Donatella Castelli

1

• Donatella Castelli
• Istituto di Elaborazione della Informazione CNR
• Pisa (Italy)

2
General information

• Funding programme
• EU 5FP - Action III-Multimedia Content and
  Tools
• Effort 205 p/m
• Duration 30 months
• Commencement date February 2001

3
General information

• Participants
• CNR (Italy, scientific coordinator)
• FORTH (Greece)
• GMD (Germany)
• University of Dortmund (Germany)
• ERCIM (France, administrative coordinator)

4
Objective

• Develop an open service archive environment
  composed by a set of interoperable services able
  to
• support scholars
• in interacting with multi-disciplinary archives
• as members of networked peer communities

5
Design decisions
Cyclades environment



Common Interface
Common Interface
Common Interface
Archive1
Archive2
ArchiveN
Built on top of the metadata harvesting layer
established by the OAI protocol
6
Design decisions
Independent services Cyclades Mediator Services
For each service - a well established access
protocol - a precise
service description
7

• Cyclades end-user functionality

8
Virtual Collections

• The user perceives an information space
  structured into a set of collections

9
Virtual Collections
simple-search Input keywords in the
abstract Preconditions keywords in
English Outputset of ltdocument.id,author,titlegt E
ffectreturns the specified output for all the
documents that contains the given keywords in
their abstract

• A set of specific operations is available for
  each collection

10
Virtual Collections

• The set of collections is dynamic

Composition
?d ? d ? ? /? (C1, C2, ,Cn) ? cond(d) ?
Selection based on description of the collection
Restriction on the document metadata fields
Time 0 one collection for each Cyclades
harvested repository
11
Information seeking

• Search

12
Information seeking

• Multilevel Browse

INRIA GMD IEI-CNR CNUCE-CNR ...
13
Personalisation

• The user creates a personal, dynamic hierarchy of
  folders for the storage of the records that match
  his information needs

• The system learns users information needs (user
  profiling)
• by observing the content of the folders,
  their organisation and the user behaviour

14
Personalisation

• New data is classified into the right topic
  folder automatically

The efficiency of the classifier may be improved
by exploiting contextual information (e.g. the
searching collection)
15

• Donatella
• (content based and collaborative)

Recommender

• By observing the behaviour of the users and their
  profiles the system identifies similar users

16
Cooperative Work

• A shared working space can be created by groups
  referencing
• user own documents
• collections
• recommendations, related links, textual
  annotations, ratings,

17

• Will we be able to build a quality service
• on the OAI low barrier interoperability
  framework?

18
An example

• Collection description is used by
• the end-user to select the collection of interest
• the collection administrator to select the
  collections required to build a new collection
• the document classifier to increase efficiency
• the recommender system to provide information
  about new collections

19
An example

• What should a collection description contain
• content descripton
• subject
• coverage
• metadata formats
• metadata language
• content format
• digitalized content yes/no

20
An example

• Collection content description we need to rely
  on the document description
• DC fields are optional
• Collection subject we need to know if the term
  in the subject field is a classification code, a
  free term, or term of a controlled vocabulary
  the controlled vocabulary the language of the
  terms
• DC is unqualified

21

• Will be Cyclades a service for a subset of the
  OAI registered archives?

22
Background Tecnology

• Browse - Desire (University of Dortmund)
• Personalisation - EUROgatherer (CNR)
• Cooperative work CSCW, Cookpit (GMD)

23
Cooperative Work

• A group of users that often access similar
  documents may enter into a long term relationship
  and eventually evolve into a working group (if
  only they become aware of each other)