1. Agents and Social Networks
Environments for Digital
Preservation
Prof. Peplluis de la Rosa
Albert Trias
EASY INNOVA @ UdG
Nov 2007 – Oct 2010
10/05/2013 1
2. PROTAGE Consortium
PReservation Organizations using Tools in AGent
Environments
National Archives of Sweden
Luleå University of Technology (Sweden)
National Archives of Estonia
Fraunhofer Gesellschaft zur Förderung
der angewandten Forschung e.V. (Germany)
University of Bradford (U. K.)
EASY Innova @ UdG (Spain)
Giunti Labs S.r.l. (Italy)
6. Digital data paradigm shift
180 Exabytes
1600 Exabytes
9X Growth
95%
Unstructured
70% Created by Individuals Enterprises responsible for 85% of this new data
(Security, privacy, reliability, compliance)
The Digital Data Paradigm Shift
2011 New Digital Data
(25% Created, 75% Replicated)
2007 New Digital Data
(Created, Captured, Replicated)
7. 10/05/2013 7
67% of DP expert users % think that the
Digital Preservation solutions they have
today are not good enough, or are
insufficient (43%) or scarce (23%). So
they look for new DP solutions
8. DP is social
Fragmented DP
Knowledge
• 67% of expert users look
for solutions provided by
other institutions.
• 90% of them consult
trusted colleagues.
• 83% consult final users
Frequent Knowledge
Exchanges
• 83% of expert users
share their knowledge
• 77% search through the
web for solutions
• 60% visit DP web sites
• 20% contribute to the
web sites
10/05/2013 8
community of DP experts are in favour of DP knowledge exchanges with
colleagues and other institutions, and they are equally happy to do it with
individual users
9. DP Price
• Distribution chanel of DP: there is slight preference for
being bundled with storage services (77%) rather than
being bundled with antivirus tools (63%). The key issue
is that PROTAGE will attract DP knowledge exchange
among expert users through Internet though it will also
distribute the DP knowledge and solutions to individual
users via storage software and antivirus tool bundles.
The expert users (64%) claim they would accept to pay
10% of the price of the storage service, being the
estimated price of 15% of the storage service and 13%
of the antivirus service.
10/05/2013 9
10. DP Opportunities
• DP Consultancy
• Indexing: deep web
• Social search
Peer DP services like
• ”keeping your copies
as you keep mine”
• Crossed services
before preservation
(i.e., enhancing
content)
10/05/2013 10
Users show a slight preference for DP being bundled with storage
services (77%) rather than being bundled with antivirus tools (63%).
Expert users (64%) claim they would accept to pay 10% of the price
of the storage service, being the estimated price of 15% of the
storage service and 13% of the antivirus service.
11. PROTAGE–intelligent agents
• Social search is
implemented
• Experts and final
users share DP
solutions
• Lists of trust guide the
social search
It might seem the
Facebook of DP
• Agents automate
the social search for
DP solutions in terms
of actions plans and
migration support
• Agents proactively
schedule the
preservation tasks
Agents are a type of peer
services
10/05/2013 11
12. First time a
DP effort that
targets not
only memory
institutions
The prototype =
IT Innovation
Agent technology can
be applied to DP
Agent technology can
simplify DP for many
users and groups …
PROTAGE = Intelligent Agents
WHY AGENTS?
1210/05/2013
13. PROTAGE = Intelligent Agents
WHY AGENTS?
13
•Agents:
• Are reactive: react after receiving a question.
• Are social: capability to communicate to
others.
• Are proactive: take initiative.
• Are autonomous: ability to work
independently.
• Agents enable the construction of information
systems from multiple heterogeneous sources
[Dignum 2005]
10/05/2013
14. How Agents do Social Search?
14
When a User is searching for a DP Plan, the query is sent to her Searcher Agent.
The Searcher Agent:
• Search in the local Knowledge Base
• Search in Institution Repositories Knowledge Base
• Asks friends’ agents (send and forward message)
• Filter Results:
• Only DP Plans well rated by friends
• Only DP which are owned by users of with some features.
• Hybrid
• Can show the question to its user.
• The effort of a search depends on the trust it has to the sender.
10/05/2013
15. Search Example
15
ActionPlan Eloy Albert Alex Access-
Point
Image conversion (3versions) X X X
Tech medatada extraction X X X
Local AV Check X X
Local AV Clean X X
Remote AV Check X X
Remote AV Clean X X
Generic Image Conversion (certified) X
Calc MD5 X
1. Alex searches the keyword “AV”.
in the “Application local DB” No result is found.
Alex’s SA Eloy’s SA Albert’s SA
AV
Local DB
10/05/2013
16. Search Example
15
ActionPlan Eloy Albert Alex Access-
Point
Image conversion (3versions) X X X
Tech medatada extraction X X X
Local AV Check X X
Local AV Clean X X
Remote AV Check X X
Remote AV Clean X X
Generic Image Conversion (certified) X
Calc MD5 X
1. Alex searches the keyword “AV” in the
“Application local DB” No result is found.
2. Alex’s Agent searches for a certified actionplan
into the Access Point (institutional point of view,
with certified plans). No result is found again.
Alex’s SA Eloy’s SA Albert’s SA
AV
Access Point DB
10/05/2013
17. Search Example
15
ActionPlan Eloy Albert Alex Access-
Point
Image conversion (3versions) X X X
Tech medatada extraction X X X
Local AV Check X X
Local AV Clean X X
Remote AV Check X X
Remote AV Clean X X
Generic Image Conversion (certified) X
Calc MD5 X
1. Alex searches the keyword “AV” in the
“Application local DB” No result is found.
