SWBPD F2F -- 1 Nov 2004

Agenda Adjustments

<danbri> Re Restaurant, I think it is here, but jjc will confirm: http://www.conradatjamesons.co.uk/location/index_location.htm

<libby> yep that's the one

propose to swap RDF-in-HTML and ADTF slots, so Bristol folk can move to the teleconf room

RDF-in-HTML will be discussed at 1530 UTC, ADTF will be discussed at 1330 UTC,

tomorrow afternoon discussion on breadth of work and ways to bring in other participants

PORT TF

<RalphS> [ PORT (aka Thesaurus) TF description is http://www.w3.org/2004/03/thes-tf/mission ]

al: brief intro to skos core spec
... SKOS Core is an RDF vocabulary for thesaurii
... vocabulary is managed like FOAF; each property in the vocabulary has its own level of stability
... classes and properties summarised in a table, with their properties
... and a brief mention of the management policy for skos core
... and a brief note re community involvement
... list of vocab details
... that bit of document generated from the schema itself, hence table created from script
... navigation at side of document

ralph: stepping back briefly...
... feeling a bit out of sync with this task force
... how does this vocab serve as being the most important answer right now to get thesauri ported to the semweb
... what's the feedback you've got that makes you confident in this as a solution

al: coming from the other point of view, within dig library community, for thesauri alone there is currently no dominant interchange format
... so v little reuse of tools etc within that community

jjc: how much is skos yet another format that won't be a standard

al: will answer that one later
... skos is part of a family of similar systems, thesauri, subject heading schemes, glossaries, which have all been developed with walls in between them
... they all share a lot of common features
... another thing in the thesaurus community is that moving beyond print environment is still a new thing

guus: do you have indication how many of the vocabs out there have such a structure that can be usefully mapped to skos
... eg mesh not something that you can...

al: i'd say 98%

DanBri: work that is written-up here [SKOS] has a heritage back to other work, including a write-up I did for the W3C Query Workshop in 1998
... vocabulary as written is good
... some commentors suggest it could be more term-centric
... this is low-hanging fruit
... could relate to future Topic Map discussions

al: i've presented this at a few digital library workshops
... the single feature that i've picked up on that people have liked...
... use of rdf means people can add new features, and extensible, you can specialise classes and properties
... v attractive as people want to refine thesauri standards to be more precise

phil: i'm kinda concerned that this might be a closed community

ralph: why are Wordnet and MESH hard examples for SKOS?

guus: wordnet isn't a conventional thesaurus in the ISO sense
... nor mesh
... but in the library world, ppl have been making thesauri for ages

ralph: can we target a specific thesaurus?

al: sure. we already have gemmet(?) who use a prev version of skos
... also uk govt categories
... in swad-e we did a cycle of review/test/etc

ralph: i didn't realise it had been used, wasn't clear from the docs

al: is linked from the swad-e pages

<libby> http://www.w3.org/2001/sw/Europe/reports/thes/

<libby> usecases: http://www.w3.org/2001/sw/Europe/reports/thes/usecase.html

DanBri: the Digital Library community likes this because RDF allows them to use many representation types and specialize
... worries me because of a view toward "facet modelling"
... I think they don't realize what can be done with OWL
... in the RDF & OWL world we can say "subClassOf"
... in the thesaurus world they are accustomed to want to say 'broader' and 'narrower'

Al:SKOS attempts to not replicate things that are already in OWL

DanBri: 'denotes' relationship bridges models
... http://www.w3.org/2001/sw/Europe/reports/thes/8.8/
... subtle distinction between modelling in RDF and OWL over modelling thesaurii -- will be important for us to explain this

al: of whole skos core vocab, the broader property is the most important one in there

david: al, you said the thesaurus community only recently thinking beyond print... can you expand?
... what are the changes?

al: single biggest impact of that, working in the print community, you have a thesaurus which is a book, terms, record cards...
... you fill terms from thesaurus into record card
... so term-centric
... as people working with this
... a problem with that, 'cos in a thesaurus you have two kinds of term, preferred descriptors and non-preferred terms
... a descriptor is a main label for a concept
... ...so confounding of term as unique id vs usage in natural language

david: i'm with you ... but ... so what?

al: althought a thesaurus is supposed to be concept oriented, in practice it is term oriented

<Benjamin> WN draft might be : http://www.w3.org/2001/sw/BestPractices/WNET/wordnet-sw-20040713.html

guus: all these communities have built vocabs, they're v domain specific, terms represent agreements within a community
... if people commit to this particular terminology

al: as we move into networked world

Guus: even within a community there will be conflicts and disagreements in usage

al: where two vocabs collide in networked world, this is a new problem/issue to pre-electronic version

<Guus> sure!

david: can you make that explicit in the document

ACTION: Alistair make explicit in skos core doc the fact that you're trying to deal with potential for multiple thesauri using the same terms, overlap etc., different from paper publishing world

Steve: this is an area in which Published Subjects might be helpful

guus: not sure ... that this is specifically work for this tf
... happy to see port tf to propose a skos-based convention for represnting real-life thesauri in rdf/xml
... could be very short
... my main problem is ?size of doc / quickstart / primer

DanBri: note comment from Brad @ Siderean

phil: you're looking at having multiple independent domain specific thesauri, but not at detailed cross domain mappings
... are we looking at the problem of translating across domains; e.g. medical to something else?

Al: that's part of SKOS Mapping

DanBri: faceted browser for Dublin Core was built using a term-centric approach
... the developer said the term-centric approach was easier to deal with, though less elegant

ralph: this discusssion positions what you've done better for me, in space of problem we're trying to solve.
... is particularly important to emphasise the point that guus just made, perhaps in skos core guide, that the motivation be given right up front...
... with examples
... there's a risk that community rivalry might project from perceived qualities of a thesaurus to skos itself
... so show several!
... so skos would be more persausive if described this way
... so important distinction, between basic thesaurus representation versus fancy stuff on mapping
... get basics out asap, showing some use for thesaurus world without getting bogged down in mapping

phil: suggesting that words on these constraints needed in the doc?

ralph: not sure we need to write it that explicitly, just make sure we work through real use case examples
... by implication this shows our priorities

jjc: felt to me that al's later examples could have been more prominent

guus: i was visiting sound and vision institute
... tv and radio archive of the netherlands
... dutch
... simple textual file, i just want to have a document to give them, to translate that
... simple use cases are there
... another one, dutch royal library, GOO (general object something)
... v simple, uses 3 or 4 things from skos
... term, broader term, related, ...
... give them something to allow their developers something to build an rdf representation

ralph: so speed is important?

al: totally agree; with skos-core guide, as short and simple as possible, ...

