The Corporate Librarian

Tom Reamy from the Knowledge Architecture Professional Services (KAPS) Group led this session. No connection to the science fiction/fantasy author that I know of.

Themes

• People categorizing their stuff on their own terms (Matt Haughey, Metafilter) – plus or minus
• Obligatory snarky Andrew Keen quote in reaction to unnamed Web 2.0 evangelist
• Yeats, “Second Coming”
• Key is social mechanism for seeing other tags

Essentials of folksonomies

• Advantages – very simple to use, lower cost of categorization (spread over large population), flexible (can respond quickly to changes), higher relevance because in user’s own terms, supports discovery (serendipitous browsing), object-neutral (can tag bookmarks, documents and photos), better than no tags at all?, gets people excited about metadata
• Disadvantages – don’t work well for finding, no structure or conceptual relationships, lack of scale, highly personal (or popularity wins), errors
• Dangers – the ‘unwisdom of crowds,’ the tyranny of the majority (popularity of tags beats quality, narrowing of choices, losing content), belief that hierarchy and taxonomy not needed

Will social networking make better folksonomies? Tom’s research suggests not:

• A few tags dominate del.icio.us
• Quality is not popularity
• Most non-techies don’t tag, let alone re-tag
• Despite following NISO guidelines, folksonomies do not seem to work to let other people find things
• Regular people and infrequent users don’t do this stuff

Flickr facets analysis

• Faceted navigation is extremely powerful and easy to use
• 90% of Flickr content can be described by six basic facets- place, events, date, things/animals, color
• Subject matter is less than 1% of tags
• Works on lower level scales

Del.icio.us tags analysis

• Tags are subject matter, not facets
• High level topics
• Related terms are grouped by popularity, not by conceptual relations (that is, people who tagged something X also tagged it Y)
• One functional facet turned up – howto, tutorial, toread, todo
• Dominated by computer terms – searching on the design tag got 1 million computer-related results versus 3,909 on interior design
• Again, the tyranny of the majority
• Top 25 tags over time – basically the same set, with an order shift
• Del.icio.us findability:

• Too many hits
• No plurals (singular preferred) or stemming (Searching on blog turns up 1.7 million items, searching on blogs turns up 516,340. The contrast is more extreme for other terms.)
• Personal tags – not good for findability (e.g., cool, fun, funny) but interesting for social research

So, how do we improve folksonomies as finding tools?

• Add automatic facets to Flickr (one-time cost, some monitoring required)
• Cluster tags (entity extraction, populate facets and subjects, types of relationships)
• Add a broad general taxonomy of popular tags (tags as natural categories*, build up and down), start and evolve a simple two level taxonomy
• Evolve the quality of tags and the emerging structure of tags (preferred term is the popular term, add mechanisms to rank tags, taggers and categories)

Folksonomies and libraries:

• Library catalog (he said library catalogs didn’t include tags, as I recall)
• Internal service (I think – my scrawl was hard for even me to read)
• Enterprise (KM) contributor

University of Pennsylvania is experimenting – PennTags

LibraryThing:

• High level concepts, either too general or too specific
• Variety of terms is an issues (cognitive science required 40 tags for an adequate coverage of the subject)
• Strange tags (he cited “book” as a tag, less than helpful, or “my living room”)
• Facets and topics
• Redundant/inconsistent tags

It was about here that Tom realized he had way more material than time to present it – and as a result, he raced through the remaining slides and I’m not sure what he meant by much of this. I’ll add a link to his slides once they’re up.

What won’t work:

• Trying to “improve” users
• Social networking
• Either/or approach – folksonomies OR LCSH (Library of Congress Subject Headings)

What might work:

• Environment and dynamic social rules
• Integrate evolving social (something) structure – people, technology, policies, procedures with feedback incorporated
• Interpenetration of opposites (top-down, bottom-up and all over)
• Reduce the folk aspects and add more -onomy (Wikipedia has 2,000 editors)
• Increase the folk aspects – add discussion and context around tags

New library now?

• Central (something) feedback system
• Communities of practice
• Technology – enterprise content management/knowledge management platforms for the enterprise, metadata, policies, tag clouds, more sophisticated displays
• Ranking – either implicit or explicit, but rank tags, taggers, categories

He recommended starting with a formal taxonomy (I believe that’s right)

Library 2.0 is about social collaboration, not tagging

Folksonomies need to evolve

Tension between ease of use and findability

90% hype

Not much time for questions afterwards – I passed on my business card and asked if I could e-mail followup questions, and he said that would be fine. It turned out I was sitting next to a guy from CiteULike, and he noted he’d found only a 10% overlap between tags and formal subject descriptors.

I have lots of questions and pushback on points, but I’m interested what the audience out there thinks.

Technorati Tags: il2007