Linking Geospatial Data Workshop Day 2

06 Mar 2014

See also: IRC log



<PhilA> Chair; PhilA

<Alex_Coley> Scribe: Alex_Coley

the first session today is: A Different Set of Problems chaired by Ed Parsons

Zhenning Shangguan - fromPitney Bowes has not been able to attend so we have a little longer for the other presenters and for discussions

<raphael> Meeting: W3C Linking Geospatial Data Workshop Day 2

Ed Parsons introducing the session

<raphael> Chair: Phil Archer

Ed Parsons: looking a little more at the applications angle

Zhenning Shangguan - from Pitney Bowes has not been able to attend so we have a little longer for the other presenters and importantly more for discussions

first up: Linked Geospatial Data, and the BBC, Chris Henden

Chris Hendon: thought we'd fixed mapping in 2005

Introducing the ways that maps were presented on the website

Chris Hendon: except we hadn't fixed it all...

Chris Hendon: moved out just from BBC Wales to cover all the BBC

Not just Wales, not just GB (slight joke at OS) all of UK and world wide interest

Location built into news - alb location an important aspect - even if the public don't know it is there or available

Location used across the BBC Website estate to help localise the site for users

A good Vocab is key - 37 Languages, place names need to be right and display as a UK audience would expect

need to be able to express shapes to a thing theta is being described - to attach content

Moving to stable relationships

now up Paul Rissen from the BBC

Paul Rissen: journalists don't like to do more than they have to

tagging happens but try to limit

Predicate Analysis example...

Journalists are very used to tagging in traditional sense

not overly useful... investigating the use of predicates to make the user experience better

Ceremonial vs Admin counties - important but most people don't care or know

BBC needs to be accurate for elections but also need to present in a familiar way

<PhilA> PhilA: fwiw I have no idea what a ceremonial county is

the BBC has had TV and Radio regions etc... does not mean much to audience

Chris Hendon: the future [of the BBC] 2022 strategy - key is user

Ed Parsons - highlighting questions will be at the end

Next a double act from the BGS - British Geological Survey

Delivery and querying of 3D spatial models, John Laxton,

<PhilA> paper http://www.w3.org/2014/03/lgd/papers/lgd14_submission_11

John Laxton - Delivery and querying 3d

<PhilA> slides http://www.w3.org/2014/03/lgd/EarthServer

Earth Server - a FP7 project using coverage data over the web

GeoSciML - GML based data transfer standard for exchange of digital geoscientific data. conceptual model

What are BGS trying to do? Some use cases being highlighted:

1. to deliver geological 3d models over web - accessible to users both expert and not

need to be explicit

2. to enable users to query the models

using Earth Server to help deliver this with partners

xWCPS being developed - xQuery enabled WCPS (web Coverage Processing Service)

<jtandy> WCPS: Web Coverage Processing Service

Used GeoSciML to describe bounding geological units in Glasgow 3d

looking if this (overall) is the best approach

looking now at how complex the querying can be and develop user-friendly interface

Next up: Tim Duffy (BGS) more on the INSPIRE side for BGS

in particular experiences with WFS2.0 services

Tim Duffy: feels already linkable data

WFS = Web Feature Service and have consistently found that implementation for real working SDIs is not great

often the optional bits in the standards are needed in the software to make it work

been working with Open Source and other software at BGS.

A lot of software doesn't meet all the needs (yet)

Need good testing facilities to test against OGC standards - these are coming

Complexity of Querying - this has been hard with rich and complex data

recognising need more dictionaries

Geology has heirachical dictionaries

Some aspects are their but not yet the tools to express or provide to users

Have had to use sub-optimal approaches such as altering schema

some query functions are VERY constrained or not well supported

Tim Duffy: recognising a disjoint generic requirements and complex information expression (information rich data)

This is common across Environmental Data - i.e. INSPIRE

move from URNs to URIs in INSPIRE is "great". Software to make this usable is behind

*noting that there are some aspects in place*

Round of applause

Panel on stage with additions of: Peter Rushfort and Tomaž Žagar

Tomaž agar: from Slovenia (Stats Office) - learning about some aspects of Linked Data but linking Stats to Location

Publishing using WFS services but looking at other approaches

Peter Rushfort from GeoGratis ( Natural Resources Canada’s web portal)

developed an geospatial api to provide access

ATOM enhanced with linkable data.

The GeoGratis API uses a hypermedia approach to delivery of linked geospatial data

Each resource is offered in multiple formats either via standard content negotiation or via typed links to API query resources - have developed tools that are available on the web

Ed Parsons: a question to all... you have complex data but you have common requirements.... how happy are you with your technology choices (1-10)?

