W3C

LLD XG F2F meeting - Day two

24 Oct 2010

Agenda

See also: IRC log

See also pictures: http://www.flickr.com/photos/paulwalk/sets/72157625231631710/

See also: Summary of F2F outcomes

Attendees

Present
Tom, Ray, Mark, Antoine, Karen, Jeff, Emmanuelle, Alexander, Kai, Gordon, Lars, Martin, Michael, Jon, Marcia, Paul
Regrets
Chair
Emmanuelle, Tom, Antoine
Scribe
Mark van Assem, Michael, Jeff

Contents


<paulwalk> Locah Project blog post: http://blogs.ukoln.ac.uk/locah/

Topic discussion

topic list for today

<kcoyle> http://www.w3.org/2005/Incubator/lld/wiki/Topics

<rayd> topic list was created as a placeholder originally

now is good time to figure it out

some relate to use cases, some are short and we haven't figured them out etc

discussion should be deliverable oriented, focused discussion

classify discussion into three areas

1. topics covered by use cases, which we should examine further, or for which a use case should be found first

2. topics to be treated as requirements

3. deliverables. things we can achieve

new topics might be created in the course of this discussion

for example, recommended software

first topic, knowledge representation

all about which vocabularies we are using.

michaelp: doesn't fit with a particular use case.

<emma> ...they are all about how we represent our domain knowledge

frsad, for example, has simple model how do we represent that, event or concept

a knowledge representation question

emanuelle: how to model domain

emanuelle: do we want to do in group, or is it for future.

marcia: more than one way to do it. decision to be made by people who assign subject terms

michaelp One of the main ideas of semantic Web: use a URI for real stuff.

<marcia> FRSAD

<marcia> FRSAD is a conceptual model. SKOS can be used to implment the model. But there are two options: SKOS only (lables are properties of a concept) or SKOS + extension for labels

antoine: hard to go into this detail for every model.

Gordon: Generally, we should be recommending VESes as ranges.

gordon: general good practice for linked data. range should be a URI.

emma: need best practice for modelling. is it possible to do this in our timeframe.

michaelp: it is a requirement rather than best practice

Marcia: Differentiate label - FRSAD - SKOS-XL. SKOS without XL works for some vocabularies. We should say: "Here are the two approaches".

Marcia: present different recipes for people to decide.

<marcia> SKOS eXtension for Labels (SKOS-XL) http://www.w3.org/TR/skos-reference/#xl

non bibliographic data

there is one circulation and an identifier use case

emma: rec. dev. is outside our scope. there are plenty of statistical ontologies
... If a vocabulary is missing, we can point it out.

gordon: appl profile for collection description.

<markva> anybody interested in statistics models should look at http://publishing-statistical-data.googlecode.com/svn/trunk/specs/src/main/html/index.html

<TomB> Gordon: There is a Dublin Core application profile for Collection Description - http://dublincore.org/groups/collections/collection-application-profile/

gordon: there are models in other domains, we don't have to invent everything

next: citations use cases

next: application profiles

tom: requirement to clarify what an app profile is and that there are different approaches, point to one or two, some issues

karen: a small number of methodologies
... Libraries should try to converge on some common application profiles.

antoine: wonder whether previous item, frs, should be with app profiles.

tom: should there be something on isbd?

gordon: yes

karen: isnt isbd itself an app model

gordon: no it's a data model

<marcia> ISBD is a data model

gordon: its flat, premarc, no concept of authority data

<TomB> Suggest that we mention role of application profiles not only in ISBD but in RDA.

marcia: applic. profile is more like what steps you need to follow
... Question is if APs are sets of documentations, or APs are technical specifications to be implmented.

tom: role of this group not to say it's one or the other (other being syntax) but point out areas like rda etc

jon: "style" of appl. profile?.

next: legacy data, first subtopic inventory available linked data

gordon: maintenance issue, anything we identify will be out of date soon

tom: do it on fringes but not a core activity

next: vocabularies statuses

gordon: moving targets

karen: difficult for us to know what's being developed and we need better communication channel.

