Molecules of Knowledge: Self-Organisation in Knowledge-Intensive Environments

Molecules of Knowledge: Self-Organisation in
Knowledge-Intensive Environments

Stefano Mariani, Andrea Omicini
{s.mariani, andrea.omicini}@unibo.it

Dipartimento di Informatica: Scienza e Ingegneria (DISI)
Alma Mater Studiorum—Universit` di Bologna
a

IDC 2012
Intelligent Distributed Computing
Calabria, Italy – 25th of September 2012

Mariani, Omicini (Universit` di Bologna)
a Molecules of Knowledge IDC 2012, 25/9/2012 1 / 35

Outline

1 Motivations

2 The Molecules of Knowledge Model
Informal MoK
Formal MoK

3 A MoK Infrastructure
The TuCSoN Coordination Infrastructure
Mapping MoK over TuCSoN

4 A Case Study: MoK-News

5 Conclusions & Further Works


Motivations

Outline

1 Motivations

Informal MoK
Formal MoK





Motivations

The Challenge

Knowledge-intensive environments (KIE)
KIE present new critical challenges in the knowledge management process:
the ever-increasing amount of information to handle, its heterogeneity in
structure, and the pace at which it is made available are just a few to
mention.

Knowledge workers
For journalists, researchers, lawyers and the like, today ICT systems
provide both new opportunities and new obstacles: ﬁnding all and only the
relevant information they need is a issue that even the most advanced
general-purpose research engines are not able to face today.


Motivations

A Tuple-based Answer I

Adaptive and self-organising systems. . .
. . . seem the only possible answer when the scale of the problem is too
huge, unpredictability too high, global control unrealistic, and
deterministic solutions simply do not work [Omicini and Viroli, 2011].

Coordination models
Tuple-based coordination models and languages have already shown their
eﬀectiveness in the engineering of complex software systems, like
knowledge-intensive, pervasive and self-organising ones
[Omicini and Viroli, 2011].


Motivations

A Tuple-based Answer II

Biochemical tuple spaces
They self-organisation features into tuple-based coordination, by exploiting
the chemical metaphor enhanced with topology aspects
[Viroli and Casadei, 2009]:
→ tuples are seen as chemical reactants, thus equipped with an
activity/pertinency value — resembling chemical concentration a

→ chemical reactions evolve tuples and possibly diﬀuse them to
neighboring chemical compartments
→ tuple spaces act as chemical solutions simulators, that is update
concentrations following the Gillespie algorithm [Gillespie, 1977], host
and execute the chemical reactions and manage the topology-related
features
a
their relative quantity w.r.t. the others


Motivations

Goals

On one hand. . .
. . . to bring the biochemical tuple space abstraction and its self-organising
features to their full realization into knowledge intensive environments, so
to harness their complexity.

On the other hand. . .
. . . to help knowledge workers, by providing them a model in which
knowledge autonomously aggregate in a meaningful and useful way and
eventually diﬀuse to autonomously reach knowledge consumers — rather
than be “searched”.


The Molecules of Knowledge Model

Outline

1 Motivations

Informal MoK
Formal MoK





The Molecules of Knowledge Model

MoK Vision

Molecules of Knowledge (MoK)
MoK is a biochemically-inspired coordination model promoting
self-organisation of knowledge toward the idea of self-organising
workspaces [Omicini, 2011]:
knowledge sources produce atoms of knowledge in biochemical
compartments, which then may diﬀuse and/or aggregate in molecules
by means of biochemical reactions, acting locally within and between
such spaces.
knowledge consumers workspaces are mapped into such
compartments, which reify information-oriented user actions to drive
atoms and molecules aggregation and diﬀusion.