2. Alex’s Agent searches for a certified actionplan
into the Access Point (institutional point of
view, with certified plans). No result is found
again.
3. The Agent asks its friend Eloy (distance = 1).
Alex’s SA Eloy’s SA Albert’s SA
AV
10/05/2013
18. Search Example
15
ActionPlan Eloy Albert Alex Access-
Point
Image conversion (3versions) X X X
Tech medatada extraction X X X
Local AV Check X X
Local AV Clean X X
Remote AV Check X X
Remote AV Clean X X
Generic Image Conversion (certified) X
Calc MD5 X
1. Alex searches the keyword “AV” in the
“Application local DB” No result is found.
2. Alex’s Agent searches for a certified actionplan
into the Access Point (institutional point of view,
with certified plans). No result is found again.
3. The Agent asks its friend Eloy (distance = 1).
4. Eloy’s agent checks whether 4 matching plans
are good for Alex (trust-guided decision). Two
plans (collection and author) do not match; their
fulfilment F = .35 is lower than the QoS of .50
(the Quality of Service threshold).
5. On the other hand, two trusted plans match
(fulfilment F=1.00 as author and collection
match higher than .50 of the QoS). Eloy’s
Agent sends these 2 actionplans to Alex. Local DB
Alex’s SA Eloy’s SA Albert’s SA
AV
2 results
found
10/05/2013
19. Search Example
15
ActionPlan Eloy Albert Alex Access-
Point
Image conversion (3versions) X X X
Tech medatada extraction X X X
Local AV Check X X
Local AV Clean X X
Remote AV Check X X
Remote AV Clean X X
Generic Image Conversion (certified) X
Calc MD5 X
1. Alex searches the keyword “AV” in the
“Application local DB” No result is found.
2. Alex’s Agent searches for a certified actionplan
into the Access Point (institutional point of
view, with certified plans). No result is found
again.
3. The Agent asks its friend Eloy (distance = 1).
4. Eloy’s agent checks whether 4 matching plans
are good for Alex (trust-guided decision). Two
plans (collection and author) do not match; their
fulfilment F = .35 is lower than the QoS of .50
(the Quality of Service threshold).
5. On the other hand, two trusted plans match
(fulfilment F=1.00 as author and collection
match higher than .50 of the QoS). Eloy’s
Agent sends these 2 actionplans to Alex.
6. Alex’s Agent receives the actionplans from
Eloy’s agent and ranks them for Alex.
7. The plans are added to Alex’s actionplan
collection and are tagged “Eloy” as provider and
author. Alex Eloy Albert
Local DBAdd and Rate
new plans
10/05/2013
20. Search Example
15
ActionPlan Eloy Albert Alex Access-
Point
Image conversion (3versions) X X X
Tech medatada extraction X X X
Local AV Check X X
Local AV Clean X X
Remote AV Check X X
Remote AV Clean X X
Generic Image Conversion (certified) X
Calc MD5 X
4. If Eloy’s Agent did not found an Action Plan, then:
1. Eloy’s Agent will show to Eloy the
question
2. Eloy’s Agent will forward the question to
Albert’s Agent.
Alex Eloy Albert
10/05/2013
AV
AV
22. What we learned: the DP relevance
• Novel approach, that way never seen before (both the
approach and the prototypic solution)
• Adequate services (tools) offered for organizational
and individual users alike
• Certain degree of “intelligence” shown by the solution
• Step forward in bringing software agent technology
into DP domain (not yet reached the end of that road)
• Personalization of user access to solution (profile,
preferences, own resources)
• No prerequisite for expertise in DP before using the
solution (depends on complexity of action plans)
1719.01.2011
23. Future work (PROTAGE++)
1. The pro-activity of the solution is essential even if a “one-size-fits-
all” solution is not expected to exist.
2. There is a need for expanding the potential of “digital preservation
intelligence” embedded into the agents.
3. The system should be able to incorporate and analyze user’s
collections. This adds to the aspect of personalization.
4. The system should make sure that it is able to efficiently point at
particular problems in the domain, to problematic overall areas,
and to specific potential risks.
5. The users need more help in formulating questions for searching
for existing action plans.
6. A solution dedicated to “ordinary” private users with little or no
experience in digital preservation requires the provision of a user-
adapted graphical user interface with more predefined
customizations but it pays off.
7. The solution must provide specific features for memory institutions
allowing them to integrate the agent based technology into their
daily procedures.
1819.01.2011
24. Main Achievements of the Project
(2) AUTOMATION
- Execution of DP
tools (locally and as
web services).
- Technology watch
function (through
monitoring agents).
(4) TRUST MODEL
- Trustable access to
trusted information on
DP.
(5) AWARENESS
- An understanding of
others’ activities that
bring context to ones
own activities
(1) AGENTS
- Agent ecosystem as
a design
concept for DP tools.
- Provide context-sensitive
access to DP information
from trusted knowledge
bases.
(3) KNOWLEDGE
MANAGEMENT
- Reduces the knowledge gap
between MIs and other users
- Provides new means for MIs
to reach their ”clients”
- Reach more user groups.
- Practical means for sharing
DP information through social
networks and crowd-sourcing
MI = memory
institutions
DP = digital
preservation
PROTAGE
1919.01.2011
25. Future Technology Challenges
User adapted GUIs Specific features for
memory institutions
Expand the intelligence potential –
Pro-active solution Analyze users´
collections
Point at problem
areas/risks
Help formulating
questions
Identify other
domains
2019.01.2011
26. Thank you
Peplluis de la Rosa and Albert Trias
peplluis@eia.udg.edu
albert.trias@udg.edu
10/05/2013 21