next: Translation of data in MARC format to linked data

mike: translation of data or translation of marc?

latter

"should marc have an rdf representation"

gordon: at least half dozen efforts, experimental, group should take note of that

next: Populating reference data models when legacy data is not perfectly fitting

<TomB> My understanding of this discussion: In Gordon's update of status of new RDF vocabularies (FRBR, etc) - comment on desirability (or not) of expressing the MARC model in RDF

<TomB> ...in addition to the issue of converting MARC records into RDF (not necessarily using an RDF representation of MARC)

<antoine> Scribe: Mark van Assem

GordonD: frbr is 4 records instead of one
... application profile bridges gap

<TomB> Gordon: Coming around to thinking: MARC to RDF triples, then build it into an ISBD record, or whatever. The promise of linked data, focus shifts from record to statement. Application profile fills the gap. Break down, then build back up.

next: [LLD. COMMON-MODEL]

is same as previous

next :[LLD. AUTHORITIES] is in use cases

TomB: problem with wording of the Topic, entities = vocabulaires?

kcoyle: SKOS is handy to put authorities into

alex: already have FRAD

GordonD: authority data is about labels, not entities themselves

Jeff: but SKOS (XL) does both

kcoyle: were two separate databases; in this new world how we model that

emma: req or not?

kcoyle: comes up in use cases

GordonD: The issue here is bibliographic entities versus real-world entities.

michaelp: this is what KR topic is about we discussed in begin
... LD challenges our notion that biblio entities are completely cut off from real-world entities.

michaelp: litmus test for FRs
... "crossing the streams" - challenges us to think of authority files in a different way.

GordonD: is there 1-1 relationship between entities and bib entity within semweb?

emma: put it in deliverable

Karen: Used to be a database in the back room.

<marma> Data is here: http://libris.kb.se/data/auth/220040?format=text%2Frdf%2Bn3

<TomB> Jeff: In VIAF, - Martin suggested using FOAF.

Martin suggest to use foaf:focus to link the foaf:person to skos:concept

Antoine: keep the two topics separate
... be aware that authority data diff of real world
... then how to articulate link
... separate issue and practical solutions, patterns, and cases that use them

Antoine: observable in VIAF, they produce skos:Concept and foaf:Person from same piece of data

next: [LLD. SPECIFIC-VOCABS]

next: [LLD. SKOS-FOR-INTEGRATED-KOS]

what's integrated?

marcia: making connections between vocabs; alignments
... generate superstructure behind linked vocabularies, like UMLS and HILT
... use case bernard and gordon from user point of view find things without being aware of underlying diff vocabs
... available tool or facility; still gap if it's not linked data

kcoyle: include in use case exploration:

GordonD: existing research into relations that appeared.
... a while ago, a group came up with 15 relations for vocabulary alignment. SKOS looked at this and rejected the complexity. But there is a requirement for something more complex than SKOS, particularly in the subject area, in order to provide terminology services to end users.
... simplistic relations are not adequate for use case in cluster of vocab merging

marcia: could be use case based on that
... Terminology mapping backend - available terminology registries - already there, but no link yet with Linked data environment.

emma: put in use case

michaelp: terminology registry use case or a mapping use case?

GordonD: use case that refers to existing research papers; This problem has been well-researched, HILT reports on interoperability - show the complexity.
... we can re-use the SKOS use case for HILT and cite them here - also important in establishing the relationship between SKOS and ...

