (Jitin's validator) and RPI has another PML validator.
  • Jitin's validator: UTEP PML Validator
  • Li's validator: RPI_PML_Validator


Paulo: should it traverse links? thinks no. how far to traverse?

todo: instance analysis for connectivity.

Nick: ProbeIt only follows to certain point. ProbeIt stops at Person. e.g. user feedback on finding errors in maps. - used that to decide when to stop.

Important "tool" to discuss: Nick's "PML-lightweight API" which is in fact a PML-Network API

Use cases

Use cases - weak?

Paulo: use cases in papers, but lost into papers. we have them, but not properly/systematically documented.

Paulo is UTEP POC for use cases.

Next steps? add pointers to use cases at http://inference-web.org/wiki/PML_2.1_-_a_Layered_Approach#PML_Use_Case_analysis

use cases of provenance group - language independent.

http://www.w3.org/2005/Incubator/prov/wiki/Use_Cases#Original_Use_Cases_Proposed

two use cases with same concerns - way to minimize that is to use list of requirements.

http://www.w3.org/2005/Incubator/prov/wiki/Requirements


3 complex "scenarios" were created to cover the variety of use cases.

wondering if we can link the tool analysis to some requirement identified by the provenance group?

Stephan is focused on validator and is looking for use cases from UTEP and Cynthia to make test cases.

Jitin's ontology supports searching with sparql, e.g. getting all antecedents.

OPM Flavors

Stephan: information model vs. data models - they def have multiple data models, probably only one information model. OPMO and OPMV are data models (ontologies) OPMV is vocab without roles (a lite version) OPMO lets you attach roles.

OPM 1.0 to OPM 1.1 - the information model changed

Object model (Java API) is a data model that produces one of the data models.

+resource-oriented model.

DONE: Stephan tweak OPM Flavors page.

Personal tools
Navigation