Hide the Stack: Toward Usable Linked Data

1,048 views

Published on

The explosion in growth of the Web of Linked Data has provided, for the first time, a plethora of information in disparate locations, yet bound together by machine-readable, semantically typed relations. Utilisation of the Web of Data has been, until now, restricted to the members of the community, eating their own dogfood, so to speak. To the regular web user browsing Facebook and watching YouTube, this utility is yet to be realised. The primary factor inhibiting uptake is the usability of the Web of Data, where users are required to have prior knowledge of elements from the Semantic Web technology stack. Our solution to this problem is to hide the stack, allowing end users to browse the Web of Data, explore the information it contains, discover knowledge, and use Linked Data. We propose a template-based visualisation approach where information attributed to a given resource is rendered according to the rdf:type of the instance.

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
1,048
On SlideShare
0
From Embeds
0
Number of Embeds
2
Actions
Shares
0
Downloads
5
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Hide the Stack: Toward Usable Linked Data

  1. 1. Hide
the
Stack:
Toward
Usable
Linked
Data
 A.‐S.
Dadzie1,
M.
Rowe2
&
D.
Petrelli3
 1.
The
OAK
Group,
Dept.
of
Computer
Science,
The
University
of
Sheffield
 2.
The
Knowledge
Media
InsPtute,
The
Open
University
 3.
Art
&
Design
Research
Centre,
Sheffield
Hallam
University

  2. 2. Key
Message
•  Linked
Data

 –  connecPons
between
disparate,
independent
(albeit
related)
data
 –  rendering
public
interest
data
accessible
 •  allowing
hidden
informaPon
to
be
discovered
more
easily
 •  enabling
quesPons
to
be
answered
more
fully
 •  PotenPal
widely
recognised,
but
 –  very
large‐scale,
wide‐coverage,
highly
inter‐linked
data
repositories
 –  under‐uPlised
outside
SemanPc
Web
community
•  Aims
of
the
research
 –  explore
new
methods
for
presenPng
Linked
Data
to
wider
audience
 –  support
more
intuiPve
exploraPon
and
knowledge
retrieval
 –  encourage
wider
use
by
web‐savvy
but
non‐technical
users

  3. 3. Outline
•  Challenges
•  IllustraPve
Scenario
•  ExisPng
Work
•  Approach
•  IniPal
EvaluaPon
•  Conclusions
&
Next
Steps
•  Acknowledgements

  4. 4. Outline
•  Challenges
•  IllustraPve
Scenario
•  ExisPng
Work
•  Approach
•  IniPal
EvaluaPon
•  Conclusions
&
Next
Steps
•  Acknowledgements

  5. 5. Challenges
in
Linked
Data
ConsumpPon
1.  CombaPng
informaPon
overload
2.  ExploraPon
starPng
point
3.  Returning
something
useful
4.  Enabling
interacPon
 How
can
we
make
Linked
Data
usable
to
real,
end
users?
•  where
end
users
broadly
classified
into
one
of:
 –  SemanPc
Web
experts
 –  web‐savvy
but
non‐technical

  6. 6. Outline
•  Challenges
•  Illustra.ve
Scenario
•  ExisPng
Work
•  Approach
•  IniPal
EvaluaPon
•  Conclusions
&
Next
Steps
•  Acknowledgements

  7. 7. Sample
dataset
‐
Data.dcs
•  research
groups
in
DCS,
University
of
Sheffield
•  ontologies
(re)used
 –  FOAF,

PRV,
SWRC,
BIB
•  by
Linked
Data
standards
very
small
 –  over
8000
statements
 –  ~3000
(disPnct)
graph
nodes
 –  however
sPll
highlights
the
scale
of
the
challenges
faced

  8. 8. Scenario
•  InformaPon‐seeking
scenario
 –  end
user:
a
primary
school
teacher
 –  task:
looking
for
research
in
local
university
on
‘Web
Technology’
 –  tools
typically
used
for
informaPon
seeking
acPviPes:

