IW Meeting 2011-05-19

From Inference Web

Jump to: navigation, search


Meeting Information


  • Deborah
  • Paulo
  • Stephan
  • Katie Dunn (special guest)
  • Tim
  • Cynthia
  • Allen Renear (invited speaker)


  • Paulo to invite http://people.lis.illinois.edu/~renear/renearcv.html to introduce FRBR on 26th May telecon.
  • Paulo's 30 June visit to RPI agenda
  • Paulo an invited expert in prov-wg
  • July 6 and 7th. Boston f2f prov-wg agenda
    • Jim Myers has conflict on 6th but may make the 7th; he is interested in the Modeling task force.
  • Tim nominated to prov-wg and on listserv
  • UTEP to move to escience svn and verify that all of their stuff works.
  • Yangfan's request for examples
  • Any comments on source vs. information?

Ongoing Items

  • Create Recipes/FAQ?
  • Tim to incorporate Deborah's feedback for layer page
  • Tim's current modeling for HTTP POST - how can he get a review? (e.g., [1] didn't work March 3rd)
  • IW Browser can't handle Tim's LOGD example
    • Cynthia to document preferred URI design
    • Tim to document his current design
    • Tim to review Cynthia's documentation and discuss disparities.
  • Jitin's Lehigh bench mark for PML.
  • Tim "we are happy to have new members" verbiage for mailing list.
  • 14 Apr Paulo gives written description of a process for data calibration, or stat analysis. [2]
  • FRBR and provenance of written papers.
  • Publishing
  • uncertainty with provenance
  • Paulo give links to youtube videos after he adds audio to it.
  • CSIRO getting IW browser code
  • PML API code examples on escience svn.
  • Tim to review Nick's PML Network API javadoc.
  • Tim to update pvload to reflect new justification-chaining design.

prov-wg telecon notes

Q: clarification of PAQ TF's objectives at F2F

 - a proposal of how provenance fits in the Web architecture
 - a number of issues raised against that draft related
 - aim to resolve some issues at the F2F1

Q: registration fee for F2F1? http://www.w3.org/2011/prov/wiki/CharterConceptsIllustration needs to be reviewed for how it associates to http://www.w3.org/2011/prov/wiki/ProvenanceExample


Allen H. Renear, Associate Dean for Research

Graduate School of Library and Information Science

University of Illinois at Urbana-Champaign



done: ask Allen to post slides to inference-web.org. he said we could. http://inference-web.org/w/images/8/8e/FRBRforInferenceWeb2011.ppt

Allen is one of the developers of FRBR.

Paulo: w3c has officially started prov-wg, RPI/UTEP is a large participant and is working to develop the concepts for provenance.

Stephan is dealing with very practical aspects of scientific activities.

Objective: We see connections to FRBR and would like to discuss connections to PML and the ongoing issues we have with PML.

12:40 - Allen, go!

Group 1 entities.

Allen's background:

a UIUC 10 years. faculty.

interested in ontologies and semantics for XML.

found FRBR in 2001/2002 as an outsider.

cultural and scientific objects

Allen is CS and Philosophy and had been developing similar notions independently.

was starting to teach courses in similar topics and began to adapt it for own purposes.

used it as a device for thinking about these models and representations.

LIS objectives of designing bibliographic ssytems.

Allens environment - frbr is not a device for intellecutaly satisfying dsicussions, but a designing for library information systems.

LIS: =? Library Information Science

slide 2

if you read one thing, read the second bullet:

•What is FRBR? A Conceptual Model for the Bibliographic Universe. Tillett, B. B

Washington: Library of Congress, Cataloging Distribution Service, 2004.

 Highly recommended short pamphlet by the FRBR lead.

very focused on people with LIS concerns, NOT computer scientists or professional data modelers. written by catalog librarians/ reference librarians.

slide 3

management of library services

generalized view independent of any particular information system.

represents current best practices.

"furber" is popular pronunciation :-(

Deborah: pretty well regarded by library information sytems and staying up to date? Yup. Some up and down over decade b/c challenging learning curve. Is highly regarded and steadily increasing.

difficulties in making distinctions. concerns of ROI.

slide 4

helicopter overview :-)

today's talk focusing on Group 1,

Group 2 and 3 are less clear and detract (make one suspicious)

done Q(Tim): why so contentious?

slide 5

slide 6

  • Expressions can be thought of as "text" (simplifying)
  • Manifestation's synonym is an "edition"
  • Item's synonym is a "copy"

Items are concrete

"interlocking cascade" - great way of describing it (OWL axioms do NOT state this... (someValuesFrom))

done Q(Tim): orchestra playing music example.