<danbri> [libraries-> web not sw; blogs, instances]

al: suggestions welcomed

DanBri: several things going on here; lots of historical context for wanting to help migrate things to Semantic Web
... librarians want to see their role in Sem Web
... many cheaper and scruffier examples being generated
... SKOS is interesting because it is right in the middle
... it fits with the light-weight data sharing things like FOAF but it also fits into the library world
... I'd like to see 2 classes of examples; a dump syntax for thesaurii and syntaxes that mix into the rest of the RDF world
... with the move from print to electronic, what's changed in the library world is how they operate
... there now are blog operators that create thousands of categories that point to a page
... hierarchical categories in SKOS have been shown to work nicely
... can now show to the library world that their work on thesaurii can show up in this new blog category world

phil: a point about pragmatics
... a solution space for semantic tech for library world
... as they're potentially an early adopter, could be some pain there
... is it within our remit to help with take up
... ? there's some responsibility within this team to act as a reference point, or respond to, frustrations from early adopters
... is that a wg responsibility?

david: this is our constant problem. charter is v broad. i wouldn't say that it is the wrong thing to do. potentially too hard for the group as it currently stands.
... we could do less of something else we're doing, or else find someone out there to help with this

phil: from my standpoint in industry, when you're trying to become an early adopter of this technology, BP seems to be an ideal home for channeled frustrations

jjc: i disagree

david: a pain registry would be good in theory...
... but because i'm sensitive to what this group is being asked to do

DanBri: the people who participate in WGs are not generic reassignable resources; they have some role in their host organization
... for example, those of us funded by EU are here to create supporting materials
... when this goes out to WD, the WG has the responsiblity to respond to public comment

ralph: SWBPD wg charter is, for better/worse, v broad
... i view the intent of the charter of this wg to help the real world use the sw tech
... migrate to it, etc
... we can't afford to take on every user community whose pain we feel
... but we should be confident that everything we work on addresses some genuine user community problem
... what i've been pushing here is for the materials that we deliver to show the real world problem that is being addressed
... depending on the prob, we may or may not be able to demonstrate we solved it all
... so be clear about what probs we want to be seen as solving
... agree w/ phil that some of that is needed, but also +1 david's point re resource limitations

david: ralph likes to (rightfully I think!) pick up the concept of use cases... If a discussion comes up, Ralph will remind us to define a use case, also in document review. This is a really good idea. When we talk about what use cases we're going to do, the same short list comes up; It's always FOAF and DOAP. We have a short list of a common candidates for use cases

<DanBri>: [DOAP vs DOPE]
drug ontology thing

<libby> DOPE: http://www.informatik.uni-bremen.de/swc/dope.html

david: if we use same list across all our note candidates...
... eg. we could look back and say, 'we did xyz to bring foaf into mainstream fold, ...
... so we should probably come up with list of these use cases

jjc: back to phil's earlier point... i think it is important that we don't do support of individual users
... ralph talked about communities not individuals
... we do latter on jena-dev list

phil: I agree

David: I see indivs supported in IG; communities supported by WG

steve: re topicmaps angle... this skos looks like a v interesting project
... collab, make sure whatever we do in the rdf/tm tf is consistent

al: thanks, agree; Kal Ahmed, ...[missed names] have been active w/ skos
... re Published Subject Indicator, currently discussing whether to have a PSI property in SKOS
... re spec document, people seem ok with that, concerns have been with the guide document
... guide is at an early stage, built from our wiki content
... feasible to publish for end of november

guus: more discussion tommorrow?

al: might be useful to run new features of skos core past the WG
... because of the subject indicator discussion

guus: so i hear two issues; 1st is potential examples, 2nd is to talk about topic maps [& denotes property]
... ie. open issues within task force; alongside discussion of rel'n to topicmaps tf

ralph: david commented about standard use casees
... guus has emphasised that there are use cases; please lets pick some.
... concern i have with foaf is that it is simple and intuitive
... but foaf is already developed, and rdf friendly

guus: foaf is a description template, not a vocab

ralph: not clear useful to list foaf as a skos use case
... anything we do, there's some community out there, who we should know about and whoese needs we should address

al: I could've written this by showing classic print thesauri followed by skos version; could do that, but foaf developer folks wouldn't follow. There are different communities who need different kinds of example.

guus: who is main community? digital library?

al: point is that this is skos core guide... is that the right title for the right document?
... another point is a primer specifically for using skos w/ thesauri

<RalphS> "Expressing Your Thesaurus in the Semantic Web"

guus: how about not having skos in title

Steve: not just thesaurii; also controlled vocabularies, subject headings

(various proposals uncaptured; discussion of whether just thesauri ...)

all: great work alistair!

<DanBri>: RalphS, I believe the intro to http://www.w3.org/2001/sw/BestPractices/WNET/wordnet-sw-20040713.html explains how SKOS, Wordnet, and RDF/OWL fit together. Would be great if you could review that bit of the doc in particular.

Wordnet TF

Guus: history: history wordnet in rdf several times - but none endorsed by princeton
... phase one - wordnet datastructures in RDF/OWL - no changes, make it so that everytime wornet is updated, get an rdf/owl version too
... Brian has been working on a draft: http://www.w3.org/2001/sw/BestPractices/WNET/wordnet-sw-20040713.html
... details to be worked on: wordnet has synsets, a group of terms. many linguistic relationships between the synsets and the terms
... main problem with the phase one conversion is lack of people's time: more work on the document and then liase with princeton
... Christianne Fellbaum someone was definitely interested in working on this
... which maybe enough

<RalphS> [ WordNet Task Force description is http://www.w3.org/2001/sw/BestPractices/WNET/tf ]

guus: issues: what shoudl the base uri be for wordnet; how shoudl updates be handled? and how shoudl words be repreesnted
... this is the only document we have really. focus discussion on how to get more effort into here

danbri: everyone wants to do the exciting stuff not phase one stuff: this is hard wor

guus: would love to do it if he had the time

danbri: good for all to review the introduction to this...easy to confuse the differnt tasks, e.g. could skos grow to address any of the wordnet tasks?
... some of these are research questions e.g. similarity of structures between thesaurus and a lexical database

jeremy: maybe just a few terms to connect those...maybe in a separate namespace

brian: issue in modelling wordnet is : what is it you're modelling - a lexical concept, a sequence of characters?