 •  web
search/browse

 •  library
•  consider
the
university
department’s
web
site
built
on
top
of
 Data.dcs

  9. 9. Outline
•  Challenges
•  IllustraPve
Scenario
•  Exis.ng
Work
•  Approach
•  IniPal
EvaluaPon
•  Conclusions
&
Next
Steps
•  Acknowledgements

  10. 10. Tools
for
consuming
Linked
Data
•  SemanPc
Web
user
 –  well
catered
for
 –  typical
tasks

 •  browsing
RDF
 •  validaPng
data
and
models
 •  extracPng
data
using
formal
query
syntax
•  mainstream
web
user
 –  lower
tool
support
 –  typical
tasks
‐
exploratory
informaPon
seeking
 •  search
and
query
(using
less
formal
methods,
e.g.,
forms)
 •  browsing
to
discover
informaPon

 •  sharing
of
informaPon
discovered,
results
of
any
analysis

  11. 11. Tools
for
consuming
Linked
Data
 Tool
Type
 Examples
of
Tools
 formaded
text
display
of
RDF
(e.g.,
 Sig.ma,
Marbles,
URI
Burner,
Haystack,
Tabulator
 using
HTML
tables,
templates)
 RDF
graph
model
 W3C
RDF
Validator,
Sindice
Inspector
 other
graph
visualisaPon
 IsaViz,
RDFGravity,
Cytoscape,
RelFinder
 other
domain‐specific
visualisaPon
 DBPedia
Mobile,
Talis
Research
Funding
Explorer
 Expected
Skill
Set
 Examples
of
Tools
 understanding
of
SW
technology
 Sig.ma,
Marbles,
URI
Burner,
W3C
RDF
Validator,
 stack
 RelFinder,
Tabulator
 formal
querying,
e.g.,
SPARQL
 LESS
 basic
to
advanced
knowledge
 DBPedia
Mobile,
RelFinder,
Talis
RFE,
IsaViz
 seeking,
exploratory
navigaPon
 web
browsing
(desktop,
mobile)
 LESS,
DBPedia
Mobile

See also:• Dadzie, A.-S. & Rowe, M. (In press). Approaches to Visualising Linked Data: A Survey, the Semantic Web Journal —Special Call for Survey articles on Semantic Web topics.• Katifori, A., Halatsis, C., Lepouras, G., Vassilakis, C. & Giannopoulou, E. (2007). Ontology visualization methods — asurvey, ACM Computing Surveys.
  12. 12. Outline
•  Challenges
•  IllustraPve
Scenario
•  ExisPng
Work
•  Approach
•  IniPal
EvaluaPon
•  Conclusions
&
Next
Steps
•  Acknowledgements

  13. 13. A
Template‐based
SoluPon
•  taking
advantage
of
self‐describing
RDF
data
 –  look
up
class
of
a
given
resource
 –  load
template
based
on
class
–
i.e.,
rdf:type
of
the
instance
•  focus
on
perPnent
informaPon
in
a
dataset
•  highlight
relaPonships
within
data
•  allow
end
users
to
retrieve
detail
in
ROIs
(regions
of
interest)
•  combine
templates
with
informaPon/knowledge
visualisaPon
 –  hide
the
complexity
of
the
underlying
data
 –  remove
the
need
for
specialist
SW
knowledge
or
skill

  14. 14. Why
this
approach?
•  Templates
 –  value
seen
in
the
wide
use
of
Fresnel
lenses
and
other
template
 development
tools
and
languages,
e.g.,
IsaViz,
LENA,
LESS
 –  simplicity,
reusability,
extensibility,
flexibility
•  VisualisaPon
 –  overview
to
support
detecPon
of
data
structure
 •  exploratory
informaPon
seeking
 •  idenPfying/highlighPng
relaPonships
 •  recognising
anomalies,
errors
 –  reducPon
in
cogniPve
load
–
through
advanced
human
percepPon
 •  especially
useful
for
analysis
of
large,
complex
data