ACTION: GordonD and antoine to study use cases that relate to vocabulary merging use case [recorded in http://www.w3.org/2005/Incubator/lld/minutes/2010/10/24-lld-minutes.html#action01]

next: [LLD. SKOS-MULTILINGUAL] is a use case

next: [LLD. SKOS-LIB-KOS] in deliverables

michaelp: is it about what's been done or what difficulties are

GordonD: it's what me and antoine just agreed to look at

next: [LLD. PERSON-NAMES]

rayd: covered in my use case

emma: put in deliverables together with authority data

antoine: and refer from there to use cases
... use cases can be moved into requirements if turns out it was not done
... merge person-names and person-metadata

next: LLD. IDENTIFIERS] is use case

next: [LLD. LEGACY-IDS] is requirement

kcoyle: issue e.g. ISBN for manifestation; need to give advice
... think about what ID means

TomB: LCSH cite as example

LarsG: related to digital preservation
... can of worms; need reqs or recommendations

GordonD: need to expose it as can of worms

<edsu> mmmm, worms

next: [LLD. NAMESPACES]

into requirements

TomB: ld principle that URI resolve to representation

antoine: could we refer to webarch?

TomB: part of TBL's four points

LarsG: not particular topic for lld

emma: do we need to address namespace policy?

TomB: yes, libraries should have persistence policies, and principles for vocab evolution
... can URI be repurposed? can meaning evolve?

kcoyle: issue what do you do with multiple copies? how do you identify them?
... important part of structure people need to understand; lot in here that people need to understand so that they do proper LD

antoine: nothing library-specific about it

kcoyle: libraries bring up interesting cases
... library experience should inform web experience

marcia: other communities gathering resources have no clear roadmap

??:

antoine: Europeana experience: for the moment URIs for digital objects are handled quite badly, after a while URIs are dead
... library practice in web context is poor
... we cannot improve that

emma: should say that practice should be better

LarsG: put persistent identification and resolution services into requirements

<antoine> scribe: Michael

emma: Next section: Semantic web environmental issues
... Group with requirements
... Next: Linking across datasets
... What links to what? Group with inventory of datasets in deliverables

Jeff: People could still use OWl to show what is being linked without relying on an inventory
... Self-description using OWL without defining new level of properties

emma: Next: Alignment of vocabs

antoine: Related to previous discussions about skos mapping properties

GordonD: Also about mapping models that are independent of SKOS
... There are different mapping approaches

emma: Next: Alignment of real-world-resource identifiers
... Environmental issue

Antoine: Put into cases for future action
... Bernard might investigate

RayD: about alignment or assignment?

antoine: Relating library authority file concepts to identifiers for the real thing

kcoyle: What is meant with alignment?
... Bringing together if there is more than one?

TomB: And specifying realtionship between them

emma: Next one: The Linked Data paradigm and the Metadata Record paradigm
... Models for packaging Linked Data in records, e.g., Named Graphs
... and Provenance of Linked Data

Jeff: Mikael's email indicates a lot of tension between metadata models and domain models.
... Lot of confusion between these paradigms

ACTION: Tom to re-categorize AGRIS under Bibliographic Data. [recorded in http://www.w3.org/2005/Incubator/lld/minutes/2010/10/24-lld-minutes.html#action02]

Jeff: How can we help people to think in these paradigms?

kcoyle: Educational vs. proof of concept. These are two different goals.
... Can we create the data we want to create without using the records paradigm.

John: I don't think we can create data in absence of a record model.
... Creation, dissemination, and consumption.
... Latter two can happen without record. First one cannot.

antoine: Some of the choices about the right URI in LD look like record building.
... Even on the basic level about which triples you send out.

<antoine> packaging in linked data dissemination context -> http://www.w3.org/Submission/CBD/

marcia: Do you mean the presence of an application profile at the time of creation

Diane: Can we use aggregated view instead of record view?

GordonD: Catalogers create a package of descriptions.

Diane: We need to carefully examine those assumptions.
... Catalogers don't start from nothing and arrive at something that they consider complete.

GordonD: Rarely info in record is created from scratch
... Reliance on external authority and other sources

Marcia: From the abstract model a record is an aggregate of other description sets.

?: But, if things are added, can this info be consumed back into your aggregated set?

GordonD: Triples will be out there. Aggregation will happen on the fly.
... Moving to a "post-coordinated" approach.

emma: We have to cut here
... It is in the requirements.

emma: provenance

kcoyle: It is an requirement. Not specific to LLD.

antoine: We can put it in the use case and probably look at the work of the provenance task group.

emma: We can extract some requirements if we put it in the use case category.