<RalphS> [Christiane attended our 27-May WG meeting to talk about WordNet-SWBP collaboration; see http://lists.w3.org/Archives/Public/public-swbp-wg/2004May/0149.html ]

brian: people using it in RDF want a large set of classes not "the word bicycle"

danrbi: yes - two traditions

guus: we decided to do the lexical one first and the classes one in phase 2

some text from guus: http://www.cs.vu.nl/~guus/public/wn.txt

jeremy: we could just discuss the resource issue?

david: do it with skos?

jeremy: complexifying stage 1

guus: problem is too many representations and people like a common reference

ralph: has princeton endorsed any rdf reepresentations?

benjamin: decker one linked form their page

david: but none of these are complete representations

<danbri> [they'll list anything that uses wordnet; they don't QA those links]

jeremy: could a grad student take this document through the process?

guus: it's not just a grad student exercise....

ralph: but we could help answer some of teh harder questions
... any candidates? even at princeton maybe?

guus: might have a candidate, btu would need a group to ask questions of

ralph: a student might drive the group by asking specific questions

david: do we gain anything as a wg by using skos for this/investigating this link

guus: doesn;t think there's much of a link betwewen skos and stage 1

david: librarian case - their terms have a lexical route, they might benefit

guus: librarians and wordnet are distinct communities
... wants to get something out fast

aliman: agrees that 1st step, skos not appropriate. pragmatic point - people don;t want two solutions for one problem: cleaner at first anyway to offer one solution

danbri: agrees with that. also peopel confuse teh powers of skos and the powers of rdf, lots of education to do there. cobining the two will confuse matters even more.
... already acknowldged in teh docs that these could become closer later, leave it at that
... wordnet is a model of the english language - work in other places to generalize the model to other languages e.g. euroowordnet, japanese

danbri: be exiting to have this language neurtral stuff later. plenty of simple work for now

ralph: generalization can make it less attractive to a particular community; specialization in documents helps us more in deployment

<danbri> (http://www.w3.org/2001/sw/Europe/talks/200404-nict/Overview-2.html I plugged wordnet (class model) to NICT folks nr Kyoto)

guus: we just say: use this one (with princeton) - and not say why ercept ina background document

danbri: "use this one" worries me - as there are naturally 2
... you (guus) are talking about adding to the lexical version with classes later....

guus: class representation is too difficult, one taht we all agree on...

danbri: make it a brutish one

guus: the revisions are the issue
... agreement in the TF telecon is to work wioth princeton to help with representation at the lexical level plus discuss with them transformations tools that could help with different versiosn
... in the maintenance phase the schema doesn;t change, just the content; for the rdf classes version the schema does change

ralph: do we know the mechanics of how they update the database?

danbri: they do a release every 6 months or so

<aharth> ftp download at ftp://ftp.cogsci.princeton.edu/pub/wordnet/

brian: 4things to do in TF: basic schema; talk to people whove done other represerntations and involve them; needs to interact with princeton, getting them to include it in their distribution etc; finally build some tools to create the RDF representation
... all depends on the first one
... lots of work in this TF above producing the document

guus: first step is critical but not the only thing

brian apologises for not having got further with this. happy for someone to pick it up and finish it; otherwise brian will find the time to get a first cut by the end of the year

danbri: q for brian: do you ahve a sense of how much more work needs ot be done on the document before a pre-working draft release of teh document. impacts on interacting with other producers

brian: betweeb 2 weeks (grad student) and end Dec (brian, given his time contraints)

ralph: would it be harmful to toss a coin to pick an existing representation?
... would like to keep the tf around until we get more resources rather than suspend it

guus started with the swiss one... [scribe missed name]

scribe: we're very close to that one

danbri: my version is not a lexical version, can't use that

<libby>[[
- KID Group, Univ. of Neuchatel, OWL representation
http://taurus.unine.ch/GroupHome/knowler/wordnet.html
]]
http://www.cs.vu.nl/~guus/public/wn.txt

<danbri> [as are half the others; this WD-draft is as good as it gets, I think]

<danbri> http://www-106.ibm.com/developerworks/xml/library/x-think6.html ... article by Uche coding to Melnik's version

<libby> [Univ. of Neuchatel version is the one the TF drafts are close to]

ralph: perhaps there is soemthign in jeremy's suggestion for a grad student to do

guus: steve - is there a link between PSIs and this work?

steve: yep

guus: could use these as publisheed subjects

steve: yep

<RalphS> i.e. if there is content that Guus and Brian already have that is not yet represented in a new version of the Working Draft, it sounds like there's a place for an editorial resource to help

danbri: might be interested in contacting say uche, re contacting developers

guus: Univ. of Neuchatel version is a more complete version of the melnik version

danbri: would like a few words on the spec about lexical vs class representation - could do this text

ralph: design rationale - useful but not essential...

guus: would be nice

ralph: basic design would not wait for the design rationale

brian: danbri's offer to interface with develeopers in this area would be very useful

ACTION: brian and danbri need to talk about what need to do for Wordnet document to be good enough

guus: review the decisions that need to be taken on this document tomoorw? 11.30 -1

[missed brian's comment sorfry]

danbri: worried about not knwoing when we've got it right

guus: this what the working draft will do
... considers all the other parts as very important, incl class representation, how to use the lingistic representtaion for annotation of images, say

<RalphS> [yes, publication of the working draft will get visibility for the design which then gets feedback on the correctness]

--break for lunch

ADTF

guus: people want to see showcase applications that show added value of the technology
... some nontechnical examples to make that point

