User-driven Semantics: Folksonomies Tags

advertisement
Semantic Web Technologies
• Brief Readings Discussion
• Class work: Research topics and Project
discussion
• Presentations
Explaining Folksonomies
• Folksonomies are when you describe something,
primarily for yourself
- For organizing
- For re-finding
• Folksonomies are (mostly) flat (no hierarchy)
• But others can take advantage of your work
- Because they think the same way
- Because they want to understand or find something
semantically
• Letting others define categories & classify
information (for you & others)
• Does time help or hinder folksonomies?
• Does the beauty pageant judging effect help or hurt?
Kinds of Folksonomies
• Broad - a lot of people are describing one object
- Delicious or Shadows
- Depth of descriptions if a lot of people describe (tag) the
object
- Lots of (potential) disagreement, but still plenty of
descriptions
- Social status & sharing bootstrap the process
- Can or will people learn & use a descriptive vocabulary?
• Narrow - one person describing (usually their own)
object
-
Flickr, Metafilter or your own Web pages/blog posts
Depth of descriptions more idiosyncratic
Fewer descriptions, but more focused (personal)
Does personal metadata become useful to others?
• Which kinds of objects are best for the two different
kinds of folksonomies?
Social Folksonomies
•
•
•
•
Can there be any other kind?
How long can sharing last?
Can people disagree?
Is all tagging considered building a
folksonomy?
• Does the ad hoc nature help folksonomies
grow or hurt their widespread use?
• How would you use a folksonomy?
- Do you use one?
- Do you tag things?
Order out of Chaos
• “We used to rely on philosophers to put the world in
order. Now we’ve got information architects. But
they’re not doing the work - we are.”
- IA is throwing the party, but you have to show up & build the
barn.
- Is this and end-run around experts?
• Google as folksonomy user?
• Machines that automate classification & social
software that makes us willing
• What about personal, automatically-assigned
metadata?
• Would you rely on a folksonomy first to find
something?
• How useful is it to keep humans involved in the
classification?
Cooperative Classification?
• User-generated metadata
- Users working for themselves or other users
• Experts do a fine job of classification
There’s not enough experts
Experts disagree
There’s too much information
How fast can experts learn to describe things for
beginners?
- Users not part of the process
- Users have to learn the experts vocabulary
-
Cooperative Classification (2)
• Make object classification tools easier for experts
- Verify rather than create classifications?
• Include metadata at object creation by authors
• Let users of objects create & share metadata
- Show popular tags as a way to show value of the system (&
sharing)
- Teach people the tagging vocabulary
• People are using tags in clever ways
-
Tags as verbs: post & buy
Tags as prompting: shop & read
Tags as votes
Me, happy, cute, sometaithurts
• Should experts learn from these vocabularies too?
Cooperative Classification Issues
• Wide varieties of tag terms (“filtering)
• Acronyms
• Spaces & Multiple Words
- Spaces_multiple_words
• Case
• Synonyms
• Face value tags vs. abstract ideas
Why Folksonomies Work
• Almost no barrier to entry
• Low cognitive costs (most of the time)
• Supports browsing and searching
- But mostly browsing
• Feedback
- Learning
• Sharing & Community building
• Can you think of other reasons why
folksonomies work?
Ontology is Overrated
• We don’t understand categorization very well
• Digital objects turn regular classification on its side
• What kind of “ontology” are we talking about?
- Canonical, never changing?
- Philosophical, AI, Semantic Web?
• Categorization has its costs
- Managing costs (shelf space)
- Managing time (experts & users)
• Aliases “@” make us think about what is real
- My perspective is real, of course!
- Are we tired of experts telling us what to do & leaning
towards being self-centered?
• So we need the “Shelf” back?
- Metaphorically, we just might
- Maybe we just need a new, better metaphor?
Overrated (again)
• We need adaptive, multi-faceted classification
- Digital objects defy structure (& tradition)
- I mean lots of facets
• Just in Time use of classification is a Google
search
- But it can’t always be right… can it?
• Who can read our minds but ourselves?
- Again, is this selfish?
- Yes, and lots of people are getting rich on that.
• “The Only Group that can categorize
everything is everybody”
- Links everywhere with URIs for everything
Tagging Advantages
•
•
•
•
•
Market Logic
Users & Time are Core Attributes
Signal Loss from Expression
Post Hoc Filtering
Merged from URLs, not categories
- Bottom up classification
• Merges are probabilistic, not binary
• Will “Organic Organization” win?
Download