Building a Digital Library with Fedora

International Conference on Developing Digital Institutional Repositories

Hong KongDecember 9, 2004

UVA Digital Library Assumptions:

• All media, all content types integrated into one collection

• A network that is built to be a part of a global network

• The global network will be built by libraries, governments and corporations

• Searching and browsing are equally important

UVA Digital Library Assumptions (cont.):

• We will provide to tools to give access and make use of our collections

• Any given resource can be presented in any number of contexts

• Increasingly, we will be faced with born-digital materials

• This is going to take a very long time …

The digital library as a web of content

Text Collections

Te xts

M o de r n E ng l i s h C o l l e c t i o n

P ag eIm ag e s

Explicit Data Modeling for Serials

C av alie r D aily

D aily I s s u eD aily I s s u e

M o de r n E ng l i s h C o l l e c t i o n

Collecting Scholarly Projects

L e ave s o f G r as s

Te xt 2Te xt 1

M o de r n E ng l i s h C o l l e c t i o n

The W hi tm an P r o je c t

Art and Architecture Data

W o rkO bje c ts

A rt a nd A rc hite c ture C o lle c tio n

Im a g eO bje c ts

Quantitative Data Collections

D a ta s e tD e s c riptio n

O be jc ts

Q ua ntita tive D a ta C o lle c tio n

D a ta ba s eO bje c ts

The Flexible Extensible Digital Repository Architecture

• Developed at Cornell under an NSF grant• UVA Library re-interpreted the architecture and

created the first practical implementation• 3 year project funded in 2001 by Andrew W.

Mellon Foundation to create open-source system• Another 3 years of development funded by Mellon

in 2004

Fedora is a set of web services that can provide a foundation for a variety of information management strategies.

– Supports client applications through SOAP or HTTP connections

– Provides back-end web services for content through behavior objects

– Provides both management and access APIs– Provides a search index aimed at repository

management– Fine-grained policy enforcement

Users access data objects through behaviors.

PI D

B e h a v io r

B e h a v io r

B e h a v io r

B e h a v io r

D a ta s tre a m

D a ta s tre a m

D a ta s tre a mUs e rs

File

File

H TTPS e rv e r

A pplica t io n s

B e h a v io rD e f in in t io n

B e h a v io rM e ch a n is m

Pro ce s s

A data object is one unit of content.

Persistent ID (PID)

Default Disseminator

System Metadata

Datastream (item)

Digital object identifier

methods for disseminating “views” of content

metadata about history and policies

Datastream (item)

Datastream (item)

Your Extension

Your Extension

set of content or metadata items

Persistent ID (PID)

Behavior DefinitionMetadata

SystemMetadata

DatastreamsData Object

Persistent ID (PID)

Service BindingMetadata (WSDL)

SystemMetadata

DatastreamsWeb

Service

behavior contract

behavior

subscriptio

n

data contract

Persistent ID (PID)

Disseminators

Datastreams

System Metadata

Behavior Mechanism Object

Behavior Definition Object

Disseminators for Data Normalization

• Can deliver datastream content directly

• Can transform content into other sizes or formats for delivery

• Can be used to hide differences among objects of a given type

Disseminators as User Interface• Can deliver a “module” of user interface

appropriate for the object

• Different user interfaces for difference purposes or audiences

• Easy to add new types of collections by adding new modules of code

• The set of all behavior objects can be used as a database of code modules

• Can provide a way to collect the “look and feel” of scholarly projects in a formal way

Relationships Among Objects

• Relationship metadata datastream in the data object

• Describes adjacency relationships among objects

• RDF data of the form:

PID – typeOfRelationship – relatedObjectPID

• Can used to assemble collections for such things as creating full-text search indexes

• Can build graphs of relationships to feed into a variety of user interfaces

• Uses Resource Description Framework (RDF)

• The repository can be configured to index any combination of the following aspects of a digital object:– System metadata properties– Dublin Core metadata– Metadata about datastreams and disseminations– Relationship metadata– Internal dependencies (e.g., between datastreams

and disseminators)

The Resource Index

I n te g ra te dS e a rch

I n te rfa ce(R o o m s )

V irg oC a ta lo g

D ig ita l D is co v e ryI n de x

PID

Disse m ina tors

S yste m M e ta da ta

De sc M e tad ata

Admin M e tadata

G DM S F ile

PID

Disse m ina tors

S yste m M e ta da ta

De sc M e tad ata

Admin M e tadata

TEI F ile

PID

Disse m ina tors

S yste m M e ta da ta

De sc M e tad ata

Admin M e tadata

EAD F ile

A rt a n d A rch .S e a rch

I n te rfa ce

Te x tS e a rch

I n te rfa ce

Fin din g A idsS e a rch

I n te rfa ce

PID

Disse m ina tors

S yste m M e ta da ta

De sc M e tad ata

Admin M e tadata

Imag eDatastre ams

PID

Disse m ina tors

S yste m M e ta da ta

De sc M e tad ata

Admin M e tadata

Imag eDatastre ams

Ele ctro n icJ o u n a ls a n dD a ta ba s e s

PID

Disse m ina tors

S yste m M e ta da ta

De sc M e tad ata

Admin M e tadata

F in d in g aid sInd e x

PID

Disse m ina tors

S yste m M e ta da ta

De sc M e tad ata

Admin M e tadata

Art an dArch ite ctu re

Ind e x

PID

Disse m ina tors

S yste m M e ta da ta

De sc M e tad ata

Admin M e tadata

Te xt Ind e x

UVA First Implementation

Arch Demo

Disseminators

• Two default disseminators on every object– Default access behaviors, i.e. getPreview, getFullView,

getLabel, getDefaultContent

– Administrative and descriptive metadata behaviors

• Class-specific disseminators, i.e. image and text disseminators

• Search services to be provided using collection object disseminators

Text Collections: three models

• TEI transcriptions of texts plus page images• TEI transcriptions only• Page images only, but the text represented by a

minimal TEI file

Future Fedora Development

• Improved infrastructure for workflow• Support for building indexes for searching• Infrastructure for building federations of Fedora

repositories• Enhance performance• Support for preservation• Begin organizing a Fedora development

consortium

Fedora Project web site:http://www.fedora.info

UVA Digital Initiatives:http://www.lib.virginia.edu/digital