<RalphS> http://esw.w3.org/mt/esw/archives/cat_applications_and_demos.html

guus: semantic web challenge 2003: AKTive Space, DOPE, Building Finder, Museum

<RalphS> [ADTF description is http://esw.w3.org/topic/SemanticWebBestPracticesTaskForceOnApplicationsAndDemos ]

guus: things in common: integrate different large data sources, RDF/OWL used for syntactical interoperability
... storage and access issues the main things to worry about
... schema mapping required
... use of owl:sameAs was an issue
... information integration and presentation is an issue
... unfortunately only in-house because it's about computer science
...DOPE: very typical based on a thesaurus
... uses EMTREE thesaurus based on mesh
... 5M Medline abstracts, 500k full-text articles
... disambiguation of search terms an issue
... use case: search for information about aspirin
... medicine is important area, professionally used
... won a technology award recently
... BuildingFinder:> USC
... use various sources (satellite images, roadmap info, address information)
... BuildingFinder: reverse address lookup not possible in some eu countries
... image alignment algorithms
... point to a satellite image and find out the name of the person who lives there
... combination of structural and image processing techniques (multimedia info)
... Finnish museums on the web
... BuildingFinder: unfortunately in finnish
... weblog providing us already with some decent material, maybe not sufficient

<RalphS> [Longwell would be another good addition to http://esw.w3.org/mt/esw/archives/cat_applications_and_demos.html ; http://simile.mit.edu/longwell/ ]

<RalphS> -> http://atlas.isi.edu/semantic/servlet/SemanticServlet BuildingFinder

libby: presents weblog
... how to get rdf descriptions out of the weblog
... weblog started out of part of the swad-e project weblog
... plan to use the weblog as part of the skos effort
... application page: title, uri, descriptions about projects
... who to contact, more information, categories
... 17 applications in the weblog already
... bit cumbersome to fill in the template
... because of spans in the template it's possible to use GRDDL and XSLT to extract the information
... uses doap (description of a project) vocab to encode information about the projects
... mixed-and-matched with dc and foaf

<danbri> ...in head of http://esw.w3.org/mt/esw/archives/cat_applications_and_demos.html

libby: uses swed: ns for categories

<danbri> DOAP, see http://usefulinc.com/doap

<danbri> (not to be confused with DOPE?)

libby: swed (semantic web environment and directory): facetted browser

<RalphS> [ SWED is at http://www.swed.org.uk/swed/ ]

libby: contains environmental information
... SWED browser worked quite nicely for the semweb applications and demos
... SWED uses SKOS to describe categories
... categorization by name, and other properties
... build-in harvester can be used to add data from remote sites

<danbri> example of Redland w/ SKOS and DOAP descriptions... [another great swad-europe deliverable :-]

<RalphS> [I wonder if SWED has any provenance yet]

libby: possible to add data and filter on properties
... creation of records about projects could be done by the application authors themselves

<jjc> http://swordfish.rdfweb.org:8080/adtf/ is top level uri for what libby is presenting

libby: then harvested
... doap uses freshmeat categories for software projects

jjc: question about harvesting and control policy

libby: possible to trust a certain domain name

ralphs: maintaining provenance data?

libby: really uses SWED as an application, keeps track of trusted sources and other sources
... jen golbeck's work is related

davidw: difficult for users to accept is presentation of raw uri's
... we are providing human-readable labels
... non-technical users focus on longish uris

ralphs: that's what rdfs:label is for

danbri: truncation mabye?

<danbri> [aside: recent sobering experience re slipping into geek assumptions; the woman sat next to me on plane on friday hadn't heard of iPods/MP3, and I found myself realising I was suprised]

ralphs: in this case truncation doesn't work

Guus: OpenDirectory is the closest existing categorization

davidw: open directory is used for categorization, but is not really appropriate

<RalphS> [yes, left-most truncation on the DOAP category example Libby was projecting contains the only human-interpretable information; that freshmeat and sourceforge were part of the classification]

libby: need labels and uris

davidw: having demos that are compelling to end users tucana is interested it

aliman: got a student who's looking at a wiki tool for building a thesaurus

guus: demos page is a demonstration in itself
... how to operate this?

<Zakim> danbri, you wanted to seek confirmation that we're happy with community project self-description (trust ppl not to be vain selfpromoting dorks)

guus: 18 new applications in the semantic web challenge
... what areas are missing?

danbri: what's the motivation for making vocabularies owl dl friendly?

guus: most applications use rdf and owl:sameAs
... maybe it's too early for owl applications, takes two years for applications to use new stuff

jjc: maybe have options on what semantic web technologies are used in the applications

ralphs: maybe a bit technical
... what's the process of maintaining the software package?

<danbri> [for listing RDF vs OWL DL vs Full etc., I find narrative content ("what we did was...:") much much more valuable than simple checkboxes ("we use OWL DL.").]

guus: later on maybe do a survey, keep the barrier low for data entry
... have to depend on ongoing projects that showcase applications

<RalphS> saying what technologies are used is also liable to be out-of-date; will someone remember to update the "Now uses full OWL" entry when they go beyond the DL subset?

guus: owl and xml datatype discussion

XML Schema Liaison

<RalphS> [XSCH Task Force description is http://lists.w3.org/Archives/Public/public-swbp-wg/2004Apr/0125.html ]

jjc: issues the task force is addressing came up in the webont and RDF Core wg
... two questions about how to use XML Schema datatypes with RDF
... current situation is a compromise between what's ideal and what's possible
... user-defined datatypes
... issue here that there's no agreement what uris to use for user-defined datatypes
... when are two type literals the same
... issue here: are 0 as a float and 0 as an integer the same?
... rdf and owl testcases don't include testcases on this question
... datatypes come from xml schema, not part of the semantic web activity

<RalphS> [JJC discussing http://lists.w3.org/Archives/Public/www-archive/2004Oct/att-0049/xsh-sw-note.html ]

jjc: some response from the xml schema wg
... need the buy-in of the xml community, but need to progress as well
... xquery/xslt are also working with the xml schema datatype
... a lot of w3c groups are potentially involved

<Guus> Related issue: numeric ranges require iser-defined datatypes: http://lists.w3.org/Archives/Public/public-swbp-wg/2004Apr/0066.html

ralphs: producing a sketch that people can comment on maybe a good idea

jjc: possible solutions;
... the DAML+OIL solution works for datatypes that has a name
... using name attribute to fragID the xml schema descriptions
... however, not in conformance with RFC2396 and xml schema mime type
... for sw people, this solution is better
...alternative: xml schema component designators wd
... powerful solution to navigate in the schema using xpointer
... however, quite complex
... possible solution: use both id and name

ralphs: it's important to distinguish between a concept and a particular description of the concept

danbri: made the conflation using uri's for identification and getting the description by dereferencing the uri

jjc: xml refers to syntactic objects, on semantic web the resource that denotes itself is interesting
... uri of a description is the uri of the thing described

ralphs: other ways of constructing an uri: bnode xpointer scheme?