Kai: Strongly related to the record / description set issue.

emma: Next: REST patterns for Linked Data

Jeff: Based on OWL, there is a one-to-many relationship between things
... This should be visible in the URI patterns.
... So they become hackable and provide for content-negotion
... All the way back to real-world objects

<TomB> Jeff: URIs can be made hackable. I like using generic resources. As you hack back, go from representation to generic resource. Hack back on the slash, go back to... etc etc

emma: Is related to best practice for indentifiers.

<TomB> Michael: This can be visualized in one slide.

antoine: This belongs to [SW-Identifier]

TomB: Best practice or research?

antoine: Best practice, comparable to best practice document for eGovernment

Jeff: Sometimes wrong URI patterns limit your choices, eg, for mobile presentation
... We shouldn't have to write one-off systems.
... We shpould create a framework that generates those pattern automatically.

Alexander: Agreed. If we can agree on ID patterns, that would be helpful to others in the community

TomB: I am uneasy about this group saying "we think this is the answer"

edsu: Cool URIs for the semantic web defines the URI patterns just fine - no need to do it again.

Michael: But that is very low-level. Does not deal with relationship between organized knowledge and the world (in detail).

Alex: Do we also consider patterns for modeling data?

Emma: No, it is in application profiles.

Alexander: Patterns not the same as application profiles.
... Software engineering perspective: things that tell me on a basic level.

emma: Created the topic "patterns" and added it to requirements.

ACTION: Jeff to review the UK eGovernment document on identifiers. [recorded in http://www.w3.org/2005/Incubator/lld/minutes/2010/10/24-lld-minutes.html#action03]

<edsu> seems like the url for that spec from the UK about URL patterns has moved or been removed, it used to be at http://www.cabinetoffice.gov.uk/cio/chief_technology_officer/public_sector_ia.aspx

<edsu> definitely ironic :-)

<edsu> ahh, here's the new url for it http://www.cabinetoffice.gov.uk/media/301253/puiblic_sector_uri.pdf

antoine: Jeff to elaborate on the document

emma: Next: Conversion issues, e.g., URIs, content negotiation, RDF compatibility

Kcoyle: Don't know what it means. Very broad.

Antoine: Could we trash it?

emma: OK
... Next: Check if SKOS extensions are needed for describing particular types of KOS (term list, name authority file (not limited to agents and works), digital gazetteer, list of subject headings, taxonomy, thesauri, classification, etc.) and provide SKOSified KOS examples.
... Related to Gordon's and Antoine's UC?

Antone: This is more related to KOS alignment.

Emma: Maybe we have a gap here in the use cases.
... We need a use case about the appropriateness of SKOS to cover controlled vocabularies in LLD.

Antoine: Some is covered in postponed SKOS issues.

Emma: Should check there.

Antoine: We should put it in the vocabulary section.

Emma: Next: extraction of semantic data

kcoyle: Perhaps Marcia can explain what is meant here

marcia: The original email was about a framework of showing things.

kcoyle: Let's put it in the deliverables so we remember to look at it when we prepare deliverable.s

emma: next: linked data management, hosting, and preservation
... vocabulary-specific aspects of management, hosting, and presentation

kcoyle: related to discussion about metadata registries. We need use case.

emma: Put it in use case.
... Next: Versioning, updates

kcoyle: Next three go together. We need a use case for all of them.
... Dissemination mechanisms: RDF schemas, RDFa, bulk download, feeds, SPARQL...
... DCMI-RDA task group would be a great use case.

GordonD: Many issues have surfaced there.

emma: Issues of Web architecture, e.g., persistent URI design best practices, HTTP

Alexander: I don't see pattern as architecture patterns, more like modeling recommendations.
... We should tell peoples about our experiences with our modeling.

emma: Should it be a use case?

kcoyle: We can require things that we don't know how to do.
... It could be a requirement.

emma: Related to "data caching"?

Alexander: Broader context. Ingestions, dissemination.
... I want to have a library system that is able to deal with linked data together with classical library data

Mark: Does that exist?

