Tag! Are You It? Taxonomies & Folksonomies in Practice Stephen Rhind-Tutt, President, May 28, 2008.

Tag! Are You It? Taxonomies & Folksonomies in Practice

Stephen Rhind-Tutt, President, May 28, 2008

Overview

• About Alexander Street Press• Is there value in social tagging?• What lies behind the value• Some practical examples• The future

Alexander Street Press

Performing Arts, Drama, and Film

World Literature

Women’s History

Religion

CounselingMusic

Social and Cultural History

Sociology

Black Studies

American Civil War

The People vs. The Machine

vs

“All tag-based classification of Internet resources (such as web sites) is done by human beings, who understand the content of the resource, as opposed to software, which algorithmically attempts to determine the meaning of a resource.”

Wikipedia, entry on social bookmarking, May 24th, 2008

The masses speak…

“All tag-based classification of Internet resources (such as web sites) is done by human beings, who understand the content of the resource, as opposed to software, which algorithmically attempts to determine the meaning of a resource.”

Wikipedia, entry on social bookmarking, May 24th, 2008

The masses speak…

• Not all tag based classifications are done by humans – machine based tagging is large and growing• Many humans don’t understand or agree on ‘the content of the resource’• Who’s to say what the content actually is?• Who’s to say what a resource is?• Social tagging requires computers to leverage collective human input using algorithms amongst other tools.

Actually it’s not the machine vs. the human – it’s the machine and the human

Another view…

Is there value in social tagging?

>3,000 photos of Alexandria, VA

Performance

Can Social Bookmarking Improve Web Search?, Paul Heymann, Georgia Koutrika, and Hector Garcia-MolinaDept. of Computer Science, Stanford University

Folksonomy

Cirrus on delicious

Item # 22 describes cirrus (the cloud)7 people tagged it.

Citeulike

Folksonomy TaxonomyEvolving terms Good Poor

New terms Fast Slow

Accuracy 80%? 95%?

Structure Low High

Hierarchy Behavior generated

(Popularity)

Editor generated(Accuracy)

Search Precision Variable Consistent

Customization High Low

Search Exhaustiveness Low High

Taxonomy vs. Folksonomy

Popularity

Effectiveness

Folksonomies

Taxonomies

Taxonomy vs. Folksonomy

49.7 m users (E-Bay)2.7 Bn searches per month (Google)

# of users * frequency of use * quality of use- (100 * bad use) ?Effectiveness

Subject understanding * MLS training ?Effectiveness

Looking closer…

Internet Trust ‘anti pattern’

Low

High

Medium

Effectiveness

1. System started by trusted parties

2. Benefits from being non-hierarchical, easy participation, quality participants

3. Unwashed masses start using it

4. Systemalmost breaks

5. Controls put in place

AfterThe Journal of the Future, Geoffrey Bilder, SSP, 2005

Time

Social Tagging

• Personalization• Speed• Discovery• Reach• Versatility

• No standard keywords• No standard structure/hierarchy• Unorthodox and personal tags

• Multiple Meanings• Mis-tagging due to spelling

• Synonym/antonym confusion• Advertising• Spamming

The Good The BadThe Mixed

ActiveUser assigns termsUser selects termsUser comments

Human and machine interactions

Passive• Most commented, blogged

e-mailed, viewed, cited…• Terms culled from associated

articles, comments…• Measuring link traffic, use,

search terms matched to results.

• Data mining, authority inference,

Google’s Page rank, etc…Delici.ous, flickr

Much social tagging is a by-product of other processes

• Central place to store your bookmarks (Delicious)• Place to share your photographs (Flickr)• Place to meet friends (My Space)• Store your bibliography (Zotero, Citeulike)• Have fun (Google Image Labeller)• Give students classwork (Alexander Street Press)

Leveraging existing processes…

Activity vs. Passivity

Active PassiveMixed

DeliciousASP Submission tools

Google Search

ZoteroCiteulike

ASP Playlists

Issues – user tagging

Tags

Philadelphia? Shirts? Women’s Rights? President?

Issues – granularity

Different views of the same item

CAB – (Husbandry)

Agricola (Agriculture)

OSH-ROM(Occupational Health and Safety)

Biosis (Species)

Long term factors influencing combustion and burn rates in North American forests. David Jones, Journal of Forest Husbandry, Sept 1999.

Humans are good at…

What works…

•Playlists on ASP’s music and video products – >20,000 users

•Over 120,000 playlists created so far

• 1,000 created by ASP

• 42,000 user created

• 80,000+ derivative playlists

Example – Dance in Video

Playlists

Playlists

Playlists

Summary

Social Tagging evolution

Descriptive Power

Suggestions, mapping

Social Network

Importance/Trust Currency

Popularity

Larger networks

Trackback, RSS Feeds

• A lot to offer• It works best

• In (very) large networks• When it is a byproduct or alongside another process

• It will always ends up being (partially) controlled • Network effects aren’t open to most of us• Discipline focus is the best way to add value

The future of social tagging…

Man empowered by machine

www.alexanderstreet.com