Peter Rushford [6]

Tomaž [8] some improving

John Laxton feedback from community is mixed, first steps [??? did i miss a number???]

Paul [7]

Chris Hendon [6] some ahhs

Chris highlighted there are internal BBC issues that are not all the tools

getting what the BBC has across all uses and scale is not there yet

Tim Duffy - part the issue is the standards and tools keeping up with the standards. now catching up, over the next year performance will improve (open source) commercial software even slower. Now [7]


Peter Parslow: Publishing data to look at vs Processing. Persistent IDs are useful for both. How much of the technology stack (that is focused now on the publishing to look at) is also usable for distributed processing?

John Laxton: recognising the issue. Trying to unpick the issues a bit at a time

deliver the data service but also better descriptions of the data on top

individually not necessarily useful - together they will be.

Peter Rushford: Kiss principle is the basis of the web. Keeping things simple (i.e. hide the complexity)

Chris Hendon: Where meaning is worked out - slightly limited by the tools. Context is important and not all the tools are there to help users

Question - Jeremy Tandy: site and conformance testing - agreed with Tim for OGC. No such approaches or tools for Linked Data yet.... thoughts?

<jtandy> Q: it's crucial that we have tools to work with the data, but how might we apply conformance testing for server implementations when there is no agreed or endorsed practice, let alone standard method ... 

Tim Duffy: Love or Loath standards - they are there... if keep inventing new ways there could be issues.

Phil Archer: in semantic web world.. there are w3c test suites for many of the standards. As part of the process it is key before a standard can be a standard. Deliberately a high bar.

scribe: there are works of fiction out there but they do try very hard

Andy Seaborne: victim of the W3c process. the w3c test are there before the standard is a standard.

<AndyS> AndyS: what are the interactions in the development of the standards (and open source software)?

Tim Duffy: change requests are fed back in OGC process. Subtle difference between reference tests and tests with real data

Bart (OGC): Implimentation of standards at the OGC and the parallel Interoperability Programme http://www.opengeospatial.org/ogc/programs/ip to try and mature standards

One final question:

We have this data... how do you share it? e.g. familiar with WFS

Tomaž agar: some raster data - note there yet. for some stats data it is less of an issue

Out of time: Applause

handing over to Stuart Harrison

<pezholio> Bill Roberts chairing this session

<pezholio> "Putting the data to work"

<pezholio> Talking about tooling

<pezholio> Three 10 minutes talks, followed by panel

<pezholio> John Goodwin from OS first up

<pezholio> Talking about research project done with University of Southampton

<pezholio> Project called RAGLD

<pezholio> We haven't have a dubious acronym for a while :)

<pezholio> 18 month project

<pezholio> Allowing 'the great unwashed' to access linked data

<pezholio> Not that precious about linked data, more about open data

<pezholio> Talking about sameAs service. Linked Data is not very linked.

<pezholio> Went round the web of data, scraped the sameAs links

<pezholio> You can put in a URI and find equivalent URIS

<pezholio> URIs

<pezholio> http://sameas.org/

<pezholio> Can we do this more generically?

<pezholio> http://apps.seme4.com/see-uk/

<pezholio> Click on a ward, and see data about that area

<pezholio> Difficult to put together

<pezholio> You need a framework to make building this sort of thing easier

<pezholio> This is where RAGLD comes in

<pezholio> A whole suite of tools

<pezholio> Relationship management, spacial query services, as well as reconciliation services

<pezholio> People don't always want to get their hands dirty with SPARQL

<pezholio> Loaded wards into the service, and exposed some APIs

<pezholio> Loaded dummy data into a RAGLD service

<pezholio> Ingested and visualised a journey and did geospatial analysis on it

<pezholio> Added a buffer on a route

<pezholio> And then loaded in all the B&Bs in the buffer

<pezholio> Piping URIs to URIs, can end up with very complex queries easily

<pezholio> John has his family tree as linked data

<pezholio> A colleague used RAGLD to build an application using RAGLD tools

<pezholio> (Didn't have much experience with linked data)

<pezholio> Simple visualisation widget, also pulls in data from the web

<pezholio> Andrew Kevis from University of Madrid

<pezholio> next up

<pezholio> (May have misheard the name)

<pezholio> Alejandro Llaves

<pezholio> Talking about Map4RDF-ios

<pezholio> Allows visualising and browning linked geo data

<pezholio> Saw there was an increase in linked geospatial data

<pezholio> Not many tools out there

<pezholio> Developed in parallel with a web-based tool

<pezholio> Allows user to test their data, but also bring the linked data to non-ld people

<pezholio> Supports mobile devices running iOS

<pezholio> Multiple map services (OSM, Google etc)

<pezholio> Allows visualisation of statistical data published using DataCube

<pezholio> Clustering also supported

<pezholio> Highly customisable

<pezholio> However, it only supports non-authenticated SPARQL endpoints

<pezholio> And only one at a time

<pezholio> Running a short video demo

<scribe> scribenick: pezholio


Showing the DataCube functionality

Loading neighbourhood info

Offers an intuitive GUI to visualise LGD

Faster than the web-based version

Easy analysis

Now in the app store!