The Molecules of Knowledge Model Informal MoK

MoK Abstractions I

MoK main abstractions
atoms the smallest unit of knowledge in MoK, contain information
from a source and belong to a compartment — thus being
subject to its “laws of nature”
molecules the MoK units for knowledge aggregation, bond together
“somehow-related” atoms
enzymes emitted by MoK catalysts, represent prosumer’s actions and
participate MoK reactions to affect the way in which atoms
and molecules evolve
reactions working at a given rate a , they regulate the evolution of each
MoK compartment, by ruling the way in which molecules
aggregate, are reinforced, diffuse, and decay
a
affected by atoms/molecules concentrations



MoK Abstractions II

MoK other abstractions
compartments the spatial abstraction of MoK, compartments represent
the conceptual loci for all MoK entities as well as for MoK
biochemical processes, also providing MoK with the notions
of locality and neighbourhood
sources each one associated to a compartment, MoK sources are
the origins of knowledge, which is continuously injected at a
certain rate in the form of MoK atoms
catalysts the abstraction for knowledge prosumers, catalysts emit
enzymes in order to attract to him/her relevant knowledge
items



Envisioning MoK Systems

MoK systems
MoK systems should be seen as a network of distributed, shared
information spaces in which some source entities continuously inject
information pieces. These may then aggregate to shape more complex
knowledge chunks and diffuse between different, networked shared spaces.

MoK users
Users can interact with the system through information-oriented actions
over knowledge, which are reified in terms of enzymes by their associated
workspace and exploited to influence knowledge lifecycle.



A MoK System


The Molecules of Knowledge Model Formal MoK

MoK Formal Model I

MoK main abstractions syntax
It is straightforwardly derived from abstractions descriptions:
atoms belong to a source, carry a piece of data, possibly some
metadata and are equipped with their concentration in the
space
atom(src, val, attr)c
molecules are unordered collections of somehow related atoms, again
with a concentration
molecule(Atoms)c
enzymes are strictly coupled to the atom/molecule being accessed,
they too equipped with their concentration
enzyme(Atoms)c



MoK Formal Model II

Reactions semantics
It should comply with the following rewriting rules
aggregation two somehow related atoms/molecules a are joined to spring
a new molecule
molecule(Atoms1 ) + molecule(Atoms2 )
r agg
−→
molecule(Atoms1 Atoms2 ) + Residual(Atoms1 , Atoms2 )
reinforcement atoms/molecules are reinforced by consuming compliant
enzymes from the space
enzyme(Atoms1 ) + molecule(Atoms2 )c
r reinf
−→
molecule(Atoms2 )c+1
a
atoms can be seen as molecules with a single atom inside



MoK Formal Model III

Reactions semantics [continue]
decay molecules naturally fade away during time
r decay
molecule(Atoms)c −→ molecule(Atoms)c−1
diffusion molecules randomly diffuse to somehow defined neighbouring
compartments
Molecules1 molecule1 σi + Molecules2 σ ii
r diffusion
−→
Molecules1 σi + Molecules2 molecule1 σ ii

N.B.: MoK other abstractions and MoK reactions syntax is left unspecified: it
can be tailored to the application domain or legacy infrastructure language at
hand.



MoK Formal Model IV
What is “somehow related/compliant”?
MoK reactions should check reactants correlation to apply, so in order to
define a MoK system one should first of all define a mok function
mok: molecule × molecule → D, which takes two atoms/molecules and
verifies if they are related.

The mok function
The exact definition of mok – that is the mathematical description of
domain D – depends on the application at hand, however will likely
depend on the fields val and attr inside MoK atoms.

N.B. The mok function could range from the simple Linda syntactical matching
– hence D = {true, false} – to more complex semantical fuzzy matching
mechanisms — for which typically D ∈ [0, 1].


A MoK Infrastructure

Outline

1 Motivations

Informal MoK
Formal MoK





A MoK Infrastructure The TuCSoN Coordination Infrastructure

TuCSoN & ReSpecT I

TuCSoN [Omicini and Zambonelli, 1999]
TuCSoN is a Linda-inspired coordination model & infrastructure
providing developers with a distributed, tuple-based middleware exploiting
programmable tuple spaces called tuple centres.

ReSpecT [Omicini, 2006]
The behaviour of such tuple centres can be programmed through the
ReSpecT logic language so to encapsulate any coordination laws directly
into the coordination abstraction.