jjc: both the id solution and the xscd solution could work
... next item: comparison of values
... comparison between float's and int's
... simplest possible solution: all the types are different
... but somewhat counterintuitive
... long and int are derived from the same primitive type

guus: number datatype would help

davidw: we have super datatype of number

<RalphS> XML Schema 'decimal' type is http://www.w3.org/TR/xmlschema-2/#decimal

davidw: implemented in product because of customer demand

<RalphS> XML Schema 'float' type is http://www.w3.org/TR/xmlschema-2/#float

jeffp: what are the current answers from the xml datatype spec?

jjc: no consensus within the xml schema group regarding the issues
... the document should be clear about that though. http://lists.w3.org/Archives/Public/www-archive/2002Nov/att-0092/04-num-hierarchy.png

jeffp: is it a subsumption hierarchy or definition hierarchy?

jjc: two extremes: all types are different vs. strong mathematical representation

<danbri> [ context: http://www.w3.org/TR/2004/WD-rdf-sparql-query-20041012/#extendedtests ]

jjc: xpath 2.0 documents mention the eq operator that says "2.0 eq 2" is true
... xslt have typed literal objects and operations can be done on these literals

guus: what next steps need to be taken?

jjc: key question: publish a document without xml community?
... continue to work on the document and then ask for input on an editor's draft
... timeline could be distribution to the other wg's before chrismas

<RalphS> [I was unaware of the XML Schema Component Designators work -- looking at it now, it does appear to be a strong connection with this datatype issue]

danbri: relationship to DAWG/SPARQL and xquery operators?; see http://www.w3.org/TR/2004/WD-rdf-sparql-query-20041012/#extendedtests

<RalphS> [XML Schema: Component Designators

<RalphS> http://www.w3.org/TR/2004/WD-xmlschema-ref-20040716/

<RalphS> W3C Working Draft 16 July 2004

<RalphS> ]

jjc: following the xslt 2.0 is more likely to be in line with the dawg group

<Zakim> RalphS, you wanted to say that mathematicians will laugh if we adopt the viewpoint that xsd:decimal has a disjoint value space from xsd:float

ACTION: jjc review SPARQL WD re http://www.w3.org/TR/2004/WD-rdf-sparql-query-20041012/#extendedtests

HTML TF

[Ben Adida calls in for this session]

-> http://lists.w3.org/Archives/Public/public-swbp-wg/2004Nov/0003.html [HTML] Status for RDF/XHTML [BenA]

Ben: RDF-in-XHTML-TF has been in existence since 1999
... users include FOAF, TrackBack, GEO-URL
... Dublin Core, and more recently Creative Commons
... main issue is how to embed RDF triples in HTML
... have been focussing now on the sections that are dependent on the HTML WG's current timetable
... working in parallel on making sure we understand our requirements

<benadida> http://www.w3.org/2003/03/rdf-in-xml.html

Ben: ... there are items in the 27 May 2003 document that we are no longer sure still make sense
... e.g. direct embedding of RDF/XML syntax

[Tom Baker calls in at this point]

Ralph: the problem is complicated because at least 2 WGs need to cooperate, and possibly 3.
... also, the HTML WG's charter does not allow them to request certain changes of XML schema (for example)
... Jeremy sent proposals to make RDF A simpler
... need to offer proposals to HTML WG before last call

Jeremy: initial review (from 1 week ago) some issues came up
... have done implementation of RDF A
... serious issue: RDF A is too complicated
... some of the rules are too complicated
... the number of different ways of representing a triple in RDF A is 432
... if you run examples from spec you get more triples than the author said

Ralph: the spec needs more work either to represent authors intent, or to fix triples
... opportunity: the words in the spec could be simpler
... and could fix triples at same time

Jeremy: some clear simplifications but doesn't go far enough l ...
... more functionality ...

<Ralph> -> http://lists.w3.org/Archives/Public/public-rdf-in-xhtml-tf/2004Nov/0001.html Jeremy's simplification of RDF/A

Ben:it's good because can represent more complex triples, literals
... and an RDF triple in an HTML clickable way
... without duplicating anything in markup
... which is good property of GRDDL

<Zakim> danbri, you wanted to check familiarity with new W3C Compound Documents WG

Danbri: new WG at W3C Compound Documents

Ben: lot's of important techinical detais re RDF A - please take a look!
... Now get some initial feedback asap for html WG
... look at 3 line statement from Ben's email
... important message to HTML WG: we';re moving in the right direction.

Ralph: This meeting output: statement that Ben put in his mail from today ...

<benadida> statement: We find RDF/A to be a big step forward and encourage the HTML WG to use it in place of the 22 July MetaInformation module. Our forthcoming detailed comments and suggestions on RDF/A are intended to perfect this work in fulfilling the long-standing needs of the RDF deployment community to embed semantic web data within HTML documents.

Ben: Can we all endorse that statement?

Guus: are willing to make this direction statement?

ralph: what questions does the WG want to ask of the TF before deciding on the endorsement?

Phil: direction is superb
... pragmatics - concerned about potential abuse within XHTML user community
... question need to push forward incolusion of metadata for XHTML 2
...rather than hold back and look at richer set of use cases
... concerned about the potential abuse of RDF/A within HTML, perhaps focus on RDF/XML embedding
... use of metadata inside XHTML may be misinterpreted by general public
... if I were writing metadata in web page now, would do it to drive search engines.

Guus: this is important point, but outside scope of this WG

Ben: what do you consider abuse?

ACTION: Phil to write up concerns on email

Danbri: how strongly are we pushing this as as a new RDF syntax?
... originally to handle FOAF namespace ...
... hoped that the FOAF namespace document could be validated for RDF content ...
... so do we support the practise that e..g. the FOAF RDF description be written as embedded in an XHTML document?
... I.e. should we write ontologies in RDF A?

<RalphS> [me says No!]

<danbri> [in which case, 1 original FOAF use case remains unmet]

Ben: goal is to produce alternative serialisation for RDF ...

Ralph: goal?

Ben: requirement to embed arbitrarily complex RDF statements in XHTML
....to satisfy this requirement is a goal.