Round of applause for getting through Apple's approval process!

Future work

Save and share snapshots

Realtime data

Developing a module to feed back suggestions to data owners

Martin Hammitzsch next up

Talking about an Android app now!

Looking for advice and inspiration

Shows information about Geohazards

Such as volcanoes et


Details, reporting

Lots of configuration options

<rhwarren> Useful for disaster tourism!

Define data sources and filters

Involved in a project involved in early warning crisis management

Trying to increase attractiveness to users

They may be potential eyewitnesses in the future

App is freely available in the Play Store

Various different data sources on geohazards

Pull only, then check what's new

Push isn't officially offered, unless you ask

Variety of formats - HTML, RSS (if you're lucky), so REST API.

Not many agencies know much about linked datas

Documentation can be poor

Showing an example of an RSS feed from the British Geological Survey

EMSC have RSS with custom namespaces

GDACS interesting as they have lots of resources and link, but only internal

GNS give encoded KML in RSS

Completely different formats

Sometimes no location information at all

USGS give volcano information

NOAA have GeoRSS, but radius is badly expressed

Ushahidi being used in an international exercise

Has a REST API to feed in data, and get data out

It is a bit weak

So had to access the data directly

A lot of room for improvement

But there are opportunities

ESRI and Google Crisis response also have APIs

Is there a shift in front of us?

Geohazards affect us all

Standards have to be anchored in the community

Speakers coming back to the stage for the panel session

Andy Seabourne and ??? coming back

Sorry, missed second name

Herbert Schentz

Herbert is a member of the Environment Agency of Austria

Scientists collecting heterogeneous data

Integrating the data with LD technologies

Happy with the results, but had problems

SPARQL is slow

Missed a lot of applications for people not familiar with SPARQL, but want simple stats and maps

Andy from Epimorphics

But representing Apache

Introducing spatial data into Apache Jena

Audience is web developers who aren't familiar with geospatial data - big scary stuff!

Looking for something simple

Two constraints: Resourcing - had a student for three months

Licensing of software: Wanted to repackage and give it away

Wasn't a lot to build on

Used Lucene spatial

Had some resource from one of the Lucene spatial devs

Wanted a relatively simple index pattern

Missed questioner's name

Why not use Android?

No funding

John's T-Shirt is game of Life themed today!

Question for Martin: is he aware of the Geo super sites?

They have selected a number of sites where they collect more data

Martin: Looking into it right now, but not in contact with the team

Some problems

Push would be really interesting. Would help with early warning

Accessing data directly is very important, but there may be some time lag

Stuart Williams: Question for John - How available is the RAGLD tool?

IP is owned by OS, Semaphone and UoS

Not sure what to do with it yet

<HadleyBeeman> s/semaphone/seme4

Quesiton for Alejandro and Martin

Which was the easiest linked data geometry to work with?

John: Built RAGLD to work with shape files and RDF

Used WKT

As well as W3G geo lat/lng

Alejandro goes for GeoSPARQL as he likes the model

Not supporting GeoJSON yet

But depends on the use case

John says they use GeoJSON on the client side

Andy - ordering of lat and long in W3C standards

Herbert Schentz: Which format is nice for the user, and what format has performance benefits?

Question from MET office

The target for tsunami warning messages is 2 seconds, but a lot of warnings aren't on the web because safety critical

Question for Herbert and Andy: We've heard that SPARQL is slow, but it can be optimised

Who's fault is it? The query writer, or the application developer

Herbert likes GeoSPARQL, but has great concerns over the performance

Andy can't speak for GeoSPARQL, but knows people who have worked on it

There is a trade off, but if you tell the system what your data looks like, it can be quicker

Optimising for data and geo queries is quite a difficult job

Missed question from Bill

John: With RAGLD, they wanted to provide RESTful layers on top of SPARQL queries - hiding SPARQL

Alejandro: Putting a visualisation on top of SPARQL does the same - non linked data people don't like SPARQL

Saves you the burden of creating your own querie


Andy: Mainly looking at the application writer, where there is some geospatail data, but other non-geo data

Martin: about putting the data to work. People generally are not interested in the technology, they want to see what's possible
... SemWeb and SPARQL is scary - you have to demonstrate the advantages

Herbert: You have to cater for all audiences, shouldn't completely hide SPARQL endpoints.