  15. 15. Template
Design
•  idenPfy
key
concepts
&
relevant
metadata
to
define
templates
 –  match
to
standard
ontologies,
e.g.,
FOAF,
PRV,
SWRC,
BIB
 –  SPARQL
queries
–
built
based
on
Fresnel
lens
SPARQL
selectors

•  presentaPon
methods
 –  visual
overview
‐
node‐link
graph
 •  collapse
informaPon
related
to
key
concepts
into
compound
nodes
 •  filter
out
less
immediately
relevant
data
–
provide
more
room
for
ROIs
 –  visual
encoding
 •  colour
coding
based
on
RDF
type
(nodes
and
links)
 •  icons
based
on
RDF
type
(nodes)
 •  size
to
encode
node
properPes,
e.g.,
no.
of
outlinks
 –  detail
view
 •  text
and
thumbnails/icons

  16. 16. Data.dcs
templates
•  informaPon
of
main
interest
–
key
concepts
 –  organisaPonal
structure
‐
research
groups
 –  people
 –  relaPonships
within
structure
•  main
concepts
 –  Organisa.on
[<http://xmlns.com/foaf/0.1/Group>]
 –  Person
 

[<http://xmlns.com/foaf/0.1/Person>]
 –  Publica.on 

[<http://zeitkunst.org/bibtex/0.1/bibtex.owl#Entry>]

  17. 17. Challenge
1:

 Comba.ng
informa.on
overload
•  very
large
amounts
of
distributed,
heterogeneous
data
•  high
inter‐linking

  18. 18. Data.dcs
–
RDF
graph 
Drawn
using
Sindice
Inspector
–
first
1000
triples
only
for
usability
reasons

  19. 19. Data.dcs
‐
RDF
Text
vs
Basic
Graph

  20. 20. Data.dcs
–
Template
Graph
View

  21. 21. Challenge
1:
SoluPon
Proposed
•  Comba.ng
informa.on
overload
 –  very
large
amounts
of
distributed,
heterogeneous
data
 –  high
inter‐linking
•  our
soluPon:
 –  visual
overview
+
filters
 –  highlighPng
key
relaPonships
 –  detail
view

 •  focus
on
graph
ROI
within
context
of
surrounding
data

 
+
text
detail
&
thumbnails
or
representaPve
icons

  22. 22. Challenge
2:

 Explora.on
star.ng
point
•  SW
users
 –  
may
have
a
specific
URI
to
explore
•  mainstream
users
 –  may
or
may
not
have
a
specific
starPng
point
 –  onen
start
with
a
vague
idea
and
browse
to
find
if
there
is
anything
 interesPng

  23. 23. Design
ideas:
Detail
templates

  24. 24. Challenge
2:
SoluPon
proposed
•  Explora.on
star.ng
point
 –  SW
users
 •  
may
have
a
specific
URI
to
explore
 –  mainstream
users
 •  may
or
may
not
have
a
specific
starPng
point
 •  onen
start
with
a
vague
idea
and
browse
to
find
what
they
want
•  our
soluPon:
 –  support
both

 •  where
input,
specific
URI
as
focus
at
start
 •  extract
list
of
potenPal
start
points
 –  selected
from
key
RDF
types
in
a
dataset
 –  randomly
chosen
focus
–
influenced
by
context
 •  centre
graph
on
focus

  25. 25. Challenge
3:

 Returning
something
useful
•  what
is
the
end
user
looking
for?
•  how
can
we
present
the
data
so
they
are
able
to
find
it?

