Structured Data Wikis
Problem
Defining ontologies for structured wikis. We need to allow people to devise folksonomies from the ground up, but we need to have a regular ontology at the top level so the structured data can be reused.
Discussion
Traditional information architecture affords:
- Searching
- Browsing
- Finding
- Refinding
With respect to structured data, IA also affords:
- Use
- Comparison
Which of these styles of information interaction are best served by free tagging versus structured hierarchy/taxonomy?
How do we collaboratively resolve the conflict between bottom-up folksonomies and top-down controlled ontologies / hierarchies?
- Faceted tagging
- Advanced navigation / filtering (refining browsing)
- Hierarchical tag clustering assembled from social tagging
- "The Wiki Approach"
Plagues any kind of wiki with structured data, e.g. OpenStreetMap.
Possible Solution: a Wikipedia of Tags
a simple wiki page listing ontological conventions adopted by the community (e.g. http://wiki.openstreetmap.org/index.php/Map_Features)
- editable ontology with keys, possible values, human readable semantics, and wiki-like affordances
- seeding editable ontologies with existing top-down taxonomies / hierarchies
- automated clean up post-processing of data entry using rules
- feeds of newly appearing key/value pairs, so people can track what is used
- ask "does this make sense to you?" turn it into a game
- a group of dedicated ontology editors / wardens
- machine readable collaborative ontologies affords autocompletion / dropdowns / other hints in data-entry and search UIs
Possible Solution: Ant Trails
Follow user behaviors as "elephant paths" or "ant trails"... see what people do in the mean, then adopt that as convention.
People tag things in a particular social, local, and/or personal context... we need to preserve that context for the tags to retain meaning.
data mine application of tags based on frequency, colocation
data mine use of tags based on user click paths
- the "utility curve of tagging" (tags used too frequently or too infrequently -- q.v. Consummating)
- allow people to have personal tag namespaces
- how are my friends tagging things? how are people nearby tagging things?
Possible Solution: Tag Equivalence
- offer a way to define tags as equivalent, or part of a tag cluster (e.g. owl:sameAs)
- allow users to decide which of a class of tags they prefer to use
- soundex / metaphone equivalence
- this is one possible solution to the multilingual problem (different word in different languages, but same concept)
- also a solution to the morphology problem ("dog" vs "dogs", "speed limit" vs "speed_limit" vs "speedLimit")
Possible Solution: Magic
- ...
- ...
- step 3: profit!