Alejandro answering question as to why Linked Data is better than standard Geo

Linking is the definite benefit

Question on integrating spatial and statistical data - what's the best tool?

Where in the stack is the best way to do it?

Alejandro: Can visualise DataCube

Anything with a SPARQL endpoint can be integrated

Session over

Now, coffee!

<AndyS> scribenick: AndyS

<scribe> scribe: Andy Seaborne

subtopic: Implementation of the Fukushima radiation LOD framework // Isao Kojima, AIST (National Institute of Advanced Industrial Science and Technology)
... Open Data-Logging for Bikes // Rich Boakes, University of Portsmouth

(reorder talks due to setup issues)

Rich: With Jacek Kopecky and Oliver Williams
... Portsmouth is nearly an island
... and very crowded
... sustainable transport ... now getting gridlock due to how busy things are becoming
... bikes. Safety - actual and perceived
... deaths in London up 18%
... and 6 in 2 weeks
... also perceived putting people off.
... as scientists, there is no data
... (video)
... a close encounter of the bus kind
... gather data and observe. Publish data and understand the reality.
... and maintain data published
... including the data gatherers : crowd, less good data vs curated data
... bicycle with instruments : RasperberryPi+GPS+ultrasonic sensor
... prototype 2 adds a panic button - records user experience
... add a camera pointing backwards -- number plate recognition -- coordinate with history of the vehicle
... better hardware - water proof
... in the future - accelerometer ,...
... example of heat map of a single bike journey. Proximity events.
... aggregate by many journeys
... effectively target council road funding - provides evidence of effect of spending
... link to other data
... http://picycle.io/
... license for software and for data to be decided
... hardware hackday coming up
... no funding currently ... looking for funding.

subtopic: Implementation of the Fukushima radiation LOD framework // Isao Kojima, AIST(National Institute of Advanced Industrial Science and Technology)

Isao Kojima: GeoSPARQL + ELDA

scribe: Monitor/construct/publish radiation database of the Fukushima area
... role -- interface to database for integrating other open data
... real time data (4K points, every 10 mins)
... (more)
... CSV to RDF for stats
... Sesame + UseekMe
... interlinked datatsets e.g. population, radiation, geo
... ELDA endpoint + extend to WKT polygon display (no search in browser)
... will contribute back to open source project
... calculation of collective radiation dose combines different sources
... issue 1 - schema design and query performance
... query touches a lot of data to get all sub details of hierarchy
... geo data at the lowest level of hierarchy
... issue 2 - support /combine two open standards
... different stores for different data
... create an aggregated database in PostGIS
... maintaining consistency hard
... been going for 6 months with RDF based framework

subtopic: A geospatial API for linking and distributing open data - researching new ways of linking similar objects from different data sets // Bert Spaan, Waag Society

Bert: CitySDK Linked Data API
... EU project
... collect these datasets and provide in a unified way
... useful for city apps (mobile)
... e.g. car park : capacity, location and name, from different silos
... pin data to spatial location
... (video)
... info delivered in JSON
... linking is quite static - want to be dynamic
... add rules for connections within the data
... example based on sameAs and geoproximity
... categorize objects maybe by crowdsourcing
... fuzzy URI -- e.g. a car park somewhere near XYZ
... City Data Ontology
... specific relationships for city data

subtopic: discussion

Rich: most important is the data license
... want to do it properly ... suggestions please

<BartDeLathouwerOGC> check out the OGC Table Join Service (http://www.opengeospatial.org/standards/tjs) for linking geospatial with tabular information

<pauljcripps> license question came in from Anthony Beck re this post on the DART project blog: http://antarch.calepin.co/dart-licences-and-the-ads.html

Isao: no general tools but we were able to map the data

Rich: tricky
... if the gain is great than maybe should be done - wide debate
... common issue that arises with open data

Bert: yes for Amsterdam , more coming

Rich: Members of the public. We need to thing about it as the public.community.
... is it a responsibility of owning a motor vehicle to have a public number plate?
... data in geoJSON but how should we publish and use internally to maximise additional use

Bert: document the many datasets and feeds. Currently assumed knowledge.

subtopic: A word about the SmartOpenData project // Jesus Mª Estrada Villegas, TRAGSA http://www.smartopendata.eu/

Jesus: FP7
... first aim to work with data sources to make them open
... environment sources
... contacts on website

<PhilA> scribe: PhilA

<scribe> scribeNick: PhilA

Bill Oates Help! Where next for sustainable statistics and geographies for Wales?

