Bonolo∗A General Digital Library System for File-based Collections

Lighton Phiri Kyle Williams Miles RobinsonStuart Hammar Hussein Suleman

Digital Libraries Laboratory

Department of Computer Science

University of Cape Town

November 13, 2012

∗

Sotho word meaning easy.

http://martinwest.uct.ac.za

3 of 23

http://lloydbleekcollection.cs.uct.ac.za

4 of 23

Motivation

Preservation costs

Preservation lifecycleHeritage funding model

Technical skills and education

Content curators skillsetSteep learning curve for most solutions

Internet bandwidth

Bandwidth intensive solutionsCloud-centric solutions not feasible

Existing solutions

Complexity

5 of 23

Design Principles

Design for least possible resources

Flexible design to facilitate extensibility

Hardware and/or software platform independence

Heterogeneous object, metadata and service integration

Minimalist design approach

Simplified preservation process

Structured organisation of data

Support for community and international standards

6 of 23

Prototype Implementation

Search

Browse Commenting

Metadata

Clients

Services

Repository

Thumbnails

File System

Indicies

Authentication

Indexing Tagging

Curator End User Solr

Content

7 of 23

Tools and Technologies

Core technologies

Apache TomcatCSSHTMLJavaJavaScriptXML

Additional software components

Apache SolrImageMagickSQLite

8 of 23

Repository Design

Metadata

Thumbnails

IndiciesContent

Hierarchical structure

Digital content and metadata stored on file system

Objects stored with corresponding metadata record

Metadata records encoded using XML

9 of 23

Repository Design (2)

archive/

notebooks

wilhelmbleeknotebooks

BC 151 A1 4 001

A1 4 1 00001.JPG

A1 4 1 00001.JPG.metadata

:

BC 151 A1 4 001.metadata

:

wilhelmbleeknotebooks.metadata

drawings

stories

10 of 23

End User Interface

Curator End User Solr

Authentication

Browse

Commenting

Search

Tagging

11 of 23

End User Interface (2)

12 of 23

Curator Interface

Curator End User Solr

Authentication

Batch import and export

Browse

Collection and Object management (add, edit, delete)

Indexing

Search

13 of 23

Curator Interface (2)

14 of 23

End User Interface UX Experiment

Objective

Feasibility of building user interface using file storeAssess impact of file store on user experience

Target Group

Individuals with basic Web experience17 participants (80% students)

Approach

Informed consentPre-experiment briefingTypical DL end user tasks (search, browse)Post-experiment questionnaire

15 of 23

End User Interface UX Experiment (2)

Search& Browse Navigability Similarity General

Average UX Ratings

Survey

Participants

0

1

2

3

4

5IntuitiveSimpleSatisfiedExpectedResponseEffectiveBeneficial

16 of 23

Curator Interface UX Experiment

Objective

Assess user experience when performing curation tasks

Target Group

Individuals with no experience working with DL toolsSocial networking site recruitment23 participants

Approach

Intrinsic Motivation InventoryFive (5) minute ’HOWTO’ screencastCuration tasks with two datasetsOnline questionnaire

17 of 23

Curator Interface UX Experiment (2)

Interest/Enjoyment Perceived Competence Value/Usefulness

Average UX Ratings

Survey

Participants

0

5

10

15

20

25

Strongly AgreeAgreeNeutralDisagreeStrongly Disagree

18 of 23

Curator Interface UX Experiment (3)

Participants general comments

”· · ·Also, I fail to see how Bonolo differentiates itself fromsomething like Dropbox. I can create a folder structure on myPC and upload it to Dropbox very easily. I can then browse myfiles and folders in Dropbox’s web interface.· · · ”

”· · · I have to say though that I managed to complete the taskswithout watching the video (which is a great sign I think). I’mimpatient with manuals but even worse with instructionalvideos· · ·”

19 of 23

Repository Performance Experiment

Objective

Impact of file store structure on performancePerformance metrics: response time

Test Environment

Intel Core 2 Duo CPU E7400@ 2.80GHz2 GB RAM32-bit Windows 7 Ultimate edition

Approach

Structured and unstructured collectionsExponential increase of files in collectionsLoad time and corresponding data transfer during navigation

20 of 23

Repository Performance Experiment (2)

1,024 2,048 4,096 8,192 16,384

0

1000

2000

3000

4000

5000

Files in Directory

Tim

e(M

illisecon

ds)

StructuredUnstructured

21 of 23

Conclusion

Experimental results look promising

EffectivenessUsabilityMedium-sized collections

Work in progressEvaluation

FlexibilityScalability

Future workReference implementation

Design principlesExtensibility

22 of 23

Thank You

Questions?

Additional Information

http://dl.cs.uct.ac.za