See also: IRC log
<HadleyBeeman> scribe: hadleybeeman
<scribe> chair: PhilA
PhilA: first thing is to welcome stefan and deiredrelee
… I'm keen for this group to expand. Core is people who are currently working group chairs or have recently chaired a group. Beyond that, want to invite people whose opinions we value, can help steer the Data Activity.
… So I've asked Stefan Decker and Deirdre Lee. I've also asked S�ren Auer , who has a teaching conflict this term.
SDe: (introduces self)
<danbri> since 1998 :) http://www.w3.org/TandS/QL/QL98/pp/queryservice.html
deirdrelee: (introduces self)
PhilA: Every time we meet, we need to have a general catch up on each working group.
<SDe> Thanks for the warm welcome!
<PhilA> Clinical Pharagenomics is not a word you hear everyday
ericp: In HCLS, there are 2 areas of interest. Clinical pharmacogenomics (deep OWL modelling of various biological/genomic features), and standarisation of RDF representations of clinical records.
… Those standards are challenging (depth/breadth of the data), and convincing a largely XML-based world to change is hard too.
… Working with CCDA in the US government, emerging standards from HL7 like FHIR, and SIMI. Want to provide mappings to a middle ground in RDF.
PhilA: The Data Activity is part of a new domain at W3C called I&K. Liam Quinn, XML editor, is beginning to contemplate a workshop end this year (ish) around XML people + RDF people.
<danbri> XML <-> RDF permathread. http://www.w3.org/TR/1999/NOTE-schema-arch-19991007
… Reasons: Liam sees more XML people using URIs as identifiers.
… Then: issue for them: should URIs be dereferenceable?
… RDF people aren't keen on XML
… xquery is being used to create a subset of RDF that can be queried using SPARQL, etc.
… Areas where each group is seeing benefits on the other side. Therefore: a workshop on that general theme.
<Zakim> danbri, you wanted to grouch "so _now_ they want to talk to us…"
danbri: A workshop and a community group would be great.
… But there aren't largely disjoint communities. Lots of crossover. Over when XML is becoming eclipsed by JSON that we have a 15-years-too-late conversion discussion.
… Merging the technologies would be ridiculously delayed. And grand.
PhilA: That's useful. We should talk more on this.
… What should we do here?
danbri: When people were terribly excited about XML, RDF was seen as a threat.
… Different tools for different jobs. Best practice guidelines would be more useful. Detailed cooperation at the spec level on data types and functions.
<davidwood> XML———————————————A good mix——————————————RDF
danbri: Update on CSVW: We've published a use case document and a doc to provide a data model for CSV
… Not a lot of feedback
… Since that, we've started to look at RDF mappings
… Most recent conversation: with Rufus from OKFN re their packaging spec (renamed now?) as a starting point.
<danbri> re datapackage.json
… editing situation: probably jeniT and rufus. But not sure yet.
<PhilA> scribe: PhilA
<scribe> scribe: deirdrelee
HadleyBeeman: working on
use-cases,grappling with nailing down the scope of best
... we had a successful F2F in London 2 weeks ago, fleshed out the details of the use-cases to map to the deliverables
... we have couple of deliverables including best practice and two vocabs
<PhilA> Arnaud's report from LDP
<HadleyBeeman> scribe: hadleybeeman
<deirdrelee> ... we want to make sure we have enough use-cases to cover the wide scope of deliverables
phila: The LDP working group is meeting F2F right now. See review from Arnaud.
… they're already talking about LDP next, which is really active. 6 implementations from within the group itself.
… Another group that Arnaud and IBM are interested in: EricP's work on RDF validation
…(and all the various other names)
<danbri> Sept 2013 workshop, http://www.w3.org/2012/12/rdf-val/
EricP: In Sept we had a workshop, 30 people trying to do RDF validation with various tools.
… Everyone wanted a simple declarative way to do this.
… 2 techs were foremost: Resource Shapes (IBM/Puppet)
… and Application Profiles (Dublin Core)
… Call for a simple user-facing syntax pushed me to create Shape Expressions. Graph topology -> people working in this area like that terminology.
… Not helpful for those not in this area.
… Trying to start a working group. IBM submission on Resource Shapes. I'll publish a submission on Shape Expressions. Hopefully that will be enough to start things.