A MoK Infrastructure The TuCSoN Coordination Infrastructure

TuCSoN & ReSpecT II

MoK requirements
TuCSoN provides the basic ingredients to enable biochemical coordination
topology multiple tuple centres can be deployed in diﬀerent nodes of a
network and/or can coexist in a single node, promoting the
notions of locality and neighbourhood
programmability chemically-inspired evolution of tuples is enabled by
implementing the Gillespie algorithm [Gillespie, 1977] as a
ReSpecT program
matching general purpose MoK reactions can be applied to actual
reactants by using ReSpecT logic tuples and templates to
represent atoms, molecules and enzymes


A MoK Infrastructure Mapping MoK over TuCSoN

TuCSoN for MoK I

MoK {atoms, molecules, enzymes} → ReSpecT logic tuples
Being generated, accessed, moved and consumed both by users and by the
MoK system itself – through reactions –, atoms, molecules and enzymes
should all be implemented as ReSpecT tuples so to be effectively managed
by the TuCSoN middleware.

MoK compartments → TuCSoN tuple centres
By definition, compartments are the locality abstraction in MoK, thus the
mapping with TuCSoN tuple centres is straightforward since they host
both:
the ordinary tuples, that is data chunks — thus atoms, molecules and
enzymes
the specification tuples, that is ReSpecT programs statements —
hence MoK reactions

A MoK Infrastructure Mapping MoK over TuCSoN

TuCSoN for MoK II

MoK reactions −→∗ ReSpecT programs
MoK reactions are simply declarative statements specifying how existing
knowledge should combine, fade away, replicate or move, hence they need
to be interpreted and executed. Furthermore, being chemically-inspired,
this should be done according to Gillespie’s chemical simulation algorithm.

MoK reactions → ReSpecT logic tuples → ReSpecT programs
So actual mapping to TuCSoN is “two-layered”:
MoK reactions are encapsulated into ReSpecT tuples of the kind
law([Inputs], Rate, [Outputs]);
such tuples basically constitute the raw data consumed by the
ReSpecT implementation of Gillespie algorithm — which
continuously, in a chemical-like fashion, schedules and executes them.


A Case Study: MoK-News

Outline

1 Motivations

Informal MoK
Formal MoK






Why News

News management systems
They are a prominent example of
heterogeneity News sources can be virtually anything, from handwritten
notes to printed oﬃcial documents through web published
articles
ubiquity Netbooks, tablets and smartphones pushed information
production, sharing and consumption to be pervasive as
never before
unpredictability News producers are no longer graduated journalists solely,
they include bloggers and whoever has access to the web
though



MoK-News Model

Formally
A generic MoK atom of the form atom(src, val, attr)c becomes a
specialised MoK-News [Mariani and Omicini, 2012] atom of the form

atom(src, val, sem(tag, catalog))c

where
src ::= news source uri
val ::= news content
attr ::= sem(tag, catalog)
tag ::= NewsML tag | NITF tag
catalog ::= NewsCode uri | ontology uri



Envisioning MoK-News Systems I

A MoK-News systems...
...should hence be seen as a self-organising news repository in which
! news pieces – “tag-content” pairs – are injected either automatically
(e.g. using XML parsers) or manually (by journalists) in the form of
MoK-News atoms
! enzymes are released by catalysts (journalists) as manifestations of
their actions over knowledge
! biochemical reactions
aggregate together semantically related atoms — based upon
catalog information
diﬀuse atoms/molecules in neighborhood compartments
reinforce them by using enzymes
decay non-relevant information



Envisioning MoK-News Systems II

“Smart” diﬀusion
It is achieved as a self-organising process caused by the cooperation among
diﬀusion, reinforcement – of relevant knowledge, that is more frequently
accessed – and decay — of useless information, ignored by catalysts.

E.g.
A journalist interested in sports news is more likely to search, read,
annotate – generally, access – sport-related atoms. In the process, he/she
releases enzymes which reinforce accessed atoms/molecules concentration.
In the very end, his/her compartment will mainly store sports-related
knowledge.



