BROWSING A STRUCTURED MULTIMEDIA REPOSITORY

Rui Lopes and Lu

ıs Carric¸o

LaSIGE, University of Lisbon, Edif

ıcio C6, Campo Grande, 1749-016 Lisboa, Portugal

Keywords:

Multimedia Repository, Hypermedia, Semantic Browsing, Web User Interfaces.

Abstract:

Browsing large amounts of assorted multimedia poses several challenges on creating a supporting user in-

terface. Even further, if metadata, structures and relations are deﬁned on top of multimedia fragments, the

complexity of a multimedia repository increases. However, this is also a path for exploiting new browsing

mechanisms. This paper presents insights on which aspects should be available on a task-oriented user in-

terface for browsing a structured multimedia repository, centred on speeding up interaction through media

independent navigation schemes and user advisory.

1 INTRODUCTION

Digital media support changed radically the way in-

formation is consumed. Nowadays, accessing dig-

ital contents is a common task, especially through

the Web and its overwhelming size. Also, recent

trends give the user an active role on producing con-

tents on different media (blogs, photographs, pod-

casts, videos) and establishing relations between me-

dia components, instead of just passively consuming

information.

With such a massive amount of information avail-

able, new challenges arise. Recent efforts (Berners-

Lee et al., 2001) envision a Web of semantically de-

scribed information, to be shared by different appli-

cations. With this framework, pieces of information

can be mixed, repurposed, and interlinked accord-

ing to different criteria (such as complementing an

e-learning environment with content from an online

encyclopedia).

To support such scenarios, there is a need for a

storage and retrieval capable repository for multime-

dia contents, metadata, and relationships. Ever fur-

ther, these contents should not be perceived as black-

boxed pieces of information. With the imposition of

content structuring, each fragment may have its own

semantics associated, allowing for its reuse in other

contexts.

In such a repository it is critical to have a large and

rich multimedia corpora, and its corresponding meta-

data and relations. To enable this, some tasks can

be automated (e.g., crawling for new content, infer

knowledge), but some tasks may only be performed

by humans (e.g, ﬁxing incorrect information origi-

nated from automation).

To cope with the amount of contents and the com-

plexity inherent of having metadata and relations in

the repository, its management user interface must be

carefully crafted. This paper discusses these browsing

aspects in a web-based interface for managing a struc-

tured multimedia repository that supports the storage

and description of multimedia information fragments.

2 GOALS AND REQUIREMENTS

The effectiveness of the user interface for browsing a

structured multimedia repository is based on achiev-

ing the following set of goals:

• Simple user interface: as a huge amount of con-

tents may need manual editing of their corre-

sponding metadata, the user interface must not be

an obstacle on performing these tasks;

• Ease browsing of multimedia fragments: brows-

ing tasks should be centred around content frag-

429

Lopes R. and Carriço L. (2007).

BROWSING A STRUCTURED MULTIMEDIA REPOSITORY.

In Proceedings of the Third International Conference on Web Information Systems and Technologies - Web Interfaces and Applications, pages 429-432

DOI: 10.5220/0001282404290432

 SciTePress

ments, their siblings, children, etc.;

• Alternate navigation methods: having different

methods for browsing the repository allows the

user to select the best way for ﬁnding or reaching

some multimedia content;

• Overall views: providing an overall view of the

repository’s contents helps the user to choose

what tasks have to be performed;

• Advice to the user: the user must be advised with

some pointers that otherwise would take too long

to be found through exploration;

• Collaboration: the user interface has to promote

multi-user activities, as a single user is unable to

perform all required tasks.

These goals can only be achieved by fulﬁlling the

following set of requirements:

• Large scale: the repository requires efﬁcient

forms for visualizing and browsing multimedia

contents, as traditional efforts are geared towards

single media types, or do not scale up;

• Multitasking: many tasks in the repository may

trigger actions that will take some time to com-

plete (e.g., uploading a video). This time can be

used to perform other tasks;

• Media agnostic: the browsing capabilities should

be independent from media types, resulting in a

coherent user interface where users will perform

better on any task;

• Cope with structures and relations: as the reposi-

tory also contains structures and relations, these

concerns should be explicitly present on the

browsing interface;

• Filtering: the browsing interface should provide

means for narrowing the browsable set of contents

through ﬁltering.

