FaceTag - IASummit 2007

FaceTagIntegrating Bottom-up and Top-down Classification in a Social

Tagging System

IA Summit 200722-26 March, Las Vegas

The Long Tail

Our culture and economy is increasingly shifting away from a focus on a relatively small number of "hits" at the head of the demand curve and toward a huge number of niches in the tail

1. Costs of production fall

2. Costs of storage and distribution fall

3. We need better filters and aggregators

The UGC Era

• Among adults, 35% of internet users have created content and posted it online

• User generated content can be anything produced by the user: text, audio, video, photos, categories or ranks, clips

(PEW, December 2005)

Information Overload

• 601 exabytes in 2010

• 70 percent of the world's digital data will be created by individuals

(IDC 2007)

• 185 exabytes (billion gigabytes) of storage needed last year

– 12 stacks of books that each reach from the Earth to the Sun

– 3 million times the information in all the books ever written

– more than 2 billion of the most capacious iPods

How can we navigate and find this information?

An Emerging Approach to Distributed Classification

• Collaborative tagging systems to organize, browse and share personal collections of resources through the introduction of simple metadata

• Folksonomy = user-generated classification, emerging through bottom-up consensus

• The basic idea is simply to make people share items annotated with keywords

Del.icio.us Homepage

Collaborative Tagging Examples

• An incomplete list of tagging systems

• They are web-based collaborative systems for:

– building a shared database of items

– a flat metadata vocabulary

– metadata driven queries

– to monitor change in areas of interest

– discover emergencies or trends

Have tagging systems solved the issue?

Del.icio.us Homepage

Del.icio.us Tagcloud

Food Tag in Del.icio.us

Down the Long Tail

TAG CLOUDS

HITS LONG TAIL

?

More Tags for the Same Thing

How many apples?

VS VS VS

Folksonomies’ Properties

Pros

– Trade-off between simplicity and precision

– Matching users’ real needs and language

– They are inclusive (nothing is left out)

– Discovery of information and Serendipity

– A forced move (the environment makes the difference)

– Better than nothing (when traditional classification is not viable)

Cons

– Language related issues

– User Experience issues

Language Issues

• As a result of intrinsic language variability, tagging systems are also implicitly plagued by:

– Polysemy (window: the hole or the pane of glass)

– Homonymy (apple, jaguar)

– Plurals (blog/blogs, folksonomy/folksonomies)

– Mistyped tags (folsonomy)

– Synonymy (tags, tagging, folksonomy)

– Ego-oriented tags (toread, funny, interesting etc..)

– Basic level variations (dog/beagle)

• These problems can dramatically reduce the effectiveness of the application and the benefits brought on by tagging systems.

User Experience Issues

– Low findability quotient and Low scalability

– High semantic density (very few well-known topics dominating the scene)

– An alphabetical criterion limits the ability to explore the tag cloud

– A flat tag cloud cannot visually support semantic relationships

• Tag clouds are visual interfaces for information retrieval that provide a global contextual view of tags assigned to resources in the system

• Flat tag clouds not sufficient to provide a semantic and multidimensional browsing experience

Navigating Large Domains

• Information seekers in large domains of objects express the desire of having to deal with meaningful groupings of related items, in order to quickly understand relationships and decide how to proceed [Marti Hearst 2006].

• How to generate and navigate such groups from a flat set of objects is anyway a different matter.

• Taxonomies, Clustering and Faceted Classification have been proposed in the past as useful techniques

Our Toolbox!

Taxonomies

• Taxonomies = coherent and complete system of meaningful labels which systematically organize a domain

• Organically crafted before starting to catalogue trying to guess user needs and content types

• Authoritative centralized view

• High precision avoiding ambiguity, hierarchical structure to give context

• Main drawbacks:

– An a priori and monolithic hierarchical organization do not have the ability to match the vocabulary and the varied ways of thinking of different users.