Bill: Asks for help by Twitter
... the Welsh Government produces data
... work to civil code of standards etc.
... concerned about security etc.
... we're a Microsoft House
... we're not data geeks
... we produce reports. It remains the most used way of accessing our data
... the stats page is the 2nd most used page after the jobs page on our Web site
... we get FOI requests
... if we publish our datasets that should fend off a lot of FOI requests. Stats Wales for example, 1k data cubes there
... no new data in the user friendly app, but it's digested
... we're being tempted by OData, and Google's offering
... feel that same people are producing data and the apps
... final slide shows what I need

How to move a relational Roads Dataset to a Roads Linked Data Dataset?, Alexander Ramage, Transport Scotland

AlexR: I'm and IT person, not an LD person. I'm convinced of LD. How dod I do it?
... got lots of data, relational
... we have a design manual for how you design roads and bridges. That includes some controlled vocabularies - we should give them URIs. That's easy
... what about stuff that isn't finite
... What about spatial information - most important slides for me
... dunno how to move into the LD realm
... what I do a lot of is to compare one bit of data with another
... Two lines are identical but they're in different places, How can I prove that they represent the same thing?
... Transport Scotland thinks this is the right way to go.
... but we're not going to move to it quickly
... putting building blocks in place is the next thing

Collaborative Geospatial Data, Stuart Harrison, The Open Data Institute

pezholio: We're planning to...
... Geodata collected by a range of organisations and they don't always use the same standards
... this is a common problem
... much of hte geo data that devs would like to use is not open
... Open Street Map is one answer
... but OSM is not the best way to show where your event is, for example
... an OSM URI doesn't necessarily give you the data back from the URI
... the Share Alike licence is not always what's needed
... storing geo data in one central place may not be the best fit
... there's lots of open data out there with geo data included
... Green space data is very disparate
... allows users to add in more data
... can be done directly or can send us data that we can include (if unincumbered)
... we want to build an open addresses database
... the land registry is an invaluable data source. Every price paid for a house since 1996
... add in Companies House, OS Locator
... started some of this work - see project on github . Need more funding to do more

Many layers of thought: connecting INSPIRE and Linked Data, Peter Parslow, Ordnance Survey

<BartDeLathouwerOGC> ODI : using POI? (OGC OpenPOI)

[No slides]

PeterP: TimbL said OS is really important so people can put data on a map. Glad we haven't been talking about that here
... I work in product development, things that will launch in 2 years or so
... trying to get our linked data out of research and into normal ops
... Do let us know if you use our LD, otherwise it might disappear
... server logs only tell you numbers, not how useful it is
... Talking about history of triples, relational data, XML anda geo data
... things have grown from small groups into large groups.
... now want to get back to troiple model that we actually started with
... It's not possible to create an ontology of everything
... INSPIRE made a lot of effort to get semantics into the model, LD people should use it. INSPIRE shouldn't try to do more semantics

Representing and Reasoning over Spatial Relations in OWL: A Rule-Based Approach, Sotiris Batsakis, University of Huddersfield

Sotoris: Talking about need to update a lot of things when changing one thing
... shows the RCC5 relations
... you can draw inferences from known relations
... we use Path Consistency
... using SWRL to encode rules
... allows us to deal with incomplete info

slides reflect the notes fully

Q & A session

<Gianfra> hello

question: Your open space project, pezholio - a lot of local gov data includes Ordnance Survey data, how can you move into open data

StuartH: Derived data and licensing is a challenge

PeterP: TOIDS are free to reuse. If local councils want to provide the data they need to talk to us. They give info about paths which they draw on our maps

<pauljcripps> sorry that question re licensing was from me.

Gothwin: A question on RCC5 reasoning - can you do it in OWL? I tired and couldn't

Sotoris: The full reasoning is incomplete. So we restricted it - 2 tricks used

(lost the details)

question - you need to sweet talk your local authorities to ask OS for exemption. With luck they'll say yes

(this for StuartH)

Some discussion of the difference between Open Gov Licence and Open Licence for Ordnance Survey, only diff is that you must credit OS

orcho: You've been exploring using LD for stats (Bill)
... have you looked at using it in the pipeline

Bill: I'd be intersted in case studies into those benefits
... I think it's not just about external apps, it's the benefits internally that I'm equally interested in
... the challenge is dealing with civil servant's addiction to Excel
... how can people benefit from LD within the organisation via desktop tools where we're locked down

<rhwarren> Hum... Bill is leading the lynching mob againts excel!

PeterP: What about sustainability, Bill?