Deborah: digital objects. identical copies of data or ebook/pdf.

Allen: in digital worlds, notion of item is [] one.

users don't interact with items directly (bits on disk?)

FRBR struggles with digital world.

Identical copy of email, but Allen is concerned that it is NOT an identical copy, different Items.

work has a subject, but it does NOT have a language.

expression has a language, but it does NOT have a subject.

attributes on any one of the 4 layers do NOT also apply to the other 3 layers.

This book in my hand is about Godel, you have to walk up from Item through manifestation through Expression to the Work that HAS the subject.

Paulo: how seriously do they take the disjointness. Allen takes VERY seriously. It is most useful to take FRBR literally and the picture it creates. Take at face value b/c it is clear (but perhaps wrong). If you don't take the separation seriously, FRBR becomes less interesting as an ontology.

"super duper disjoint"

the ink on paper is not in a subjective mood, it is the higher layer that has the subjective mood.

physical objects are not both composed of carbon AND in the subjective mood.

Paulo: wants to open pandora's box.

Q(Tim): many attributes span the layers (in the ontology).

NO inheritance across named arcs. (OWL is cool with this.)

ER modellers have problems with this notion.

oversimplification with notion of Expression. same expression with same RDF, N3 - all at expression level BUT are more fine-grained concern. FRBR outstands expressions as alphanumeric after another.

brick and morter - item vs edition (items are trackable and interact with them)

we don't interact with items in the digitial world.

we are more likely to treat a manifestation as an item.

Q(Tim) but doesn't it land to bits, the physical?

slide 11

expression is NOT a TYPE.

person and student. people cease to be students but not cease to be people.

expression is a Role that a Symbol structure enters into.

Q(Tim) history of the OWL version of FRBR. How well does it reflect the spec?

slide 7

Paulo: cardinalities?

many to many expression <=> manifestation

  • done to accomidate an anthology

a single edition c

"run of cds" is a manifestation.

a single cd with multiple composers' music. (anthology example?)

"symbol structure" - expressions. (back to "text" synonym earlier)

an expression is a symbol structure that realize works.

symbol structures that don't realize works are not expressions

(back to student/person role issue comes in)

Allen is seeing fewer entities and more and more roles.

orchestra playing music - and what are the challenges in digital age.

a little nervous.

music catalogers took up FRBR

page 17 of 1998 j.s. bach's

composer' for organ is expression

arrangement for quartet is expression.

performances recorded are an expression.

recording release on sound disk is manifestation

http://archive.ifla.org/VII/s13/frbr/frbr.pdf 3.2.1

^^ Thanks!

slide 10

Data in RDF vs data in Relational.

in RDF, RDF/XML vs turtle.

character mappings.

still not worked out how to apply FRBR to digital world.

what to consider a work or expression for scientific dataset?

what makes a dataset representative of work or expression?

Allen can send links - Dubin did work in this.

John and Joe used the same data - John's was tab delim, Jills was comma delimited and we dont' say "OH NO, they didnt' use the same data!"

different expressions are understood to carry the same data.

two expressions; tab-delim vs. XML - CORRECT (allen)

Paulo: Geoscience. seismic data, magnetic data, remote sensing.

they keep mentioning, whatever the outcome is, they should be the same structure b/c it is the same Earth. Work is the thing.

Observational science - trying to make observation of same fature of interest, but get different observation based on different observational techniques.

We each do separate observations - two different works.

but they have the same feature.

observations try to approxmiate the feature.

Trying to estimate an aspect of the feature.

paiting a photorealistic of one person

so does another.

the person is the phenomenon we're tryign to capture.

Work's subject is an Observations' feature of interest.

observed property is light you see.

same feature of interest and same observed property.

Work is interpretation of the feature, not the feature itself.

Two interpretations of what should be the same thing.

Each painter is tryign to converge to same result.

groups two and three.

not as powerful b/c they didn't receive attention.

groups two and three are stub-ish.

Allen is comfortable with how two and three relates to group one entities.

Allen: one entities are sufficiently coherent and valueable part of discussion of ontologie for sematnic objects like datasets.

Allen: abstract entities instead of sets of objects.

Allen: observation ontologies are a can of worms.

Looking at RDf triples.

file: I'm looking at in a formal langauge; it is making assertions as propositional content. that propositional content that is carried by those assertions it th ework that corresponds to the expression (the triples themselves).

notation in file (tirples

expression is making assertions

set of assertions articulated by RDF triples is the work.

Work serves as "same dataset" when two scientists are affirming they use the same data. they may not have used same expressions, but used different expressions wiht same propositional content.