<RalphS> [but it would be an interesting exercise to see how much of the expression of an ontology in RDF/A could be done]

Guus: meant to e used mainly for annotation purposes ..

Jeremy: RDF A IS a new RDF syntax ...

Ralph: but this in itself is not a goal.
... i.e. implies something wrong with existing syntaxes.
... The result is to create a new syntax
.... but this is soultion to original requirement.
... Use RDF A to express an OWL ontology? No, not a goal or recommendation.

Danbri: what about my original FOAF use case?
... I want a single resource for both humans and machines
... my use case was to write a single document at my namespace that could be both presentation and RDF without content negotiation

<DavidW> RDF/XML has known problems as an XML vocabulary and as a serialization format

Phil: This new variant of RDF will become the defacto standard syntax ...

Guus: can people outside the TF support positive statement for RDFA

danbri: on the fence, wants to talk to compound docs WG

SteveP: opening a can of worms ... lead to trouble
... but in the mandate of the TF support for complete RDF
... think RDF subset for simple annotations sufficient.
... the problem is that the requirement was for complete RDF support, not "just enough" to write metadata _about_ a document

ACTION: Steve to email on concerns for RDF in XHTML

David: should not treat RDF/XML as sacrosanct
... problems with it ... we have an opportunity to recognise that RDF is to concept, NOT the syntax
... this proposal leads to use cases for HTML authors ... and more
... would rather clean this up than see RDF/XML fixed, or RDF/XML in XHTML.

Danbri: In RDF Core, test cases used rigorously
... worried that design work so far happened without test case infrastructure ...

ralph: we have an action to express all RDF test cases ...

Ben: I have an action to make sure this works with creative commons and with FOAF ...

<RalphS> [make sure that DanBri differentiates between users of FOAF and the FOAF namespace document]

David: we have to get formal comments to HTML WG before use case work.

Guus: very positive about this work, with reservations about test cases
... but if going to happen, then happy.

Jeremy: political goal to be positive.

Guus: propose to make general positive statement, with technical caveats

<danbri> me: "It's great and it's useful and it's progress... but it doesn't address my use case (FOAF namespace documentation: RDF/XML inside XHTML); I want to know how that'll be progressed. Shoudl we begin a conversation w/ Compound Documents WG?"

Guus: add wording about test cases

David concurs

Danbri: fulfil SOME of the needs; who should do test cases?

Ralph: ask them to do all of our test cases

<danbri> maybe 'many of the...'

Danbri: they should be using test case driven framework.

Ralph: don't want them to feel that all test cases must be met.

Guus: rephrase the statement?

ACTION: David to reword the statement on RDF A to HTML WG.

Jeremy: they need DIFFERENT test cases
... RDF Core test cases are a starting point ... may not want to use all ...

Danbri: framework for doing it using NTRIPLES.

Ben: HTML WG meets next week ... not all will agree to RDFA
... if impose too constraints now, they may revert to metainformaiton approach
... can we include test cases in detailed comment s to come?

Guus: in fact test cases are implicit in any spec
... so can leave this out of comment

<DavidW> Try this one: We find RDF/A to be a big step forward and encourage the HTML WG to use it in place of the 22 July MetaInformation module. Our forthcoming detailed comments and suggestions on RDF/A are intended to perfect this work in fulfilling long-standing needs of the RDF deployment community to embed semantic web data within HTML documents.

Guus: proposed to send this as comment to HTML WG

second Jeremy

motion carried

ACTION: Ben to send this statement to HTML WG via email

VM TF

Guus: outline current docs and issues please.

<RalphS> [VM Task Force description is http://www.w3.org/2001/sw/BestPractices/VM/ ]

Tom: sent out timetable ...

<RalphS> [Tom's timetable is in http://lists.w3.org/Archives/Public/public-swbp-wg/2004Oct/0147.html ]

Tom: put document on list before holidays
... Move to CVS for finer editing by January
... First public release mid Feb
... Other section try to describe some principles of good practise
... then third section discussing unclear issues
... After discussion ... there are some points could be added
... Do we have large scalse vocab to use as example of good practise ?
... or discuss this in open issues?
... but before getting into issues - is the timetable realistic, does the outline look good?
... aim to quickly flesh out the draft with low hanging fruit
... agree on main points ... then begin refinement.

Guus: timeschedule?
... ambitious but feasible (good to be ambitious)
... also fits well with charter for TF to produce results within 2-4months

David: there has been interest re ontaria
... interest in evaluating ontologies posted on ontaria against VM recommendations
... and posting compliance ont he site
... i.e. this is the kind of use this thing will be put to.

Tom: compliance is a heavy word
... we are trying to get agreement on some basic principles
... to evaluate ontologies against these principles could be good but
... we are talking about quite geenral principles ...

Guus: this note does say anything about 'compliance' but customers may ask about 'compliance' ...

ralph: this WG is producing docs with are 'best practises'
... whatever can be mechanically tested will be.

Danbri: part of this TF leaving machine readable evidence for management of a vocab
... but machines cannot tell if statements are true ot not
... wories me when people look for a big pile of 'good' or 'bad' ontologies ...

Guus: N.B. we are talking about vocabulary management
... proper usage criteria
... we can endorse this without going into a good/bad debate.

David: there is good/bad URIs and issues without going near whether an ontology is itself good for a specific job ...

<Tbaker> agree with David

Jeremy: can these principles say: use RDFS label?
... and re 'conformance' this says rec rather than note ... would rather not go there.

David: we can put out a series of good ideas as a best practise group
... without taling about compliance
... but as Guus says, people may choose to evaluate compliance relative to our note
... even though it is not a W3C recommendation.

David withdraws the word 'compliance' :)

Tom: a suggestion came up to have an example vocab that provides example of good practise points
... examples of different types of vocabulary
... describe FOAF, DC, SKOS, .... illustrate range of vocab tyupes
... pointer to how management is done for these vocabs
... still not clear to do about the really large ontologies
... there are some big vocabs that do not use URIrefs
... what should we use re large complex ontology as example?
... e.g. FAO fisheries, wordent, NCI ..

guus: we take some simple vocab for section 2
... want to keep as simple as poss.

Tom: so OK to feature simpler vocabs in section 2
... leave high end onts for esction three?

Guus: exactly.
... which vocabs on the table as examples?