… We have not started chartering. Mailing list for discussion: firstname.lastname@example.org
philA: I see huge demand for this. Research Data Alliance is looking at something similar.
… Many people think something like this already exists.
… Problem: we don't necessarily have the demand within our membership.
… What do other people here think about this?
FabGandon: What do you need?
philA: No clear answer there. Sufficient members who believe it's important enough and will contribute enough for that group to be successful
… at least 4-5.
… On top of that, want other members to vote yes (though they may not take part in it)
… It does help if there is support from big members.
EricP: DERI is already supproting this they siad they would cosign on the submission.
danbri: Is there a community group?
EricP: There is a group of people who gather for Dublin Core Application Profiles, shifting toward Shape Expressions because they fit their use cases.
<FabGandon> Olivier Corby in my team reviewed ShEx, gave feedback and showed interest
<Zakim> davidwood, you wanted to suggest cooperation with RDA
… So there is a community group but not a group of people hacking on it.
davidwood: The Research Data Alliance told me they don't interact much with the W3C.
… But they said they would like to. Is this an opportunity to start things with them?
phila: I've been to last 2 RDA plenary meetings
… There is a discussion in process to create a formal, deep liaison between W3C and RDA
… We're trying to work out where each org fits. I've joined the advisory board of the RDA.
… There are areas where the scientific research community isn't interested in our work.
… But where they are using the web to publish/exchange data, there is potential.
… It's like what we have with the IETF
<danbri> (url for RDA?)
… Then: discussions about where the money comes from? We're talking to the European Commission.
… Hoping there will be a joint W3C/RDA workshop in next 12 months.
davidwood: Thanks, PhilA. The people I was speaking to didn't seem to be aware of that.
PhilA: When did you speak to them?
davidwood: a month or three weeks ago?
PhilA: Conversation ramped up a lot at plenary they held last month.
SDe: I've been talking to S�ren Auer who agreed to cochair a linked data interest group within the RDA
PhilA: Thanks, could be helpful.
danbri: When we did the RDF schema work, we had people with XML background. This piece of work could address some of the big quirks that separate the two.
… With a Google hat on and maybe schema.org hat on, we might be interested in seeing this happen. For Google: patterns, RDF graphs.
… schema.org: patterns against RDF graphs. Stepping stone to a more standardised solution
… We've just published schema.org actions
<ivan> http://blog.schema.org/2014/04/announcing-schemaorg-actions.html schema.org actions' announcement
PhilA: I asked FabGandon to come along because I saw your tweets. Hope you can give this group a taste.
FabGandon: (introduces self)
… WWW2014: 300 participants, 24? countries
… Plus a big conference.
… I don't know how to give a full overview, but I can remind you there were a number of tutorials. 1 on entity resolution on the Web of Data
… Tells me the subject has come to some maturity
… 1 on online learning and linked data.
… I did one on Semantic Web and linked data.
… Usual suspects were IDO? on the Web.
… There was a presentation on linked data fragments. Offload the queiring from the server to the client. Lightweight mechanism between the clients and linked data server.
PhilA: Tell us more about linked data fragments? Can you focus on stuff that was unexpected/interesting?
FabGandon: Linked data fragments. Idea is that you have the mechanism for the linked data server to publish subqueries that can be recompiled by the client to get the data that the need for a full complex query.
<danbri> (kicked off for posting too much?)
<danbri> summary of Dan's hasty comments earlier re RDF validation and shapes: at Google we are potentially interested in following/using and maybe joining WG in this area, to help characterize graph patterns, we use prose + examples so far, e.g. https://developers.google.com/webmasters/business-location-pages/schema.org-examples
<danbri> Also today schema.org launched 'potential actions' design, http://lists.w3.org/Archives/Public/public-vocabs/2014Apr/0242.html http://blog.schema.org/2014/04/announcing-schemaorg-actions.html … this is about providing descriptions of _future_ actions, e.g. 'book a flight', 'RSVP an invitation'. In creating this we used an ad hoc mechanism for characterising graph patterns; may be interest later in plugging in something more standard.
… That way the server can be very efficient in caching.
