Tagging v2: Advanced machine categorization, Manual categorization, Attribute Tags

[GoogleCodeExporter]

Ginny mentioned the idea of categorization.

Roughly, there's two techniques:
1. machine categorization
   1a) keyword-based
   1b) more sophisticated models
2. manual categorization (aka labels)
   2a) fixed category names
   2b) flexible category names

then for search:
 - use these as the FPXML category names.
 - index the category names for keyword search
 - provide restricts by category (must-have, must-not-have) in the API
 - expose to consumers in the UI (and gadget?)

for v1, I recommend 1a followed by 2a, since we don't have the staffing for
anything more sophisticated.  For the consumer UI, keep it simple e.g. add
as a set of restrict-links for each listing.

for the actual categories, I recommend names that:
 - users want to search on but which aren't simple keywords
 - are reasonably non-subjective  (ok to say: when in doubt, tag it)
 - can be seeded (started) with keywords, e.g. not completely abstract
   concepts.
e.g. "fun" could be a category meaning that you'll have fun doing this
activity (e.g. playground build or reading to a child, not a suicide
hotline or blood donation), and we seed it with a bunch of keyword searches
and hand-tagging.
e.g. "skill building" could be a category meaning that you'll learn a
valuable skill (e.g. habitat for humanity and not soup kitchen), and we can
easily seed them with keywords on various skills. (though could be tricky
if the skills are pre-requisites)

Original issue reported on code.google.com by adam.sah on 30 Jun 2009 at 11:13

[GoogleCodeExporter]

Original comment by adam.sah on 30 Jun 2009 at 11:14

• Added labels: Milestone-Roadmap
• Removed labels: Milestone-TBD

[GoogleCodeExporter]

Original comment by ehysen on 6 Jul 2009 at 7:12

[GoogleCodeExporter]

Version one is in 1.7.  I'm taking myself off as owner since I'm not sure if 
I'll be
working on the next major features, but here are a couple of ways tagging could
evolve in v2 (or v3 or future versions):

* Analyze categoryTags and audienceTags coming in from providers and build a 
proposal
for new tags and how we should factor tags from providers in our tags.
* Build more advanced machine categorization taggers beyond the keyword method,
potentially using the data from providers.
* Add capability for direct tags from partners (we'd read certain tags from 
providers
and directly tag them ourselves) for events like 9/11.
* Develop a plan for "attribute" tags - things like "skill building", "fun", or
"good-for-seniors".
* As an extension of the attribute tags, look at using tags for skill-based
volunteering - figure out what the top required skills are in listings and how 
to
identify them in taggers.

Original comment by ehysen on 21 Aug 2009 at 2:54

• Changed title: Tagging v2: Advanced machine categorization, Manual categorization, Attribute Tags

[GoogleCodeExporter]

+jess -- could be fun + high-leverage 20% projects if you wanted...

Original comment by adam.sah on 21 Aug 2009 at 3:45

[GoogleCodeExporter]

Original comment by danstryk...@gmail.com on 8 Jan 2011 at 1:10

• Changed state: Deferred