Abstract. In order use our personal data within our day to day activities, we need to manage it in a way that is easy to consume, which currently is not an easy task. People have found their own ways to organize their personal data, such as categorizing files in folders, labeling emails etc. This is acceptable to a certain degree, since we have to deal with have some (human) difficulties such as our limited capacity of categorization and our incapacity of maintaining highly structured artifacts for long periods of time. We believe that to organize this great amount of personal data, we need the help of our communities. In this work, we apply the emergent semantics field to personal data management, aiming to decrease our cognitive efforts spent in simple tasks, handling semantic evolution in conjunction with our close peers.
Social Emergent Semantics for Personal Data Management
1. Social Emergent Semantics
for Personal Data Management
Cristian Vasquez ( cvasquez[at]vub.ac.be )
Semantics Technology and Applications Research Lab
Vrije Universiteit Brussel
2. Agenda:
β Motivation
β Personal Data management
β Use case
β Shared Ontology Views
β Blackboard anatomy
β Experiment dynamics
β Summary
3. Motivation
Use case
Let's suppose....
β¦ that in a far away country...
A bar that is frequently visited by
sailors...
And they exchange experiences...
4. Motivation
Use case
These sailors enjoy talking about:
Practical things: And not so practical things:
β Geographical information
β Histories about their trips
β Journey advice
β Gossip
β Weather
β Big sea monsters
β hazards...
β Phantom ships
β Mermaids...
5. Motivation
Use case
These sailors would like to share information
such as
β Maps
β Drawings
β Travel logs etc
Which are useful to their community of sailors.
6. Motivation
Let's suppose that....
They count with:
Advanced technological devices,
And they use them to record and
store movies,photographs, sound,
geographical information etc.
On all their journeys.
7. Motivation
These sailors would like to share information with other sailors.
The problem:
β Every sailor has its own way of organizing its
information
β It's already difficult for them to find their own
information... since the volume is huge.
β Data is not well structured
8. Motivation
Current solutions:
To 'Attach' pieces of information (structured or not) to other pieces
of information, in order to find and manage them. 'Metadata'
Measurements,
Models, (I.e: 'taxonomies') (I.e: 'coordinates')
Written symbols
(I.e: 'tags')
9. Motivation
Sharing information is easier with the help of:
β Structured meta-data
β Artifacts that reflect our agreements (ontologies)
β But to come up with agreements, is
already a difficult task.
10. Motivation
Sharing information is easier with the help of:
β Structured meta-data
β Artifacts that reflect our agreements (ontologies)
β But to come up with agreements, is
already a difficult task.
β Example:
β Tree of our sailors want to share the
pictures and position of the mermaids
that they have seen
β Sailor 1 (Greek) Mermaids appear in the folklore of many
cultures including east, europe, china
β Sailor 2 (British isles) and india, they are usually considered
dangerous, and are associated with
β Sailor 3 (Slavic) floods storms, shiprecks and drownings.
However in other folk traditions, they can
be benevolent and can fall in love with
humans
11. Motivation
Sailor 1 (greek):
- These creatures are called 'seirines'
- They live in the sea
- They are woman
- They have beautiful and long hair.
- They have enchanting voices
12. Motivation
Sailor 1 (greek):
- These creatures are called 'seirines'
- They live in the sea
- They are woman
- They have beautiful and long hair.
- They have enchanting voices
Sailor 2 (british isles):
- These creatures are called 'mermaids'
- They live in the sea
- They are woman
- They can be giant
- They don't have inmortal souls
13. Motivation
Sailor 1 (greek):
- These creatures are called 'seirines'
- They live in the sea
- They are woman
- They have beautiful and long hair.
- They have enchanting voices
Sailor 2 (british isles):
- These creatures are called 'mermaids'
- They live in the sea
- They are woman
- They can be giant
- They don't have inmortal souls
For sailor 1 & 2, is
direct to share
artifacts about woman
that live in the sea...
14. Motivation
Sailor 1 (greek):
- These creatures are called 'seirines'
- They live in the sea
- They are woman
- They have beautiful and long hair.
- They have enchanting voices
Sailor 2 (british isles):
- These creatures are called 'mermaids'
- They live in the sea
- They are woman
- They can be giant
- They don't have inmortal souls
Sailor 3 (Slavic):
- These creatures are called 'Rusalkas'
- They live in the sea
- They are woman
- They do not have a fish-like tail
- They are beautiful young women with long green hair
15. Motivation
Sailor 1 (greek):
- These creatures are called 'seirines'
- They live in the sea
- They are woman
- They have beautiful and long hair.