  26. 26. Data.dcs
–
RDF/XML

  27. 27. Co‐ordinated
Template
Views

  28. 28. Co‐ordinated
Template
Views

  29. 29. Under
the
Hood‐
Detail
View
•  e.g.,
SPARQL
query
template
for
the
full
publicaPon
view PREFIX foaf: <http://xmlns.com/foaf/0.1/> PREFIX bib: <http://zeitkunst.org/bibtex/0.1/bibtex.owl#> SELECT DISTINCT ?publicationTitle ?year ?bookTitle ?personUri ?author ? imageUri WHERE { <data.dcs:publicationUri> bib:title ?publicationTitle ; bib:hasYear ?year ; bib:hasBookTitle ?bookTitle ; foaf:maker ?personUri . ?personUri foaf:name ?author ; foaf:img ?imageUri } ORDER BY DESC(?year) ?publicationTitle
  30. 30. Challenge
3:
SoluPon
•  Returning
something
useful
 –  what
is
the
end
user
looking
for?
 –  how
can
we
present
the
data
so
they
are
able
to
find
it?
•  Our
soluPon
 –  graph
overview
+
detail
template
view
 –  reuse
familiar
web
browser
look
and
feel
(detail)
 –  interacPve
graph

 •  to
support
exploratory
navigaPon
 •  retain
context
of
surrounding
informaPon
 •  colour
coding
to
highlight
key
resource
types
and
relaPonships

  31. 31. Challenge
4:

 Enabling
interac.on
•  Our
soluPon
 –  graph
overview
+
detail
template
view
 –  reuse
familiar
web
browsing
interacPon
 –  click
to
navigate
through
data
(in
both
views)
 –  pan
+
zoom
for
graph
view
 –  filters
to
remove
less
relevant
informaPon

  32. 32. Outline
•  Challenges
•  IllustraPve
Scenario
•  ExisPng
Work
•  Approach
•  Ini.al
Evalua.on
•  Conclusions
&
Next
Steps
•  Acknowledgements

  33. 33. FormaPve
EvaluaPon
•  at
ESWC
2010
‘EssenPal
HCI
for
the
SemanPc
Web’
tutorial
 –  focus
group
of
(14)
“expert
reviewers”
 –  assessing
usability
for
both
mainstream
and
expert
users
 –  training
task
and
an
informaPon
exploraPon
exercise

•  graphs
found
to
be
expressive

•  graph
view
effecPve
in
giving
a
sense
of
data
distribuPon
•  detail
view
effecPvely
displayed
key
resources
 in
a
neat
and
 concise
way 
•  prototype
seen
to
have
potenPal
for
exploring
and
debugging
LD
•  some
difficulty
for
users
not
familiar
with
interacPve
graph
layout
 –  eventually
you
got
a
big
picture
of
the
data 
 –  I
liked
the
direct
manipulaPon
but
the
graph
should
stay
put

 [when
I
click] 

  34. 34. Outline
•  Challenges
•  IllustraPve
Scenario
•  ExisPng
Work
•  Approach
•  IniPal
EvaluaPon
•  Conclusions
&
Next
Steps
•  Acknowledgements

  35. 35. Conclusions 
•  explored
human‐centred
soluPon
for
consuming
Linked
Data
 –  exploiPng
templates
(via
SW
technology)
 –  combined
with
visualisaPon
•  evaluaPon
highlighted
challenges
sPll
remaining
‐
among
others:
 –  scale

 –  complexity
•  however
–
promising
start....
•  Next
Steps
 –  more
user
control
‐
(intuiPve)
support
for
defining
templates,
filters
 –  dynamic
update
with
new
Linked
Data
 –  formal
usability
evaluaPon
with
wider
range

of
users

  36. 36. Outline
•  Challenges
•  IllustraPve
Scenario
•  ExisPng
Work
•  Approach
•  IniPal
EvaluaPon
•  Conclusions
&
Next
Steps
•  Acknowledgements
 –  parPcipants

of
ESWC
2010
‘EssenPal
HCI
for
the
SemanPc
Web’
tutorial
 –  Funding
 •  A‐S
Dadzie
–
SmartProducts
&
WeKnowIt
(EU
FP7),
X‐Media
(EU
FP6)
 •  M
Rowe
–
WeGov
(EU
FP7)

 •  D
Petrelli
–
X‐Media
(EU
FP6)

  37. 37. Hide
the
Stack
@
SW
Dogfood


×