Bill: The key point for me is not doing Linked geospatial Data, it's doing it again x lots
... if people download our data and so stuff, great, but will they be there next year, the year after?
... what's the first thing I do to move from Excel to open data publishing
... the retention period on gov data is 20 years. After that we have to follow one of a few routes

question: Lots of people dealing with the validity of data, called citizen science

scribe: several years ago I was reading about folksonomies. Are you working on that

bill: all the issues around provenance and reliability of the data are important.

<Gianfra> I meant citizen science

BillRoberts: On the issues of digital preservation... standard desktop tools are part of the problem. Publishing LD gets around a lot of preservation probs.

SteveP: I'm an OD person, I'm also an OSM enthusiast - people at OSM are cross referencing physical things on the ground with boundaries

AlexR: That's not happening in Scotland. Only seed points open in Scotland, not available nationally

<HadleyBeeman> scribe: hadleybeeman

<scribe> scribenick: hadleybeeman

Panel: Show Me The Way

<scribe> chair: stuart:

Stuart Kerry Taylor will be speaking on behalf of herself and Simon Cox (who isn't here)

subtopic: Geospatial standards and linked geospatial data (Kerry Taylor)

Kerry: Subtitle of my talk is "Simon Says"

… I'm standing in for Simon today.

… Simon wanted to present to this community, esp W3C audience, state of work in the OGC and ISO.

… OGC has developed a URI scheme and a naming authority. A definitions service.

… ISO/TC11: rules for using OWL, ontology management group.

… URIs for all OGC artefacts

… ISO/TC211: rules for developing ontologies in OWL as a draft international standard

… Preview at http://def.seegrid.csiro.au/isotc211

… Includes rules for translation from UML modelling into our modelling

… Not geospatial data. Can help modelling and classifying it though.

… End of Simon's points. This is Kerry speaking now.

… These are technical as well as cultural bridges.

… The OGC community has lots of knowledge in spatial data, and industry consumers.

… The W3C has a good profile, high uptake, and semantic web knowledge to contribute.

… Specifically, we can get inside complex data representations.

… Finally, with W3C hat on towards OGC world: keep the geo-location and geometry out of other stuff that goes with it. Don't put these things in single packages.

… Hardly anyone here belongs to both W3C and OGC.

… Needs to be considered for the sake of interoperability.

subtopic: Keith Jeffery

Keith: I come from the big data world

… 2 projects: EPOS - volcanoes (?) and Engage - research data.

… I'm interested in relationship between linked open data and heavy big data.

… Linked open data isn't new. Used to call it 'interoperability'. Goes back to the 1960s.

… Triple stores: IBM in the 1970s. In the 80s: 9 column store. Triples RDF + spatial + temporal + modal

… Dealing with stuff modelled into RDF can be hard. 2 problems:

… Danger of losing referential and functional integrity. Queries can produce wrong results or not be computable.

… Metadata: important to let you do something with data. Critical. 1. If good, contextual metadata, helps end user to judge quality and applicability. 2. Allows disambiguation. 3. provides lowest common level across datasets to combine.

… Way forward: I go back to Tim's view of the Web: to publish and share information.

… Use the LOD semantic web for browsing/discovery, close to user interface. Need high throughput systems for real world data processing. Link together with metadata.

… Virtualise on cloud infrastructures

… Use it all to look for potential links in data. Use to generate converting software to bring together data sets at lower level, high volume.

subtopic: John Goodwin, in place of Raj Singh

<richboakes> Isn't LOD is more that interoperability? Interoperability suggests (to me) that services/data are designed to be used together. The challenge of good LOD is to design data so than it can be used _somehow_, but we may have no idea what it might one day be used for.

gothwin: 3 years ago, we wanted to publish linked data for 3 of our products. Best practice was to use ontologies, or extend existing ones.

… I needed ontologies for modelling spatial info, but they didn't exist. Had to write mine own. Nothing to model geometries apart from WGS84.

… At a Geocamp, got to brainstorming with database vendors… ended up being what GeoSPARQL looked like.

… A few things missing from GeoSPARQL. I'd like other predicates, not just geometry. Boundary, extent, centroid.

… Also missing metadata re spatial data quality.

… I invented my own, but I don't really like it.

… As a data consumer, I work in a research dept. Look at: how do we ingest other linked data?

… We need simpler ways to ingest it.

… People say linked data is a steep learning curve. But it's actually really simple.

… How do we remove that perception that this stuff is hard?

subtopic: Steve Peters

Steve: I'm a publisher and a hobbyist developer, so looking at this from both roles.

… Start with a question: Is the spatial planning system encouraging or discouraging the building of new homes in flood risk areas?

… Ask a floating voter , they'll say "place" means "my local village/community".