Alexander: No.

Mark: We have that covered in software recommendation.

Alexander: Not so much the issue what to use, but how to use the tools.
... The IT departments have systems that are going to stay there for a long time.
... We have to come up with ways with doing new stuff with existing infrastructure.
... We are generating LD at runtime. This is not the right approach.

<michaelp> Jeff: I care about that also

Martin: Me also.

michaelp: Perhaps we can do something together

<scribe> ACTION: Alex, Jeff, Martin, MichaelP elaborabe on general purpose IT archtiecture for dealing with linked data with caching feature [recorded in http://www.w3.org/2005/Incubator/lld/minutes/2010/10/24-lld-minutes.html#action04]

<antoine> Scribe: Jeff

<antoine>

Ontology discovery and dissemination [DATA. ONTOLOGY-DISCOVERY]

kcoyle: covered in registry part, discovery part, vocabulary, need a way to find ontologies

marcia: difference between vocabularies/ontologies. format-related

kcoyle: different perspectives on vocabularies: things divided into class, instance, properties, (ontologies?) vs. different vocabularies naming concepts
... no vocabulary of vocabularies
... need to be clear about what we mean when we use the term "vocabulary"

# Search Engine Optimization for Library Data Google Rich Snippets, Yahoo SearchMonkey, Facbook's OpenGraph Protocol [edsu, jphipps] [DATA. SEARCH-OPTIMISATION]

alexander: seems to be close to architecture topic

antoine: these systems may be able to understand library models in the future

emma: this needs a use case

antoine: Europeana wants to put RDFa in HTML

<edsu> facebook's rdfa has a notion of book, author, movie http://opengraphprotocol.org/

<edsu> also isbn :-)

TomB: is there a role of application profiles in search (e.g. Google)

antoine: if you search Google for a book, you will get a Google Book results near the top. It has special status.

ACTION: Emma and Antoine to create use case DATA.SEARCH-OPTIMIZATION [recorded in http://www.w3.org/2005/Incubator/lld/minutes/2010/10/24-lld-minutes.html#action06]

# Licenses, IP, DRM, other availability/rights/access restriction info [antoine, kcoyle, emmanuelle, aseiler] [MGT. LICENSES]

michael: related to provenance and rights discovery

antoine: need common way (RDF) to discover these things

kcoyle: need a use case for provenance and rights

# Workflows or roadmaps for different kinds Linked Data projects [keckert, emmanuelle] [MGT. WORKFLOWS]

# Examples of business models of managing linked library resources (metadata, vocabulary, and KOS resources) [digikim] [MGT. BIZ-MODELS]

# Common patterns in Linked Data, with examples, and with best practices for "Linked Data friendly" output from traditional library data - to provide guidance and save time - maybe several best practices when there are several good ways to solve a problem. [MGT. PATTERNS]

kcoyle: 1&3 have been covered? 2 is new?

Alexander: more concerned with common software (architecture) patterns
... it's analogous to Java classes (built in classes)

<edsu> http://patterns.dataincubator.org/book/

<TomB> http://www.ldodds.com/blog/2010/04/linked-data-patterns-a-free-book-for-practitioners/

kcoyle: we need library examples that refer to the "free book"

Alexander: What are the patterns that are pecularly useful in Library Linked Data?

Emma: examples of business models. no use cases.

marcia: Somebody needs to manage

Karen: sustainability is essential

kcoyle: ROI isn't necessarily money. it can also be cost savings

antoine: abstract a business model from existing use cases?

marcia: somebody needs to envision patterns of business models

# Need for training and documentation (a Linked Data primer for libraries ?) [gneher, Jschneid4, keckert, digikim, antoine, emmanuelle, aseiler] [MGT. TRAINING]

emma: a UTube video?
... can we deliver training and documentation?

antoine: our report should be readable as a primer

<Marcia> +1 Antoine primer idea

kcoyle: the community needs to commit to education in this area

TomB: do we need to specify the skillset?