Tom: FOAF DC SKOS Wordnet
... + maybe major medical/life sciences vocab
... Wordent section 2/3?

Guus: FOAF excellent example

Danbri: good example, shares stuff with DC ...

Guus: also nice to hive more terminology style vocab
... e.g. FAO thesaurus
... I.e. maintained RDF representation by owning authority

Danrbi: two classes of thing to do:
... 1. interview people on how they managed older vocabs
... 2. manage specifically in relation to RDF representations ...

Guus: my preference would be to choose something that is already expressed in RDF by its owning organization

Alistair: opportunity to pick something that already has a history of evolution

David: spoke to NCI guys last week
... have 5-6 guys
... + chief editor who merges by hand changes
... process is painful
... want standard tools to handle change ... i.e. real world problem
... difficult for large onts edited by multiple peoplle
... so people appreciate guidance on how to markup an ontology
... to support change management.

<Zakim> DavidW, you wanted to ask Tom whether the VM TF has addressed guidance for multi-user editing and merging of edits for very large ontologies.

Guus: any more on sample vocabs?

Tom: for people coming new to W3C
... want to collect issues into one place
... this note would be helpful if could summarise in 2-3 paras
... what the major papers are, what there scope is etc.

Guus: agree

ACTION: VM TF to compile list of sample vocabs for the note ...

Guus: Candidates FOAF, DC, and one thesaurus style vocab (missing candidate)
... look into candidate for this ...

Ralph: we need to find someone who'll keep maintaining things in good way ...

Jeremy: also need to choose example with good modelling, even though modelling is not the focus.

Steve: (on published subjects)
... what about vocabs where don't use URIs ...

<Zakim> RalphS, you wanted to say that in the VM case the only reason to push to an external example (e.g. non-FOAF) is to engage some specific community

ralph: whatever we pick for third, need to be reasonably confident about their current practise, or that they will follow our best practise
... OASIS may be possibility ...

<RalphS> [specifically, if we can nudge OASIS in a better direction by involving their Published Subjects in this work, that might improve the world]

Jeremy: published subjects good if meet quality threshold.

Steve: may meet in DC at XML conf 2004 week of nov 14-

Guus: summarising:
... positive feedback on outline and timeschedule
... endorsemenet of using DC and FOAF as examples

ACTION: VM TF with help from Guus to find thesaurus like example and high end onts to section 3

Tom: final point: appreciate help setting up wiki.

danbri volunteers

<RalphS> [I accept the actions listed by my name in http://lists.w3.org/Archives/Public/public-swbp-wg/2004Oct/0150.html ]

OEP Task Force

Guus issue to discuss: how do we get additional people involved in some task forces? e.g. Sandpiper
... according to current W3C policy it is difficult to ask people to become invited experts from companies

Ralph: Reminds all members of W3C economics and the benefits of joining. There are good reasons not to invite experts from non-member companies unless they have skills we specifically need and cannot get in another manner.
... Invited experts may also join when an organization's joining is in progress.

Guus: Good for everyone to know policy.

Deb: This particular request was from a very small company. Joining may be difficult for very small companies.

Ralph has an action to discuss this with the company.

David: Companies are responsible for making the decision to join or justify the reason not to.

Guus: Editors of documents do not have to be members of WGs or W3C.

Ralph: Not sure that is a good precident to set.

David: Since Ralph has an action, we should move onto OEP business.

Deb: There are other companies who are in the same position. Should we generate our Note with them as authors, but not members?
... OWL Time would be another note that would benefit from the involvement of individuals who are from non-Member companies

Jeremy: I have a preference that editors and authors be bound by the patent policy.

<jjc> http://lists.w3.org/Archives/Public/public-rdf-in-xhtml-tf/2004Nov/0004.html was msg from DanC I mentioned half hour ago

Guus: More about the priorities for the OEP work?

Deb: Reviewing http://www.w3.org/2001/sw/BestPractices/OEP/

discussion of breakout groups

breakouts will be:

1. RDF/TM TF description (Steve, Alistair, David, Libby)

2. RDF/XHTML issues + XML Schema datatypes (Benjamin, Felix, Jeff, Ralph, Phil, Jeremy)

3. WordNet phase (Andreas, Guus, Dan, Brian)

Breakout report - RDF/TM

<aliman> http://www.w3.org/2004/11/02-swbptm-irc#T14-24-51

Aliman: happy with TF description

<aliman> http://www.ontopia.net/tmp/RDFTM-TF-DoW.html

Aliman: only those in the meeting today

<danbri> [re TF membership, I have a few things I want to contribute, not sure yet quite how much time I'll have for TF overall...]

mailist identier?

Guus: suggest rdftm

Aliman: we frame the initial statement
... to combine the two families
... to provide transformation between the two objects:
... three longer term objects

Ralph: note additional W3C Process requirements if the RDFTM TF intends to produce a Recommendation (or a "Recommendation-track document")

Steve: that means sth not part of the short term objectives

Alima
n: The initial focus is on features defined in ISO 13250 Topic Maps.
... Other Topic Maps-related standards (such as TMCL and TMQL) may be considered at a later date.

Steve: we discuss on various detailed issues, and suggest consider them later on

Alistair: DELIVERABLES

<RalphS> [discussion of dropping the word 'complete' in approach "2. Choose one or more of these as a starting point for defining a complete methodology." ]

Guus: what do mean by "Note"?

David: we can intent to have a "note"

Guus: WG note on this issue.

Steve: any W3C member can provide note?

Ralph: note true; member submission vs. WG note

Guus: WG note is different from recommendation
... we can use working draft
... our note is not for public review yet

Alistair: DEPENDENCIES are left updated by steve, dependency on PORT TF for SKOS

David: by the next teleconf
... are you addressing any use case in this TF?

Guus: it might get lots of attention from other communities
... suggest two co-ordinaters; one for W3C and the other for ISO

David: not sure if we need coordinators

Steve: we can get some publications out of it

ACTION: David to contact Eric Miller re his interest in joining the RDFTM TF.

Dan: someone might write about it at xml.com

ACTION: Steve to finish rdftm TF description

Ralph: are you soliciting public feedback?

Steve: the note should cover all the existing approaches

Ralph: WG Note implies we don't have further version of it
... otherwise it is a Working Draft

Steve: TF only produce draft?

