[atlas-l] back to knuggets

[jer]

>> So is a meta-token equivalent to a phrase? Are factories expected to
>> identify all phrases in a document?
>
> Right, a phrase. This is optional for a Factory, and just a great way
> to add value.  It only has to identify any or as many phrases as it
> can or has resources to.

Also I want to add that I believe there should be something along the  
lines of a "roots" attribute on knuggets.  This attribute would be a  
tab-delimited list of strings, those being single words or phrases,  
and they would also be entirely up to the Factory to decide on.  Any  
single root is intended to be essentially a "category" and must be  
human readable/usable (Finance, Gaming, Iowa Agriculture, etc).  This  
attribute is mostly useful on the doc knuggets (since a single  
document generally has the same roots for all of it's content), but  
could be used on individual knuggets if deemed appropriate by the  
Factory.

A Collector can decide how or if it even wants to use the roots, such  
as providing rank clustering based on overlapping roots.  A Broker  
can use this for end-user display as well.

I could see some standard classification schemes arising eventually,  
but again this should be organic and not some top-down-all- 
encompassing effort.

Jer