IW Meeting 2010-11-18

From Inference Web

Jump to: navigation, search


Meeting Info

Meeting Info

IW Overall Objectives

     1) PML ontology: (OWL level only) potential non-monotonic  changes (Li) 
     2) PML practice  
         2.1) Use cases and examples (Tim) 
         2.2) RDF level issues (Cynthia) 
         2.3) PML Tools / API issues (Cynthia) 
         2.4) PML primer (Tim) 
     3) PML and OPM mapping (Paulo) 
     4) IW management (TBD)
        4.1) inference web infrastructure and research plan
        4.2) technologies for engagement 
     5) PML Graphical Notation


  • Tim
  • Cynthia
  • Paulo
  • Geoff
  • Li
  • Deborah
  • Stephan
  • Patrick - here
  • James M
  • Jim M


Deborah putting agenda back:


Combing Proofs

Geoff discussing http://www.cs.miami.edu/~geoff/Guest/gcc.html

Combing proof a and proof b Generalized clusterin coefficients. Four columns are four measures. History column shows substitution. four combined proofs

  • Axiom measures - .28 smaller is better than original .71

Last table: starts with a combined proof and continues combining (second round)?

CNF Leaf measure

Is smaller better for all four measures? yes.

Paulo: an example of proof combinations with an example that is not logic?

Science domain example: Represent science example in TPTP (or PML'ed TPTP) - TPTP will handle it irrelevent of content. Process control? set of derivations of same conclusion from same axioms/ancestors. We need to think of an example for two or more paths to the same conclusion.

Paulo: having similar antecedents but different conclusions. Geoff - all they look for is common nodes in derivation structure.

Paulo - tomography of earth domain. They have measurement files. How to arrange sets of datasets to create a model that converts?

Australia trip debrief CIRO

  • They created and presented demo that SPARQL'd PML.
  • Done in TopBraid.

data collection and workflow to estimate rainfall for next (2 hours, 2 weeks?) so they can decide if farmers can draw from river for fields.

Paulo: what are the scientific questions they are trying to answer? List of ontologies they are using? Stephan: they are building a forecasting system. They are trying to answer policy questions.

Turn off your pumps, there's going to be a drought. Farmers pay for water if they turn off pumps.

Modeling change needs?

one major - integration with PML and O&M model - when observing something, it has a result (complex record or a primitive value) observation

point hasRawString to same thing as om's hasResult.

Paulo: conclusion is information (a raw string). but raw string can be structured. put hasLanguage=RDF, then put whatever you want.

Paulo; the Information can be a whole complex beast.

Stephan: need to be able to query both the data and the provenance without "reaching into" the raw string.

Stephan needs PML to integrate with W3c's semantic sensor ontology and O&M domain ontologies (observations and measurements)

(Tim: couldn't you also just describe the instance of Information?)

General issue: "losing the structure" by materializing it into a string.

Li: PML-OWL.owl - http://inference-web.org/2.0/pml-owl.owl InformationRDFInstance subclass of Information Adds a property to point to an OWL instance instead of a materialization of the information. (TAMI wanted to point to information that was RDF)

DIFF ontology - difference between two owl instances

  • a use case for the PML-OWL.owl ontology

CIRO was looking at pml:Query

Deborah: does the pml:Query example give mutliple paths to a single conclusion? Stephan: yes. alternative justification that "bypasses" the database. processingp provenance ends at the query to the service. add the extra provenance to get to the sensor. provided two separate justifications hitting the organizational boundaries stops the provenance, but we want to get to the sensor. e.g. knowing it came from a sensor, but we dont' know what processing was done on it.

two justifications: 1) pulled it out of a service 2) came from a sensor

database problems:

1) structure coming out is not same as going in 2) what comes out could have come from mutliple additions at different times.

two sources at two different times, one query.

Previous action items

1 - cynthia publish 2.01 that has ONLY one change over 2.0 which is the range of hasInferenceEngine is now Agent (instead of inference engine) DONE

2 - we expect that we will still be in the 2.0 namespace

  • OBE Paulo to send Tim examples motivating the promotion of Source Usage "next to" an inference step.
  • Cynthia to incorporate hasAuthor to 2.1 list
  • Cynthia to incorporate change the range of isFromEngine of Query to Agent request to 2.1 list
  • Cynthia to review PML-S ontology to see how it addresses the hasAuthor
  • Cynthia to add isRoot to 2.1 list - describe need/want
  • Paulo to provide pointer to PML-S proposal
  • Paulo to provide InformationContainer proposal
  • DONE Tim to provide SourceUsage promoted proposal (also needs example from Paulo)
  • DONE Tim to find Stephan's renaming proposal
  • Geoff refinement for geoff - geoff will do two annotaitons - one annotation that is generic and another more detailed for one particular example that points to some details
  • DONE paulo will respond to the PML query email from stephan

Action items

  • todo - Geoff will send email by Wednesday on Proof combine work
  • todo - target write paper for cade - due date feb 7 - http://cade23.ii.uni.wroc.pl/ - - plan to start from some material from aaai draft
  • todo - a non-logic (science domain) example for combining proofs.
  • todo - geoff look for examples in process control or management theory from the tptp library. idea is to get at least one more example for the paper
  • todo - Stephan to find the presentation that conveys the demo that put together in Australia (Ching's presentation)
  • todo - Stephan tries the approach that paulo suggested on using pml:Information, sends out email with questions/requests. All, but particularly Paulo help with getting the approach to work
Personal tools