Work corresponds nicely with propositional content of expressions in the file.

No inheritence anywhere? it's a slogan. ER digrams there is no inheritence implied by a named arc between two classes. nothing about FRBR that implies inheritence. nothing in spec that implies inheritence. FRBR attributes are designed to be entity specific with NO indication of inheritence. The inheritence that is mentioned by FRBR expositors.

"an expression is about whales because it's work is about whales" - NO.

attributes of higher level entities do NOT inherent attributes on lower level entities.

works do not have typeface or language.

Allen uses term "propogation" as generic term that includes [?]

part_of and gathered_into

(item and the collection it is in)

Allen hasn't considered value restrictions.

Paulo: classes (i.e. sets) vs. instances?

top three levels are abstract

Item is concrete

individuals of abstact objects.

a maniffestation as a set of indivudal copies - Allen: this is an editing error.

identify set of indivudal objects that exhaust the extension of the manifestation, but the fact is not identical with a manifestation.

Paulo: when you see an item, do you know the work for it?

Allen: picking up an item give visual cues, add background knowledge to determine what manifestation, which expression, which work is carried by the item.

daily ordinary epistemic problem of having an item and wondering what it is.

Paulo concerned about cardinality up tree and "knowing" the higher level abstract association.

Allen: separate into two parts:

In an anthology, just pick out the one expression that the antology elemennt embodies, then you're on your way to the one work.

in rework: manifestation is associated with infinite [manifestation?].

no rung is harder than any other. physical object, perceptual experiences, drawing conclusions about what is in hand. when you read a story, you have intellectual encounter with the story. mysterious but it happens; we get to that "work".

Paulo: An item has provenance as an attribute. Do we not have provenance for the other three?

  • if used on his own, space-time movement to where it is now.
  • if from FRBR spec,

provenance of abst

e.g., who is an author of a given item?

pick up book, title page (know b/c bibliographic convention), visual experiences, have belief (no fraud), conclude book is written by David. If not annotated, then search by word usage.

done at item level? Items don't ahve authors. Items ahve manufactuers. Creator of item is a craftsperson that produces books one by one. If Item had text written by person, author of text is the dct:creator of the text.

Paulo: provenance of abstract works.

Allen's provenance is archival interpretation.

Kathyrn: these are different senses.

reformat or incorporate.

combingin - you get whole new work.

provenance of a stain on a book (item)

Allen: provenance in art world. idealized: spatio-temporal location. a career of an object.

sci data provenance - we don't have an enduring object, even if you look at OPM (which focuses on real world). Still, at any time we are not looking at physical object, we are asking about the same question about it's career in space/time, but WHY is a particular physical object/context signifying what it signifies at this time? what is the chain of events in my computer system that is telling me that it is 10 degrees celcious. I'm not intersted in molecules on disk, I'm concerned bout partiuclar representatin tht came to be instantiated in those molecules. In a sense, we still want to know something about physical state (matter and energy). We're not concerned about ehspatial temproal, but it's REPRESENTATIONAL CONTENT.

Paulo: Does the definition of provenance concern space/time, but also include other things at other times. Does the same strategy to handle space/time be used for another (abstract) dimension?

Allen: PML and OPM and others, there are conceptual problems. Traditional archival provenance is just entirely about spatial temporal career of a physical object, but that is not the concern in [scientific/data] domain.

Allen reads FRBR as it is an ontology. you can represent what is being said in FOL.

Owl encoding? There are a few. Some that are wrong.

follow ups?

where does it fit? limitations of inheritence vs. propogation.

lack of minCardinality.

from skype

stephan and i were on but we seemed to be the only 2.

[2:23:58 PM] deborah_mcguinness: we just signed off but i could get back on for about 4 minutes

[2:24:14 PM] deborah_mcguinness: i would like to see the attempted alignment of PML information with FRBR

[2:25:25 PM] Stephan Zednik: sounds to me like PMLP:Source would be a Manifestation, based on Paulo's prior arguments for a differentiation between Source and Information

[2:25:58 PM] Stephan Zednik: * at least it sounds to me that Paulo had previously argued for a pmlp:Source frbr:Manifestation alignment

[2:26:14 PM] Stephan Zednik: anyway, I have to take lunch now

[2:26:22 PM] Stephan Zednik: Will follow-up in email

[2:26:51 PM] deborah_mcguinness: good thx. i also have to get on a different telecon in the next 2 minutes

Tim L


Paulo Pinheiro

Stephan Zednik

30th of June, we are meeting.

TODO: Tim to send Allen to join inference web mailing list.

TODO: tell Allen about hackathon.


Personal tools