Envisioning MoK-News Systems III


Conclusions & Further Works

Outline

1 Motivations

Informal MoK
Formal MoK






Final Remarks

Molecules of Knowledge
The MoK model
→ provides knowledge workers in general with a novel approach both in
thinking and managing knowledge
→ supports their work with self-organising shared workspaces able to
autonomously cluster and spread information

TuCSoN for MoK
The MoK implementation over TuCSoN is a ﬁrst step toward a full
implementation of the MoK model, featuring topology-related aspects,
self-aggregation of information and smart diﬀusion toward interested users.



Open Issues & Further Developments

How to. . .
? push the MoK model toward the idea of self-organising workspace
[Omicini, 2011], fully supporting adaptiveness of compartments rather
than information solely?
? effectively implement efficient semantic matching mechanisms
[Nardini et al., 2012] to lift Linda purely syntactical one currently
exploited in TuCSoN?

Further works
! explore the literature to address first issue
! improve the current prototypal implementation of the MoK model
upon TuCSoN coordination infrastructure
! test MoK against other application domains — e.g. research
publications, sensor networks, social media, healthcare...

Thanks

Thanks to. . .

. . . everybody here for listening
. . . the Sapere team for supporting our work 1

1
This work has been supported by the EU-FP7-FET Proactive project Sapere
Self-aware Pervasive Service Ecosystems, under contract no.256873.

Bibliography

Bibliography I

Gillespie, D. T. (1977).
Exact stochastic simulation of coupled chemical reactions.
The Journal of Physical Chemistry, 81(25):2340–2361.
Mariani, S. and Omicini, A. (2012).
Self-organising news management: The Molecules of Knowledge approach.
In Fernandez-Marquez, J. L., Montagna, S., Omicini, A., and Zambonelli, F., editors, 1st
International Workshop on Adaptive Service Ecosystems: Natural and Socially Inspired
Solutions (ASENSIS 2012), pages 11–16, SASO 2012, Lyon, France.
Pre-proceedings.
Nardini, E., Omicini, A., and Viroli, M. (2012).
Semantic tuple centres.
Science of Computer Programming.
Special Issue on Self-Organizing Coordination.
Omicini, A. (2006).
Formal ReSpecT in the A&A perspective.
In Canal, C. and Viroli, M., editors, 5th International Workshop on Foundations of
Coordination Languages and Software Architectures (FOCLASA’06), pages 93–115,
CONCUR 2006, Bonn, Germany. University of M´laga, Spain.
a
Proceedings.


Bibliography

Bibliography II

Omicini, A. (2011).
Self-organising knowledge-intensive workspaces.
In Ferscha, A., editor, Pervasive Adaptation. The Next Generation Pervasive Computing
Research Agenda, chapter VII: Human-Centric Adaptation, pages 71–72. Institute for
Pervasive Computing, Johannes Kepler University Linz, Austria.
Omicini, A. and Viroli, M. (2011).
Coordination models and languages: From parallel computing to self-organisation.
The Knowledge Engineering Review, 26(1).
Special Issue for the 25th Years of the Knowledge Engineering Review.
Omicini, A. and Zambonelli, F. (1999).
Coordination for Internet application development.
Autonomous Agents and Multi-Agent Systems, 2(3):251–269.
Special Issue: Coordination Mechanisms for Web Agents.
Viroli, M. and Casadei, M. (2009).
Biochemical tuple spaces for self-organising coordination.
In Field, J. and Vasconcelos, V. T., editors, Coordination Languages and Models, volume
5521 of LNCS, pages 143–162. Springer, Lisbon, Portugal.


Molecules of Knowledge: Self-Organisation in
Knowledge-Intensive Environments

Stefano Mariani, Andrea Omicini
{s.mariani, andrea.omicini}@unibo.it

Dipartimento di Informatica: Scienza e Ingegneria (DISI)
Alma Mater Studiorum—Universit` di Bologna
a

IDC 2012
Intelligent Distributed Computing
Calabria, Italy – 25th of September 2012


Molecules of Knowledge: Self-Organisation in Knowledge-Intensive Environments

More Related Content

Similar to Molecules of Knowledge: Self-Organisation in Knowledge-Intensive Environments (20)

More from Andrea Omicini (20)

Recently uploaded (20)

Molecules of Knowledge: Self-Organisation in Knowledge-Intensive Environments