SAmgI: Automatic Metadata Generation v.2

SAmgI: Automatic Metadata Generation v2.0

Michael Meire, KULeuven, Belgium

Xavier Ochoa, ESPOL, Ecuador

Erik Duval, KULeuven, Belgium

2007

Agenda

• Why Automatic Metadata Generation?

• AMG v.1 – What Went Wrong

• SAmgI – What is New• SAmgI in Action

• SAmgI to the Test

• Make SAmgI yours!

Economy of Abundance

Economy of Abundance

Put your LMS here!

Managing Abundance

• But those resources have to be:– Found– Shared among tools

– Evaluated by users– Assembled together– Related to each other– Branched in new versions– …

• But those resources have to be:– Found– Shared among tools

– Evaluated by users– Assembled together– Related to each other– Branched in new versions– …

Managing Abundance

This is a job for …

METADATAMETADATA

But WHO will create all this

metadata?

ARIADNE Grow

ARIADNE Grow

Librarian metadata

doesn’t scale!

Automatic is the Way…

• GATE (Cunningham et al. 2002)

• AMeGA (Greenberg et al. 2005)

• Magic system (Ying et al. 2005) • AMG (Cardinaels et al. 2005)

Use information about the Object and its Context to extract or generate its metadata

AMG v.1 Design

• Integrate existing extraction algorithms

• Two groups of Generators/Indexers:– Object Based Extractors– Context Based Extractors

• Extraction/Generation Toolbox– Keywords, Summarization, Language, EXIF

AMG v.1 Design

AMG v.1 What Went Wrong

• It was an application

• Not as flexible as initially thought

• Initial set of extraction tools was too basic

• It was Java only

• It was an application

• Not as flexible as initially thought

• Initial set of extraction tools was too basic

• It was Java only

AMG v.1 What Went Wrong

If at first you don't succeed…

SAmgI – What is New

• Simple AMG Interface

• Main Design Goals:– Easily Extendable– Service Oriented

Easily Extendable

Service Oriented

Service Oriented

Service Oriented

SAmgI in Action

• Projects deploying SAmgI:– MACE (http://www.mace-project.eu/)

– MELT (http://info.melt-project.eu/)

– ACKNOWLEDGE (http://projects.ibbt.be/acknowledge)

– GLOBE network (http://globe-info.org/)

SAmgI in Action

• Try it online:– http://www.cs.kuleuven.ac.be/~hmdb/amg

SAmgI to the Test

• 22 reviewers for 20 LOM records– 10 human generated (from Ariadne KPS)– 10 SAmgI generated (from Prolearn deliverables)

• 7 quality parameters (Bruce & Hillman framework)

– Completeness– Accuracy– Provenance– Conformance to expectations

– Logical consistency and coherence

– Timeliness– Accessibility

SAmgI to the TestAverage Grade

0

0,5

1

1,5

2

2,5

3

3,5

4

Comple

tnes

Accur

acy

Prove

nanc

e

Confo

rman

ce

Coher

ence

Timeli

ness

Acces

ibility

Quality Parameter

Qu

alit

y V

alu

e (0

- 6

)

AutomatedManual

SAmgI to the TestAverage Grade

0

0,5

1

1,5

2

2,5

3

3,5

4

Comple

tnes

Accur

acy

Prove

nanc

e

Confo

rman

ce

Coher

ence

Timeli

ness

Acces

ibility

Quality Parameter

Qu

alit

y V

alu

e (0

- 6

)

AutomatedManualNo

significative difference in Quality!

Make SAmgI yours!

• Option 1: SAmgI as a black box– Minimum programming – No context information

– Small set of metadata.

• Option 2: Create a CBG– Reuse all the other services

– Rich metadata set– Some programming required

Make SAmgI yours!

• Source:– Sourceforge – ARIADNE KPS Project

(http://ariadnekps.cvs.sourceforge.net/ariadnekps/SAmgI/)

• Binaries and Examples:– SAmgI Website

(http://www.cs.kuleuven.ac.be/~hmdb/amg)

Thank You! Dank U! Gracias!• Questions, Comments, Critics…

are all welcome!!

Michael [email protected] http://www.cs.kuleuven.ac.be/~michael

Xavier [email protected]

http://www.cti.espol.edu.ec/xavier

Erik [email protected] http://www.cs.kuleuven.ac.be/~erikd