- They have enchanting voices
- They DO have a fish-like tail
Sailor 2 (british isles):
- These creatures are called 'mermaids'
- They live in the sea
- They are woman
- They can be giant
- They don't have inmortal souls
- They DO have a fish-like tail
Sailor 3 (Slavic):
- These creatures are called 'Rusalkas' Sailors learn gradually from the
- They live in the sea conceptualizations of others.....
- They are woman
- They do not have a fish-like tail
- They are beautiful young women with long green hair
16. Motivation
β To make agreements can be easier for some domains than
for others.
β Example: can be easy for these sailors to agree about:
β System of coordinates for the islands.
β Weather conditions (distinct types of weather).
β Price of a good.
β But it may be difficult to come up with agreements about
personal (custom) data.
Example: How we can store, classify and
annotate digital data about?
β Sailor 1 (Greek) 'seirines'
β Sailor 2 (British isles) 'Mermaids'
β Sailor 3 (Slavic):'Rusalkas'
In order to share it?
17. Proposal:
Blackboard networks
β Users interact through multiple 'canvas' or 'blackboards', in order to build 'semantic bridges'
β These networks are constructed incrementally, and organically.
β Network objective: To build and represent local agreements, collaboratively.
Application
seirines
is a
Mermaid
Part of
Tail
Application
18. State of art
β Essential components:
β Semantic desktop (I.e [1] Nepomuk Framework)
β Personal Information Model (PIMO) a local
'ontology' to annotate our personal data.
[1] http://nepomuk.semanticdesktop.org/nepomuk/
19. State of art
β How to elicit custom ontologies?
β Ontology views
An ontology view is not just a portion of a complete
ontology. Rather is a collection of concepts and
relationships that allows a unique representation by
some participants of a certain domain. In the same way
as ontologies, ontology views may be described using
metadata representation languages such as RDF, RDFs
and OWL among others. They evolve using change
operators that allow coherent ontology view mutations.Mermaids
Ontology
Variant
Sailor 1 seirines Sailor 2
Ontology
Variant
Shared
Ontology Service
Elizabeth Chang, Tharam S Dillon, and Ling Feng. Modeling Ontology
Views : An Abstract View Model for Semantic Web. Proceedings of
the First International IFIP/WG12.5Working Conference on Industrial
Applications for Semantic Web (IASW), pages 227β246, 2005.
Example of elicitation of local ontology
20. Ontology views in the Web:
We want to describe our referents, to
Be used by computers
β Structured descriptions
β Identified referents (observed subjects)
Conceptualization
Thought + Observer
βThese 3 components cannot
be separated!
Referent (observed subject)
Symbols
'seirines'
21. Ontology views in the Web + personal dataspaces
Sailor 1(british) Shared Sailor 2(greek)
Entity
URI
Ontology View
Mermaids Ie: rdf schema seirines
(terminology) (terminology)
Sailor 2's
Sailor 1's Perspective personal
dataspace
22. Ontology views in the Web + personal dataspaces.
How to manage them?
Research proposal: Web blackboards
Blackboards can be seen as extensions of a semantic
wiki web page, where participants collaboratively
describe a subject using distinct description mechanisms
and formalisms. A participant is allowed to subscribe to
multiple blackboards, contributing content in order to
converge into acceptable conceptualizations. The
blackboards collected by an user constitute a network
what he can bind directly with his own Personal data
(extending his Personal Information Model)
23. Anatomy of a blackboard
Blackboard as a playground
β Multiple of observers
β Multiple representation layers
Conceptualization Conceptualization
Thought + Observer A Web Observer B + Thought
Blackboard
(Public space)
Observers
Blackboard's
Metadata
Referent (observed subject)
Symbols Symbols
Representation
Layer
24. Anatomy of a blackboard
Multi Layer Blackboard variant Example
Conceptualization Conceptualization
Thought + Observer A Web Observer B + Thought
Blackboard
(Public space)
Observers
Blackboard's
Language Metadata RDF
(practical)
Referent (observed subject)
Measures
(empirical) Semantic layer Controlled
Vocabulary
Empirical layer
Models Natural
(ontology) Language
Pragmatical layer
Observer B
private space
25. Anatomy of a blackboard
Multi Layer Blackboard variant Example
Conceptualization Conceptualization
Thought + Observer A Web Observer B + Thought
Blackboard
(Public space)
Observers
Blackboard's
Language Metadata RDF
(practical)
Referent (observed subject)
Measures
(empirical) Semantic layer Controlled
Vocabulary
Empirical layer
Models Natural
(ontology) Language
Pragmatical layer
Observer B
private space
Written symbols (I.e: 'tags')
26. Anatomy of a blackboard
Multi Layer Blackboard variant Example
Conceptualization Conceptualization
Thought + Observer A Web Observer B + Thought
Blackboard
(Public space)
Observers
Blackboard's