… Ask a policy person in central government, they'll say "place" means "the nation." Want to compare different areas and regions, suceptibility to flooding.

… Need data from different sources: local authority, Environment Agency, nationally. Need to link those sources together.

… This is much more than maps. Need spatial intelligence and insights about those areas.

… Gaps: Do we need new standards? We probably have enough. I've been hearing here that standards exist and we aren't useing them.

… We need to better understand what's out there, how to reuse the patterns.

… Re GeoSPARQL, looks promising as a publisher. But not many examples of where it's running in the real world.

… To make standards stick, we have to put them in procurement processes.

… We have legacy systems, with third party suppliers. How to educate them on these benefits?

… Re INSPIRE, it's a focal point for action. Maybe we can use this agenda to drive out the new standards?

… But INSPIRE is hard to get into, as a developer.

… Re culture and language: if you're a geospatial developer, AJAX is not a domestic cleaning product. If you're a web dev, a polygon is not a dead parrot.

… I see the emergence of open layers, leaflet JS, etc. Good stuff.

… Linked data IS the answer, bridging the geospatial conventional world and the stuff we need to know about it.

subtopic: Raphael Troncy

raphael_: Andreas was supposed to take this slot, but it's me.

… We agree on what the future work should be.

… Re vocabs for describing features, geometries: Need more predicates

… We start with the existing vocabularies — no controversies in them. We need more properties to express the relationships

… I agree with gothwin

… We talked about how we could structure the representations (nodes and points)

… We talked about GeoJSON output

… We need a vocab for precision of maps.

… Many maps charge when you get below a certain resolution.

… A vocabulary for the semantics there would be useful.

… SPARQL and GeoSPARQL are nice, along with APIs.

subtopic: Alex Coley

alexrcoley: I'm with the Department for the Environment, Food and Rural Affairs in the UK

… We have 39 arms-length bodies connected to Defra, small to large organisations.

… Most of them use data to drive their business, most of it is environmental and attached to a place. Location is important.

… We're the lead department for INSPIRE in the UK.

… Open data and maximising the use of our data are also key.

… We're learning about interoperability from our incidents (like current flooding). Sharing data is easy.

… Money is tight. For us, moving away from the silos. The problem is duplicative standards.

… If different communities are doing things in parallel, and we have to do things more than once — that's not good for us.

<AndyS> sharing data is easy ... but hard if you want to do it now

… There isn't necessarily always one way. We have a lot of existing standards we can reuse, but it's hard to know which to use when.

… We should be a bit more coordinated.

… Less is better than more.

… I may have come to this a little biased, and I've learned through our discussions.

… We have the traditional Web community, and the Geospatial community: discussing the same things, but with different words.

… Coming together would be useful.

… What's the best way, for data available on the web (as opposed to through the web)? Still not convinced we've demonstrated one way.

… Joint W3C/OGC working group? Or just a community group?

subtopic: Panel Q&A

Stuart: One pass down the line of panelists before we open up to questions from the audience.

… What is on the top of your shopping list for some combined OGC/W3C effort?

Steve: Helping me as a user navigate through the maze of standards, best way to prepare, publish and encourage reuse of data. Best practice, support, peer networking.

Keith: Standardising terminology of metadata.

Alex: Lots of parts of the picture. Bringing those together to demonstrate a whole way it can be done.

John: +1 to what's been said.

… Agreement on best practice and the standards, tools to implement them.

Kerry: Shared spatial vocabulary. Useful and easy to do.

Raphael: +1 to Kerry.

Question: Bill Roberts: My wishlist: I think we have enough standards, but I don't know how to use them. So: design patterns where people can document how they use standards.

Raphael: For GeoSPARQL, hard to see good examples of endpoints. More would be great.

Steve: To convince my bosses to fund this, evidence of the benefits are important. Case studies, evidence of take-up. What you can do now that you couldn't do before.

Question: Peter: +1 to last point re management oriented business benefit. On previous point, sounds like the SDI cookbook of 10 years ago or so. Maybe an update? SDI was good for publishers, but not for consumers.

Question (Bill Oates): open standards gain traction. Last year, we published what we thought was WMS, and also published WFS because we forgot to switch it off. Therefore, didn't need to make a business case to implement it.

bill Oates: Tools can fall behind the standards. Would they help here, or make it worse?

Alex: I agree. That balance: new standards? tweaks to the standards to make things easier? How to link between different approaches? And when to? Doesn't necessarily mean changing the implementations.

Keith: Horses for courses. Linked open data is good for somethings. We need to build up experiences and case studies. Share experience. Reflect that back into the standards bodies. Bring things forward with the whole community, not just us in this room.