3 RELATED WORK

Typically, multimedia browsing is centred either on

a single media type (H

urst and Stiegeler, 2002; Al-

banese et al., 2004). The fact that real scenarios must

take into account all media types, poses severe dif-

ﬁculties on interacting with a multimedia repository.

Even further, search tasks consume more time, as the

user has to perceive the semantics of each content (in-

stead of using metadata to provide this information

instantly).

Recent studies (Lew et al., 2006) on content-based

multimedia information retrieval have identiﬁed what

challenges have not been achieved: creating new hu-

man centred methods based on exploratory interac-

tion; enforce collaboration efforts; and provide mul-

timedia assets taxonomic classiﬁcation and browsing.

Consequently, the architecture for a new multimedia

repository must take into account these challenges.

4 OVERALL ARCHITECTURE

An architecture was deﬁned as the ground basis for

an environment based on structured multimedia frag-

ments stored in a repository (Figure 1).

Structured Multimedia Repository

Admin

User Interface

Data

Mining

Doc.

Creation

API

Browsing

WWW

third-party

service

user

Figure 1: Overall System Architecture.

The main architectural decision relates to the

Web’s distributed nature as its central point. Every as-

pect within the architecture is a producer and/or con-

sumer of resources through services with speciﬁc in-

terfaces: an API is provided for developing these ser-

vices on top of the repository (e.g., documents cre-

ation or third-party services), whereas a user inter-

face provides repository management (including the

browsing aspects).

5 BROWSING INTERFACE

A Web-based user interface was created for manag-

ing the structured multimedia repository. This inter-

face supports different tasks, such as importing new

content in different media formats, triggering mul-

timedia document production tasks, editing content

metadata, or creating structures and establishing re-

lations between multimedia fragments. However, to

perform these tasks successfully, users need to grasp

the repository’s contents through different points of

view. Therefore, several browsing aspects within the

user interface were deﬁned and classiﬁed into two

main concerns: overviews and instances.

WEBIST 2007 - International Conference on Web Information Systems and Technologies

430

Faced with millions of multimedia fragments,

users may feel daunted on choosing a starting point

for working on repository’s contents. Overviews pro-

vide different possibilities for browsing the reposi-

tory and start new tasks, as seen on Figure 2. Five

overviews are provided:

Figure 2: Browsing Overviews.

1. Filter: a keyword-based ﬁltering capability is pro-

vided to the user, allowing to dynamically adapt

all other overviews accordingly. This feature en-

ables the user to explore different perspectives and

subsets of the repository without browsing a spe-

ciﬁc repository instance, helping the user to better

perceive which contents the repository has;

2. Pinned Blocks: users are provided with mecha-

nisms for pinning contents and viewing all pins,

thus leveraging collaboration efforts;

3. Tag Cloud: with tags, users can freely classify a

content according to their own perception. The

cloud feature displays all tags that have been ap-

plied by users, with font sizes relative to each tag’s

count. This results on visualizing a user-created

taxonomy of the repository’s content;

4. Suggestions: a mechanism searches the repository

for incomplete metadata, advising the user to ﬁll

the blanks. This feature enforces the complete-

ness of the repository;

5. Latest: the user is presented with the latest con-

tents and relations added to the repository, new

starting points for browsing activities.

After choosing a starting point for browsing, the

user is presented with its respective instance, as seen

on Figure 3. Three different aspects can be perceived

within the browsing instance:

1. Search: while browsing, the user has the option

to search the repository. However, instance search

feature enables more complex exploratory tasks,

such as multiviews;

2. Metadata summary: to quickly grasp the subject

of a given instance, the interface presents its corre-

sponding ontology, tags and title, links for down-

loading content and metadata. Both ontology and

Figure 3: Browsing Instance.

tags summaries are navigable objects mapped into

search tasks, providing the user with other brows-

ing possibilities;

3. Content: this section presents the current in-

stance’s content. As the repository is based on

structuring multimedia fragments, each structure

is dynamically built and presented, enabling the

user to grasp its full content. Also, each structure

node is itself a navigation point for browsing its

children within the repository.

Expading the metadata summary, a full view is

presented (Figure 4). Here, users are able to: (1) view

and edit metadata ﬁelds, and (2) view different avail-

able navigation possibilities (inclusion, composition,

and relations) for exploring the repository, based on

the current browsing instance. It is worth noticing that

metadata ﬁelds are also starting points for exploratory

tasks from the user (similar to Tag Cloud browsing).