Ralph: TF provides proposed draft
... WG decide if it can become WG working draft

Steve: what are the final output from WG

Ralph: recommendation or a WG note
... WG provides last call working draft
... provide evidence to director

Ralph: change the second point [change 'Working Draft' to 'Note']

Guus: WG consensus can lead to a WG note
... time?
... rdf/xhtml

Breakout Report - RDF/HTML & XSCH

Ralph has sent notes to the mailing list

Ralph: Phil concerned that RDF/A addresses a closed community; the opportunity
... to express a variety of use cases around RDF/A is limited.
... Phil agrees to provide some use cases
... Jeremy says good tools are part of the solution to manage proper usage.
... Benjamin and Jeff willing to help
... Jeff agrees to give comments of current draft
... Jeremy points out there are two related working drafts had last call last year
... hard to get further comemnts

Guus: can we handle the late suggestion?

Jeremy: we should have done it much earlier

Guus: is it critical?

David: waht about rdf/a

Jeremy: w.r.t. the value space question, the most critical thing is that implementors do the same thing -- which choice is made won't matter as much

David: we should be prepared to comment on rdfa in the next few months

Jeremy: i have made all the negative points in their mailing list

David: Mark has been had his way

Ralph: we have to be persuade

<danbri> today's negative comment on rdf/a: http://lists.w3.org/Archives/Public/public-rdf-in-xhtml-tf/2004Nov/0007.html

Breakout Report - Wordnet

<danbri> see: wordnet breakout: raw notes

Guus: TF draft will be ready after their f2f meeting

Guus: we revise the TF
... we look at the prolog source of WN
... and went through the list of all issues
... requirement that URI should be humna readable
... we should do some test on URIs
... we postpone discussion on some issues
... ask Jan to write transformation into RDF/OWL
... we include wn:lexicaForm
... we also resolve Prinston team re requirements
... resolved not to add a verb group for mow
... we provide two versions of WN
... we can complte by the end of the year
... move to class-centred representation

Wrapup

Guus: 1. UML stuff
... we can also talk about location of next f2f
... teleconf is useful
... ODM group is more ambitious now
... also do metamodel mapping
... they use OWL full
... as the anchor point for translations
... other metamodels: rdf, topic maps
... also mapping to scl but only one way
... they will have a two way UML - OWL Full mapping
... mapping between ER models and OWL Full
... possible for express as well
... they expect our feedback
... we should review their draft in Dec
... we should plan in advance for review

Jeremy: HP agees to review

Dan: comments? 6 pages or 2 lines comments?

Phil: they mainly want to be awared

Guus: which part HP want to review?

Jeremy: not sure

Guus: TM TF should review the connection between OWL full and TM
... if their work is good, we can take it

Steve: we can include it into our previous work

ACTION: Jeremy Clarify which parts of UML docs HP is most interested in reviewing

ACTION: find someone to do the review the part of UML about TM

Guus: two chapters; TM metamodels and its mapping to OWL Full

Guus: Dec will be the review period

Guus: I can do metamodel of OWL full and mapping to UML
... ask PatH to review the scl part
... who are interested in OWL full to ER?

David: Tate Jones can do that

Guus: it makes sense to help them as much as possible

David: will the OWL to ER be chapter review?

Guus: yes

next f2f at W3C Tech Plenary

<libby> 28 February- 4 March 2005, Boston, MA, USA, Hyatt Harborside Hotel

Guus: whole week meeting; WG f2f meeting and TF meetings; up to us to decide

Jeremy: should be working with XML schema working group etc.

Steve: we can have TF meetings first then WG meetings; i.e. 4 days of meetings?

Ralph: we can ask

Guus: ok for everyone?
... Nov 18 next teleconf

<libby> http://www.timeanddate.com/worldclock/fixedtime.html?month=11&day=18&year=2004&hour=15&min=0&sec=0&p1=0

Ralph: 2pm boston time

<libby> http://www.timeanddate.com/worldclock/fixedtime.html?month=11&day=18&year=2004&hour=19&min=0&sec=0&p1=0

<RalphS> 2pm Boston is now 1900 UTC

Guus: Nov 18, 1900 UTC; after that we stick to two weeks schedule

<RalphS> 2 December 1500 UTC

Guus: www panel

<RalphS> 16 December 1900 UTC

<RalphS> (noting that 2 December is during the Advisory Committee meeting)

David: WWW2005 May 10 2005, japan
... we need strong representation from this WG
... who will be in Japan then and want to be involved

Steve: I am tempted but need to talk to the boss
... good chance to present our work

David: panel submission dl: shortly
... Steve, Jeremy, ...

David: someone from NI ...

Guus: review
... table about TF and members
... Felix agrees to help ADTF
... we decided to drop some TFs ...

<RalphS> Guus is showing the table in the 2004-03-04 F2F minutes: http://www.w3.org/2004/03/04-SWBPD

David: one to explain SW to developpers

Phil willing to participate in TF revise

<DavidW> WRLD TF to be considered again in March/April 2005, after some TFs complete (maybe VM, WORDNET, RDFHTML?).

<DavidW> David, Phil, probably Jim H willing to participate in WRLD TF next year.

thank the local host Jeremy!

very nice service!

<danbri> jjc++

We will do a good job in our first year!

Steve: thanks to Free University of Amsterdam for dinner!

SemWeb Best Practices and Deployment F2F
1-2 Nov 2004

Attendees

Contents

Day One

Agenda Adjustments

PORT TF

Wordnet TF

ADTF

XML Schema Liaison

HTML TF

VM TF

OEP Task Force

Other Topics, Day 2 planning

Day Two

RDF/OWL and Topic Maps

discussion of breakout groups

Breakout report - RDF/TM

Breakout Report - RDF/HTML & XSCH

Breakout Report - Wordnet

Wrapup

Summary of Action Items

SemWeb Best Practices and Deployment F2F 1-2 Nov 2004

Attendees

Contents

Day One

Agenda Adjustments

PORT TF

Wordnet TF

ADTF

XML Schema Liaison

HTML TF

VM TF

OEP Task Force

Other Topics, Day 2 planning

Day Two

RDF/OWL and Topic Maps

discussion of breakout groups

Breakout report - RDF/TM

Breakout Report - RDF/HTML & XSCH

Breakout Report - Wordnet

Wrapup

Summary of Action Items

SemWeb Best Practices and Deployment F2F
1-2 Nov 2004