Question: Jack Kareski(?): Many say linked data is hard. We have to reuse ontologies. When building an SQL database, you don't have to make a schema with others. One good practice could be: do it in SQL, give us a read view in RDF.

PhilA: D2RQ.

Question: Jeremy Tandy: We've identified 4 geospatial vocabs for identifying geometry. Could we get it down to 1? Agree it? NeoGeo? Core location? GeoSPARQL? W3C Geo? Can we have a vote?

Raphael: I'm not sure of OGC practice, re GeoSPARQL vocab. More familiar with W3C. Now, small sets of properties we want to add are being discussed in the Core Vocab group.

… We should add these to that discussion, or feed back to OGC for their process.

Question (Tim Duffy): +1 to Jeremy's point.

Stuart: Any dissenting views?

Audience: none

Question: Chris Little: As an outsider, is it possible to get to just one?

Question: PhilA: When we began this workshop, we didn't know what the outcome would be. However, reasonable to think now in terms of a chartered working group. Now need to discuss legalities and practicalities. Outputs jointly badged W3C & OGC.

PhilA: It needs an actual charter, which fits both orgs. Members are in control.

… Pretty quickly, we could standardise GeoJSON. Been around a long time.

… Cry for a shared spatial vocabulary. We can do that. May need converters/mapping.

… cookbooks, best practices… we can do tht.

… Usage studies are harder. Perhaps an EU-funded H2020 project? We could learn from it?

… Do we need to take a wicket literal and split into semantic bits with predicates. Is that useful?

Raphael: yes, though there are different ways to write it.

PhilA: K

… sometimes you need the metadata for each value in a coordinate pair. Can some of this be standardised?

… Compression may be an issue as well. Some of this data is big.

… SPARQL performance increases, etc.

Bart: My notes are similar as PhilA's.

… It's not PhilA or me who make the standards, it's you. We help you and guide you through the procedures.

… To bring the experience of using the standards back to the standards bodies — to make the standards better. That's the way to go.

… That's an action item for all of you.

… Action item for me: OGC's upcoming technical committee meeting, 3 weeks from now: I'll report back to working groups what happened today. Will start looking into taking on this work.

… I get great value from this. We have to work out how to keep the momentum going.

… Thanks to the panel.

Kerry: Geospatial data is not a European issue, travels further outside. Plea for us further away. We have a lot to contribute.

Bart: Upcoming TC meeting is worldwide. It's in the US, next one is in Geneva, then Tokyo.

PhilA: W3C is worldwide.

Stuart: Lots of us are from small companies. Global travel can be hard.

Jeremy: most of the work happens between meetings, remotely. You can participate remotely.

Kerry: in the middle of the night.

Question: to add to Bart's comment re OGC looking into spatial vocab: there is a group already discussing this issue. Until this joint W3C/OGC group starts, could we have the discussion there?

<PhilA> The Cg that Michael talked about

Raphael: We're glad to support this as W3C members.

Alex: Most positive thing I heard was Bart's list is similar to Phil's. There are definitely things we can do together.

Keith: +1 to phil. Horses for courses. Layered metadata. Interesting thing is the relationships between those layers ensuring consistency and integrity.

Steve: Rapid incremental progress is good. Quick visible fixes to JSON or JSON-LD.

Stuart: thanks to the panelists.

Barcamp session: Fuzzy URIs

The notes from this session are in the form of a diagram.

Barcamp session: Determining which RDF ontologies provide best practices in areas of overlap

There were 12 people involved in this barcamp discussion (names listed at the end of this report).

As a starting point we noted that vocabulary reuse is important, but that the adoption of existing vocabularies needs care. It is possible to use the Linked Open Vocabularies site as a first port of call, but there are even issues here around the “semantics” of “vocabulary” and that Linked Data begins with the first star. It was also noted that sometimes the vocabularies deemed authoritative may not be up to date, and so governance is as issue. For example, the EPSG register of Coordinate Reference Systems (CRSs) may not be in sync with the, e.g., French register of CRSs which may deprecate and introduce CRSs frequently. Can issues of a federated mechanism for authority / governance be investigated?

Following these opening conversations, we examined the topics of the Open Geospatial Consortium (OGC) Abstract Specifications, and brainstormed for potential vocabularies that we use / know of which cover these topics.

What are the overlaps between the OGC / ISO / W3C vocabularies

It is worth noting that some of the Abstract Specifications cover outdated topics, so we concentrated on those that we continue to use. The overall concepts of the Abstract Specifications are quite durable as the central concepts are fairly stable.

There many more domain specific ontologies we could look at in a second sweep of the available resources. There are also other documents that could be looked at such as OWS Context.

What can we do to fix it?