Figure 4: Metadata view for Browsing Instance.

6 IMPLEMENTATION

The browsing user interface for the structured multi-

media repository was implemented with Web-based

technologies. As most users are comfortable on in-

teracting with the Web, this will ease interaction with

the browsing user interface. Simple multitasking is

also leveraged just by opening a new window or tab

BROWSING A STRUCTURED MULTIMEDIA REPOSITORY

431

in the browser, enabling its fast response. Another di-

rect consequence is the instant availability of the user

interface (no software installed on the client side), in-

creasing its dissemination and usage.

The repository was implemented on top of eX-

ist (Meier, 2002), a native XML database. This deci-

sion allows using XML formats for document struc-

tures (Walsh et al., 2002), metadata and relations.

Also, using an XML database allows for volatile data

schemes, as opposed to relational databases. Con-

sequently, any ontology instance can be inserted on

multimedia fragments metadata, thus enforcing ex-

tensibility to any knowledge inference mechanism.

Executing queries on the repository is performed with

XQuery (Boag et al., 2006). Different abstraction lay-

ers on information retrieval have been implemented,

and may be used in the future as basic constructs for

smarter mechanisms. On top of XQuery modules, the

browsing interface has been implemented as a thin

layer returning XHTML.

On top of the user interface, a set of unobtrusive

javascript functionalities were added to improve the

usability of its browsing characteristics. This way,

tasks that require high computational resources (e.g.,

complex searching within the repository) can be trig-

gered asynchronously, leveraging the user interface’s

responsiveness.

7 CONCLUSIONS AND FUTURE

WORK

This paper presented new browsing aspects for man-

aging a structured multimedia repository. These as-

pects were centred on speeding up exploratory activi-

ties within the repository, typically based on navigat-

ing between relations, structures, and semantic infor-

mation of a huge amount of multimedia fragments.

Two different browsing concerns were presented,

overviews and instances, providing exploratory cues

to start browsing the repository, and visualizing con-

crete fragments and their own browsing opportunities,

respectively. These concerns are crucial, given the

amount of fragments and the complexity of structures

that may be related to each other.

As future work, it is desirable to support searching

by example for images, sound and video resources.

Also, allowing the repository metadata layer to ref-

erence any Web resource (instead of just referencing

multimedia fragments stored inside of it) will ease

coping with repository feeding. Furthermore, by in-

troducing data mining techniques, better advice can

be provided to users. Lastly, other collaboration tech-

niques may improve the repository’s browsing tasks,

such as communities specialized in speciﬁc topics.

ACKNOWLEDGEMENTS

This work is being funded by Fundac¸

ao para a Ci

encia

e Tecnologia, through grant POSI/EIA/61042/2004,

and scholarship SFRH/BD/29150/2006.

REFERENCES

Albanese, M., Cesarano, C., and Picariello, A. (2004). A

multimedia data base browsing system. In CVDB

’04: Proceedings of the 1st international workshop on

Computer vision meets databases, pages 35–42, New

York, NY, USA. ACM Press.

Berners-Lee, T., Hendler, J., and Lassila, O. (2001). The

semantic web. Scientiﬁc American.

Boag, S., Chamberlin, D., Fern

andez, M. F., Florescu, D.,

Robie, J., and Sim

eon, J. (2006). XQuery 1.0: An

XML Query Language. W3C candidate recommen-

dation, World Wide Web Consortium (W3C).

urst, W. and Stiegeler, P. (2002). User interfaces for

browsing and navigation of continuous multimedia

data. In NordiCHI ’02: Proceedings of the second

Nordic conference on Human-computer interaction,

pages 267–270, New York, NY, USA. ACM Press.

Lew, M. S., Sebe, N., Djeraba, C., and Jain, R. (2006).

Content-based multimedia information retrieval: State

of the art and challenges. ACM Trans. Multimedia

Comput. Commun. Appl., 2(1):1–19.

Meier, W. (2002). eXist: An open source native XML

database. In Proc. of 2nd Annual International Work-

shop on Web and Databases, NetObjectDays 2002,

Erfurt, Germany.

Walsh, N., Muellner, L., and Stayton, B. (2002). Docbook:

The Deﬁnitive Guide. O’Reilly, 2.0.8 edition.

WEBIST 2007 - International Conference on Web Information Systems and Technologies

432