kcoyle: a lot of people as that question, but few answers

emma: use cases address this in problems and limitations

# Mapping Linked Data terminology to library terminology and concepts [kcoyle] [MGT. LEGACY-MAPPING]

antoine: can glossary make these connections? http://www.w3.org/2001/sw/wiki/Library_terminology_informally_explained

TomB: part of training and documentation

antoine: can this be a deliverable

emma: just listing the terms is a hard task

# Liaison with standardisation bodies and initiatives (ISO and national bodies, IFLA, International Council on Archives, CIDOC...) [GordonD, emmanuelle] [MGT. STANDARDS-PARTICIPATION]

kcoyle: it's a big one

TomB: Gordon and IFLA are a good example

gordon: need on going organizational commitments

TomB: we need to have ongoing communication

# Outreach to other communities (archives, museums, publishers, the Web) [Jschneid4, GordonD, antoine] [MGT. OUTREACH]

emma: we do have a use case related to archives

kcoyle: these communities also have "bodies" that can become involved

antoine: identify a list of these communities and keep it up to date

emma: use the people in this group to create connections to there groups

ray: "collaboration" is different from "liaison". Liaison is too hard.

kcoyle: but necessary.

TomB: try to disseminate our results as broadly as possible.

ACTION: on everyone to update the Events page (http://www.w3.org/2001/sw/wiki/LLD/Events) on the wiki regularly [recorded in http://www.w3.org/2005/Incubator/lld/minutes/2010/10/24-lld-minutes.html#action05]

# How to announce new efforts, build appropriate communities around those efforts, get the right players to the table. [kcoyle] [MGT. NEW-EFFORTS]

emma: it's very general

kcoyle: in the future, make sure we outreach to right people

<paulwalk> Re lldvis: The vocabs are mapped to use cases, but the topics have not yet been mapped at all yet - will do this following today's meeting

emma: group with next steps , new efforts, and future working groups.

<TomB> don't we have a page for linking articles, such as my TWR blog post http://metadaten-twr.org/2010/06/23/new-w3c-incubator-group-on-library-linked-data/?

# pulling in linked data for end users [USE.END_USERS]

# Computational use of library linked data [USE.COMPU]

# Linked data to enhance professional processes or workflows, for librarians, cataloguers, etc. [USE.PRO]

emma: special effort in use cases to demonstrate these points
... use cases to enhance current practices

antoine: can we make this a deliverable?

emma: need a specific section in the deliverable
... that's the end of the list

See post meeting cleaning:Outcome of the topics discussion

time for a group photo

Summary of Action Items

[NEW] ACTION: GordonD and antoine to study use cases that relate to vocabulary merging use case [recorded in http://www.w3.org/2005/Incubator/lld/minutes/2010/10/24-lld-minutes.html#action01]
[NEW] ACTION: Tom to re-categorize AGRIS under Bibliographic Data. [recorded in http://www.w3.org/2005/Incubator/lld/minutes/2010/10/24-lld-minutes.html#action02]
[NEW] ACTION: Jeff to review the UK eGovernment document on identifiers. [recorded in http://www.w3.org/2005/Incubator/lld/minutes/2010/10/24-lld-minutes.html#action03]
[NEW] ACTION: Alex, Jeff, Martin, MichaelP elaborabe on general purpose IT archtiecture for dealing with linked data with caching feature [recorded in http://www.w3.org/2005/Incubator/lld/minutes/2010/10/24-lld-minutes.html#action04]
[NEW] ACTION: on everyone to update the Events page (http://www.w3.org/2001/sw/wiki/LLD/Events) on the wiki regularly [recorded in http://www.w3.org/2005/Incubator/lld/minutes/2010/10/24-lld-minutes.html#action05]


[NEW] ACTION: Emma and Antoine to create use case DATA.SEARCH-OPTIMIZATION [recorded in http://www.w3.org/2005/Incubator/lld/minutes/2010/10/24-lld-minutes.html#action06]
 
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.135 (CVS log)
$Date: 2010/11/02 12:00:26 $