… You know how big each fragment is. Can decrease the data flowing between the client and the server
<PhilA> Linked data Fragments
PhilA: Do you think this is a good thing?
FabGandon: It's a completely different approach from the previous one using indexes
… Several people agreed that a bottle neck for linked data is that we're trying to make the server too intelligent. Therefore we should make the server dumber, and to distribute the computation.
<PhilA> Mr Linked data (S�ren Auer) and Peter Colpeart
PhilA: At the European Data Forum in Athens. ^
… This puts more intelligence in the client.
… Remember the beauty of linked data is that the intelligence is IN the data. Is there a danger in moving the intelligence to the client?
FabGandon: This is moving the intelligence from the server to the client, not from the data.
… They're doing this for SPARQL.
… You can imagine a client doing all the things that SPARQL solves.
… Data mining, using the same approach, not relying on the server.
… The processing needs to happen somewhere, and this distributes the server and client loads.
… This is not about replacing the old approach, just creating an intermediary solution
Hadleybeeman: can you explain that?
FabGandon: If you wanted to send a full SPARQL query, you could do that. And the load would be on the server side.
… But through the fragment would let you do it on the client side, if you wanted.
PhilA: That matches what Peter was telling me
FabGandon: Another thing that caught my attention: a presentation on RNL source-agnostic mapping
philA: Surely there must be some constraints on that?
FabGandon: I think the idea is that they all comply with a relational model but don't have to be in a relational database.
… Their model mentioned was CSV
<ivan> anastasia (sp.)
<danbri> who is in the wg
PHilA: Were they aware of CSVW? Do you remember who it was?
<Zakim> ericP, you wanted to ask if that was a relational model or an SQL model
ericP: Clarification: was it really a relational model, or a full-on SQL model?
FabGandon: No idea.
<danbri> ("does it have 10 types of nulls, joins etc"?)
Ivan: Generalisation of R2RML.
… Instead of SQL queries, you can put something else.
<PhilA> Anastasia Dimou https://twitter.com/natadimou
… I hope I wasn't tough on her. My feeling is that this is way too complicated for 90% of the users of CSV, so I pushed back on doing this on a working group level
PhilA: Anything else that caught your attention, ivan?
Ivan: No, I spent more time on Web Science track.
… There were possibly relevant presentations there though.
… One more thing: workshop from Web Obervatory. Maybe more useful for DWBP.
… Set up a joint agreement, "standards", to publish data about the Web and Web-related things.
… Store visualisation methods and tools.
… Jim Hendler presented on vocabulary he's put on schema.org to describe this kind of data.
… There is a move there in that scientific community to set up big data storage/repositories.
… Wendy Hall is very active in this.
… They are setting up such repositories by RPI in Korea and other places. A bit of an old boy's network, but active.
FabGandon: Keynotes weren't especially relevant for this call, but Microsoft's mentioned their Microsoft Office Graph.
… They'll use it to connect people to documents, events, etc. And to do machine learning on that graph.
PhilA: underlying technology for that?
FabGandon: I'm pretty sure it can be an RDF graph. He didn't get into the detail.
… They were saying they have big opportunities through the Office graph and do machine learning on top of it
<danbri> clarification: Phil's ahead of his time not 15 years out of date! The XML community's belated enthusiasm for RDF is 15 years out of date.
hadleybeeman: is this a tool with implementations of Office? Or will they run on it on all the data they're hosting for users?
FabGandon: He didn't go into that
level of detail.
... On the Semantic Web track: 5 of the 6 papers are relevant here.
… Everyone is looking at a different way of doing semantics.
… Nothing is being done with OWL
… 3 papers on data management. One of efficient provenance. One on cube analytics on RDF.
… One on rewriting R2RNL rewriting of SPARQL queries to SQL
… RDF reification.
… Introduction of unit tests for linked data, released with the schemas for the data
philA: Which takes us back to RDF validation
<danbri> (aside q: how many attended?)
… We have events coming up.
… Opportunities to get together?
<danbri> I'll be at TPAC
… Who is coming to TPAC?
<davidwood> I will be at SemTech and hopefully ISWC
PHilA: perhaps dinner at TPAC?
… Next meeting is on 7 may
<PhilA> DNM 7 May 2014
… Usually first Weds of the month. Unless rescheduled for the third Weds