IW Meeting 2011-10-06

From Inference Web

Jump to: navigation, search

Contents

Meeting Information

Agenda

Attendees

  • Tim (skypeless)
  • Jim Mc
  • Deborah (via car phone)
  • Jim H (via Deborah via car phone)
  • Jim Mic (?)

Discussions

named graphs for accounts

http://www.w3.org/2011/prov/wiki/Using_named_graphs_to_model_Accounts

cache graphs: http://www.w3.org/2011/prov/wiki/Using_named_graphs_to_model_Accounts#Named_meta_graphs_of_cache_graphs

https://github.com/timrdf/csv2rdf4lod-automation/raw/master/doc/ontology-diagrams/sparql-service-description-2010-10-31.pdf

http://www.w3.org/2001/sw/BestPractices/OEP/SimplePartWhole/index.html

http://www.w3.org/2011/prov/wiki/PROV_OWL_ontology_components introduces/summarizes:

http://dvcs.w3.org/hg/prov/file/tip/ontology


FUSE debrief

identify emerging up and coming research topics. patents, newspaper articles (twitter feeds perhaps).

A lot of initial work has already been done.

Three points of this program (why it got funded):

  • Evidence, representation and quality.

Patent literature is 

Multilanguage is another aspect, but not our strength.

Chinese and German to start (others to be added Russion, Spanish, Japenese)

Infrastructure looks good.

"Evidence representation" model - 5 slides and discussed feedback.

  • very early proposal.
  • 6 questions (set in stone)
    • 1) is CoP around research area? conference exists, multiple pubs, sustained pubs over time period.
      • gov identified SMEs

5 use cases

  • Human gene transfer
    • have experts, asked 6 questions. when is 1 CoP start, 2 was debate?, 3 commercialization of tpic., 4 industrialization, 5 is there infrastructure?

6 questions at most, program will scope down to same questions for each team

4 teams

  • BAE team (+= RPI)
  • Columbia team
  • SRI team
  • BBN team

PI/PM distinction (visionary vs. programatics)

HW: evidence representation not mature. 4 teams and gov team asked to go through of 1 of 2

Use case 1: Human gene transfer.

Use case 2: Genetic Algorithms (not as developed of a use case)

each team finding indicators for emergence.

  • seeing "concept", seeing approach to 
  • identifying faetures of GAs.
  • finding indicators of emergence

how to represent and explain that we belief some topic is emerging b/c indicators were fused in some fashion.

BAE - working on finding 1) features of topics and 2) indicators of emergence.

then how would we use it to present an explanation.

explain, provide follow up

allow access to SourceUsage and get back to snippets of information that supports the claims. (this will probably happen in phase 2, not phase I)

infrastructurally and organizationally looks good.

How much explanation will happen in Phase I?

  • criteria for passing phase I must exceed 60% comprehensibility
  • quantifative measure for comprehensibility
  • e.g. ask follow up questions - accuracy.
  • handout examples - for how many years did the CoP exist?
    • all were numeric answers

when is next deliverable? features and indicators early next week. telecon next week to brainstorm. draft by Oct 31st for representation and explanation.

FUSENet wiki has 5 use cases publshed. One of 5 is genalgorithms.

Corpus is RODG - related document groups.

FUSENet wiki has description of use case.

RODG is not available yet.

Multiple ROGDs, gov with identify which to use.

When will one of these be available?

3 method for getting input:

  • sciento___ometrics

Queries against Elsevier -provided corpus

HGT - human gene transfer (worth looking at b/c it was the last one they did and was the most developed)

16 use cases

  • 8 advanced
  • 5 posted
  • 8 held back - some used for the test.

HGT use cases shows result

  • used 3-4 SMEs
  • interviewing experts interviewed SMEs.
  • what "anchor" - went to conference in 1993 in XXX
  • 90 min interview
  • highly vetted SMEs
  • looked for what evidence indicated that it is an emerging

"diffusion" going into other areas people coming in and leaving startups, etc.

guess: 1998-2000

3 RDGs - for GA, was it emerging and why do we think so and how sure are we?

  • get asked 6 questions, CoP, infrastructure, commercialization, industrialization.

most useful on FUSENet wiki: GAs and HGT.

  • slides posted today.
  • under forms tab.
  • look at T&E slides @ slide 61-66
  • representation that they are suggesting.
  • need to score >60% for comprehensibility.
  • idea: applying templates centered around questions.

Reemphasized: must be answering quetsions from FULL TEXT - not just metadata. o/w program will NOT continue.

  • metadata derived from full text.
  • will need to focus on extraction, understandability, fusing.

gov has 70 indicators of emergence. will make available.

each of 4 teams will have own indicators.

  • coreference resolution?

cooperative or competative? But each team needs something distinct.

  • we need to give the kinds of questions we are good at answering.

explanations of _debates_? - what makes more sense?

e.g. "here are the 5-50 examples" vs. "here is the summary"

e.g. explanation: "authorittive sources based on rules you've previoiusly agreed with"

Personal tools
Navigation