– Expensive to build and maintain by professional indexers

Clusters

• Clustering = the act of grouping items according to some measure of similarity

• It reduces the semantic density and improve the visual consistency of tag clouds

• But it generate messy groups, conflate many different dimensions and does not allow refinement and follow-up queries

• Users prefer clear hierarchies with categories at uniform levels of granularity over the messy, unpredictable and unlabeled groupings typical of clustering techniques

Facets

• Facets = orthogonal descriptors (categories) within a metadata system

• Each facet has a name and addresses a different conceptual dimension or feature type relevant to the collection

• Each object is classified combining labels from different facets.

• Facets can be:

– Flat or hierarchical

– Assigned single or multiple values

It is hierarchial

Facet name

393 resources here

Facets

– add structure and context to tags

– navigate along several dimensions simultaneously

– seamlessly integrate browsing and searching

– refine and broaden filtering criteria

• Benefits

• Hierarchical faceted metadata can be used to

– Better support for exploration, discovery and iterative query refinement.

– Easier to understand the meaning of tags

– Large tag clouds more browsable

– Reduction of the mental work (favoring recognition over recall)

• Usability studies show how this approach is preferred over single hierarchies and clusters

What do we need next?

What do we need yet?

• Middle ground (between the pure democracy of bottom-up tagging and the empirical determinism of top-down controlled vocabularies*)

• Metadata ecology: merge and leverage emerging and traditional tools to improve findability

• Metadata ecology as a fusion not only a coexistence

*Alex Wright: http://www.agwright.com/blog/archives/000900.html

FaceTag Contributions

• FaceTag tries to limit the impact of polysemy, homonymy and basic level variation introducing a multidimensional and more semantic paradigm

• Goals: improving usability, findability, browsability, serendipity and scalability of the system.

• FaceTag mixes three contributions to social tagging systems:

– Tag hierarchies

– Facets of Tags

– Tagging and searching mixed

Faceted Analysis

How to choice facets?

2 main roads

Freehand made

• Each facet is created at the moment

• The subject is freely deconstructed in several aspects

Standard based

• Each facet is chosen using the guidelines fixed by Ranganathan or CRG

• The subject is deconstructed following a “general scheme”

• “General scheme” works as a prototype of every particular faceted scheme

Facets derived from the CRG general scheme

CRG FACETAG

Thing [Documents, Resources]

Type Resource Types (e.g. online report, case study, blog...)

Part --

Property Language

Material Themes

Process --

Operation [Activities]

Product Deliverables

By-product --

Patient Purposes, [Markets] (e.g. Industry, Health ...)

Agent People

Space [Country]

Time Date

FaceTag facets

FACETAG FACET SAMPLES

Resource Types white papercase studyblog > enterprise web

Language predefined values (select box ISO Standard ISO 639-2)

Themes competitive analysisclassification > facetsweb 2.0 > folksonomiesinformation design > navigation design > breadcrumbs

People Weinbergere. ReissMorville

Purposes public administrationhealth, education > conferences > www2006

Date date of publication (through a calendar)

The Real Application

Preliminary User Research

• Preliminary facets were derived from the CRG schema

• These facets will be revised through an iterative bottom-up (card sorting) process

• The goal: to elicit the best facets from a wide set of IA bookmarks already online

Facets Evaluation

• An intuitive, easy to learn and easy to use interface is probably the single most effective way to support and stimulate participation

• The new user interface has been designed through documented heuristics and patterns

• Verified at each iterative step by small usability tests.

• Critical task: the assignment of new bookmarks and the association of tags to relevant facets.

User Interface Evaluation

Conclusions

Folksonomies are dead. Long Life to folksonomies

• Low Benefit & Mainstream Ready?• 28% of online Americans tag content (PEW December 2006)

• Tagging is commoditized, but it has a long way to go• A few ideas with FaceTag

Hiera

rchie

s

Facet

s

Tag

Suggestio

nHie

rarc

hical

Clust

erin

gTaggin

g

Decay

Advance

d

Navig

atio

n

Synonym

s

Synta

x

Control

User

Ratin

gs

Questions?

Thanks for your time!

www.facetag.org

FaceTag CrewEmanuele Quintarelli ([email protected])

Andrea Resmini ([email protected])

Luca Rosati ([email protected])

Luca Mascaro ([email protected])

Diego La Monica ([email protected])