Language Metadata RDF
(practical)
Referent (observed subject)
Measures
(empirical) Semantic layer Controlled
Vocabulary
Empirical layer
Models Natural
(ontology) Language
Pragmatical layer
Observer B
private space
Measurements,
(I.e: 'coordinates')
27. Anatomy of a blackboard
Multi Layer Blackboard variant Example
Conceptualization Conceptualization
Thought + Observer A Web Observer B + Thought
Blackboard
(Public space)
Observers
Blackboard's
Language Metadata RDF
(practical)
Referent (observed subject)
Measures
(empirical) Semantic layer Controlled
Vocabulary
Empirical layer
Models Natural
(ontology) Language
Pragmatical layer
Observer B
private space
Models, (I.e: 'taxonomies')
28. Anatomy of a blackboard
Multi Layer Blackboard variant Example
Conceptualization Conceptualization
Thought + Observer A Web Observer B + Thought
Blackboard
(Public space)
Observers
Blackboard's
Language Metadata RDF
(practical)
Referent (observed subject)
Measures
(empirical) Semantic layer Controlled
Vocabulary
Empirical layer
Models Natural
(ontology) Language
Pragmatical layer
Observer B
private space
29. Anatomy of a blackboard
Blackboards as a network
β Relations to other blackboards (links)
β Wiki paradigm variant
Conceptualization
Thought + Observer A Web Web
Blackboard Blackboard
(Public space) (Public space)
Observers Observers
Blackboard's
Is related to Blackboard's
Metadata Metadata
Referent (observed subject) Referent (observed subject)
Symbols
Representation Representation
Layer Layer
β Sailor 1 (Greek) 'seirines' β Sailor 3 (Slavic):'Rusalkas'
β Sailor 2 (British isles) 'Mermaids'
'With tail' 'Without tail'
30. Anatomy of a blackboard
Blackboard networks
β’ Users can relate blackboards using relationships such as causality, location
function etc. forming a network. Pattern analysis is used then to provide
feedback to the communities, increasing their awareness.
β’ During the interplay within a blackboard, there will be cases where some
participants disagree with others regarding some representation. Thus
agreement mechanisms can be used in order to reach convergence.
β’ If the distinct participant's views become irreconcilable, then the blackboard
itself may diverge into distinct variants, intended to capture distinct semantics.
31. Anatomy of a blackboard
Blackboard networks
β user constructs a perspective via selecting distinct blackboard variants
β are decentralized
β are constructed incrementally in an organic way (emerging)
Application
seirines
is a
Mermaid
Part of
Tail
Application
32. Anatomy of a blackboard
Blackboard networks
β Since one user only have a partial view of the blackboard network,
β We need mechanisms to promote awareness
β One possibility is pattern recognition
Application
βIs-aβ
relationship
cycle
is a
is a
is a
Application
33. Anatomy of a blackboard
Blackboard networks
β Since one user only have a partial view of the blackboard network,
β We need mechanisms to promote awareness
β One possibility is pattern recognition
Application
Part of
Part of
Part of
Application
βpart ofβ
Relationship
pattern
34. Anatomy of a blackboard
Application: An user augments their own Personal
Information Model Ontology (PIMO) by means of
binding their own concepts to the subjects
described within the blackboards
User context
Application
Application
35. Anatomy of a blackboard
User context
Application
Application: An user links elements
from Linked Open Data to their own
view of blackboards, creating
'bridges' to query for example using
local terminology.
Application
LOD cloud
38. Blackboard dynamics
Snapshot based versioning
All the layers are versioned together forming a
snapshot that is identified as a whole (With an URI).
Web V1 V2 V3 Snapshot
Blackboard
(Public space)
Observers O O O O
1 1 2 2
Blackboard's M M M M
Metadata 1 2 3 4
Referent (observed subject)
Semantic layer S S S S
0 1 1 2
Empirical layer E E E E
0 0 1 1
Pragmatical layer P P P P
1 1 1 1
39. Blackboard dynamics
Snapshot based versioning
Users interacts selecting some
blackboards and pulling them to
their local spaces, where they can
augment or use the blackboards. Sailor's Local
Sailor's
Staging Blackboard
if they make contributions then Working space
area clone
they have to push them through
multiple stages.
Web V1 V2 V3 Snapshot
Blackboard
(Public space)
Observers O O O O
1 1 2 2
Blackboard's M M M M
Metadata 1 2 3 4
Referent (observed subject)
Semantic layer S S S S
0 1 1 2
Empirical layer E E E E
0 0 1 1
Pragmatical layer P P P P
1 1 1 1
40. Blackboard dynamics
Snapshot based versioning
Expect
consistency &
Users interacts selecting some A draft space or some degree of
blackboards and pulling them to playground with agreement the
no constraints local community
their local spaces, where they can
augment or use the blackboards. Sailor's Local
Sailor's
Staging Blackboard
if they make contributions then Working space
area clone
they have to push them through
multiple stages.
