This document discusses connecting and synchronizing scientific knowledge through an associationist model. It proposes a process model for tracking the evolution of categories in scientific knowledge over time. An example tool called AdvoCate is described that would allow researchers to model changes to categories and maintain connections between changing categories and tools like databases and ontologies. Future work includes formalizing the data model, developing change recognition rules for different category models, visualizing category evolution, and integrating with ontology and database evolution tools.
3. The
current
state
of
scienBfic
pracBces
How
well
are
we
carrying
forward
the
core
principles
of
science
(communicaBon,
repeatability
and
refutability)
with
the
new
scienBfic
pracBces?
6. Learning
from
the
past
Map
Categories
How
do
we
connect
them
back
to
synthesize
an
integrated
view
?
7. Knowledge
producer
and
consumer
perspecBve
Data Methods
Analysis
Map
Workflow
Data model
Knowledge
producer
8. Knowledge
producer
and
consumer
perspecBve
Data Methods
Analysis
Map
Workflow
Data model
Knowledge
producer
It’s
very
confusing.
They
are
all
disconnected.
Hard
to
say
how
they
were
used.
I
wish
they
had
some
explicit
connecBons.
Knowledge
consumer
9. FragmentaBon
of
scienBfic
arBfacts
and
processes
among
communiBes
Community
2
Image
processing
tools
CSV/XML/database
Image
processing
DigiBzed
data
Community
1
Remote
sensing
system
Satellite
Imagery
Data
observaBon
Concepts
Database
Community
4
ClassificaBon
Machine
learning
tool
Land-‐cover
dataset
Web-‐mapping
tool
Taxonomy
tool
Community
5
Land-‐cover
map
Taxonomy
ApplicaBons
Community
3
Aerial
Field
work
photography
Training
data
CSV/XML/
database
10. FragmentaBon
of
scienBfic
arBfacts
and
processes
among
communiBes
Community
2
Image
processing
tools
Richer/beXer
data
CSV/XML/database
Image
processing
DigiBzed
data
New
validaBon
techniques
Conceptual
change
Algorithmic
improvement
New
ideas
Community
1
Remote
sensing
system
Satellite
Imagery
Data
observaBon
Concepts
Database
Community
4
ClassificaBon
Machine
learning
tool
Land-‐cover
dataset
Web-‐mapping
tool
Taxonomy
tool
Community
5
Land-‐cover
map
Taxonomy
ApplicaBons
Community
3
Aerial
Field
work
photography
Training
data
CSV/XML/
database
New/beXer
technology
13. Other
knowledge
integraBon
models..
• Research
Objects
S.
Bechhofer,
D.
De
Roure,
M.
Gamble,
C.
Goble,
and
I.
Buchan,
“Research
objects:
Towards
exchange
and
reuse
of
digital
knowledge,”
presented
at
The
Future
of
the
Web
for
CollaboraBve
Science,
NC,
USA,
2010.
• Reproducible
Research
System
J.
P.
Mesirov,
“Accessible
reproducible
research,”
Science,
Jan.
2010.
• Linked
Science
T.
Kauppinen
and
G.
M.
Espindola,
“Linked
open
science
communicaBng,
sharing
and
evaluaBng
data,
methods
and
results
for
executable
papers,”
presented
at
the
Int.
Conf.
ComputaBonal
Science
(ICCS),
2011.
•
Workflows
14. Other
knowledge
integraBon
models..
• Research
What
Objects
are
the
shortcomings?
S.
Bechhofer,
D.
De
Roure,
M.
Gamble,
C.
Goble,
and
I.
Buchan,
“Research
objects:
Towards
exchange
• and
reuse
Focus
of
digital
knowledge,”
presented
at
The
Future
of
the
Web
for
CollaboraBve
Science,
NC,
USA,
2010.
• Reproducible
Research
System
J.
P.
Mesirov,
“Accessible
reproducible
research,”
Science,
Jan.
2010.
• Linked
Science
T.
Kauppinen
and
G.
M.
Espindola,
“Linked
open
science
communicaBng,
sharing
and
evaluaBng
data,
methods
and
results
for
executable
papers,”
presented
at
the
Int.
Conf.
ComputaBonal
Science
(ICCS),
2011.
•
Workflows
on
a
single
experiment
of
science,
rather
than
science
as
an
ongoing
and
evolving
process
• Provide
a
linear
view
of
science,
but
science
is
instead
exploratory,
dynamic
and
cyclic
• Focus
typically
on
data
and
not
on
conceptual
structures
16. AssociaBonist
A
model
that
supports
living
and
linked
scienBfic
knowledge
view
Organic
view
–
born,
evolve
and
die
17. ConnecBng
scienBfic
arBfacts
Database
Data
schema
Includes
e-‐
Science
tools
and
process
models
Sogware
tools
Categories
Map
Ontology
Live
connecBons
among
scienBfic
arBfacts
18. Example
1. If
a
new
classifier
method
is
used
for
land
cover
classificaBon,
it
may
lead
changes
to
categories
2. The
extension
of
the
category
‘Forest’
changes,
leading
to
change
in
the
data
stored
under
that
category.
3. Finally,
the
change
in
data
is
reflected
in
the
land
cover
map
Database
Data
schema
Sogware
tools
Categories
Map
Ontology
1
2
3
19. Adventures
of
Categories
(AdvoCate)
• An
e-‐Science
tool
that
incorporates
the
process
model
of
category
evoluBon
• The
system
allows
researchers
to
model
changes
in
categories,
captures
the
process
of
evoluBon
and
maintains
a
category-‐versioning
system
• Connect
changes
in
categories
with
the
tools
supporBng
database
and
ontology
evoluBon
20. Process
model
of
category
evoluBon
External
change
drivers
Revising
categorical
model
EvaluaBon
of
categorical
model
Change
approval
Change
report
(using
elementary
and
complex
change
operaBons)
ImplemenBng
the
changes
&
updaBng
category
versioning
system
Change
PropagaBon
21. Process
model
of
category
evoluBon
• New
observaBon
(training
External
data)
• Societal
drivers
change
drivers
Revising
categorical
model
EvaluaBon
of
categorical
model
Change
approval
Change
report
(using
elementary
and
complex
change
operaBons)
• New
understanding
ImplemenBng
the
changes
&
updaBng
category
versioning
system
Change
PropagaBon
• New
category
• Splikng
or
merging
of
categories
• Drig
in
categories
• Elementary
changes:
• Add/Delete
category
• Add/Delete
relaBonship
• Change
label
• Change
intension
• Composite
changes:
• Born
• Die
• Merge
• Split
• Drig
22. Process
model
of
category
evoluBon
External
change
drivers
Process
of
science
Revising
categorical
model
EvaluaBon
of
categorical
model
Change
approval
Change
report
(using
elementary
and
complex
change
operaBons)
ImplemenBng
the
changes
&
updaBng
category
versioning
system
Change
PropagaBon
Ontology/database
evoluBon
tools
24. An
example
of
category
evoluBon
from
land
cover
mapping
25. An
example
of
category
evoluBon
from
land
cover
mapping
26. An
example
of
category
evoluBon
from
land
cover
mapping
27. Future
work
• Formalize
data
model
using
semanBc
technologies
• Change
recogniBon
rules
for
various
categories
models
(probability
distribuBon
models,
rule-‐
based
models,
etc.)
• VisualizaBon
of
categories
evoluBon
• Change
broadcasBng
service
to
ontology
and
database
evoluBon
tools