Web V1 V2 V3 Snapshot
Blackboard
(Public space)
Observers O O O O
1 1 2 2
Blackboard's M M M M
Metadata 1 2 3 4
Referent (observed subject)
Semantic layer S S S S
0 1 1 2
Empirical layer E E E E
0 0 1 1
Pragmatical layer P P P P
1 1 1 1
41. Blackboard dynamics
Example:
β Managing inconsistency
Mermaid
Web Blackboard
β Sailor 1 (Greek) 'Seirines'
Observers β Sailor 2 (British isles) 'Mermaids'
Blackboard's
Metadata
Referent (observed subject)
Semantic layer
β They live in the sea
Empirical layer
β They are woman
Pragmatical layer
42. Blackboard dynamics
Example:
β Managing inconsistency
Mermaid β Sailor 1 (Greek) 'Seirines'
Web Blackboard β Sailor 2 (British isles) 'Mermaids'
β Sailor 3 (Slavic):'Rusalkas'
Observers
- They DO have a fish-like tail
Blackboard's
Metadata Variant
B
Referent (observed subject) 0
Semantic layer
Empirical layer Variant
A
0
Pragmatical layer
- They do NOT have a fish-like tail
43. Blackboard dynamics
Example:
β Managing inconsistency
Mermaid β Sailor 1 (Greek) 'Seirines'
Web Blackboard β Sailor 2 (British isles) 'Mermaids'
β Sailor 3 (Slavic):'Rusalkas'
Observers
Blackboard's
Metadata Variant Why divergence is useful?
B
Referent (observed subject) 0 β Irreconcilable world views
β Practical reasons
Semantic layer β (I.e distinct degrees of
complexity needed)
Empirical layer Variant
A
0
Pragmatical layer
Sometimes we don't want global Interoperability
Our scope is our community.
44. Blackboard dynamics
Root
Web
Blackboard
Variants mutate
Observers independently
Blackboard's
Metadata Variant Variant
B B
Referent (observed subject) 0 1
Semantic layer
Empirical layer Variant Variant
A A
0 0
Pragmatical layer
45. Blackboard dynamics
Root
Web
Blackboard
Observers
Blackboard's
Metadata Variant Variant
Variant
B B B
Referent (observed subject) 0 1 1
Semantic layer
Empirical layer Variant Variant Variant
A A A
0 0 1
Pragmatical layer
46. Blackboard dynamics
Convergence
example:
Root
Web
β Sailor 1 (Greek) 'Seirines'
Blackboard β Sailor 2 (British isles) 'Mermaids'
β Sailor 3 (Slavic):'Rusalkas'
Observers
Blackboard's
Metadata Variant Variant Variant
Variant
B B B B
Referent (observed subject) 0 1 1 1
Semantic layer
Variant - MAY have a fish-like tail
C
0
Empirical layer Variant Variant Variant Variant
A A A A
0 0 1 1
Pragmatical layer
How can we support convergence?
B3
With computer aided support:
β
B4 B2 β I.E: Relationship pattern recognition
β 'Seirines' & 'Mermaids' very similar to 'Rusalkas'
B1 β suggest MAY have a fish-like tail
47. Blackboard dynamics
Service layer
example:
Why versioning and
Root convergence is useful?
Web
Blackboard
βIts easier to construct and
Observers
maintain services
Blackboard's
Metadata Variant Variant Variant
Variant
B B B B
Referent (observed subject) 0 1 1 1
Variant
Semantic layer C
0
Empirical layer Variant Variant Variant Variant
A A A A
0 0 1 1
Pragmatical layer
Service layer Services Services
0
1
50. The experiment
Nepomuk Framework to
β Local metadata-extraction
β PIMO management
Semantic media Wiki + iMapping
β Blackboard description interface
(This is under evaluation)
51. The experiment
Nepomuk Framework to
β Local metadata-extraction
β PIMO management
JGIT Semantic media Wiki + iMapping
β Dataspace versioning
β Blackboard description interface
βConvergence and divergence capability (This is under evaluation)
52. The experiment
RDF as representation model
β Fundamental 'glue' to put all the pieces together
β Straightforward possibility to use the Web as publishing and distribution mechanism.
53. Summary
β’ This framework explores notions such as personal context and
emergent semantics, making use of artifacts such as blackboards that can
diverge and converge in order to support meaning evolution, in order to
improve our personal data management capabilities.
β’ In this work we don't aim to distill global semantics. Instead we want
our own semantics, taking as hypothesis that they are incrementally
constructed by our close communities.