[XML-DEV Mailing List Archive Home] [By Thread] [By Date] [Recent Entries] [Reply To This Message]

RE: More on Vector Models

• To: xml-dev@l...
• Subject: RE: More on Vector Models
• From: "Bullard, Claude L (Len)" <len.bullard@i...>
• Date: Wed, 4 May 2005 12:33:31 -0500

It's a service.  Given language variants and 
use of apriori knowledge about the event, a 
VSM engine works for any lump of text you 
give it.  Then you leverage what the 
relational DB knows about the metadata.  

VSM is yet another means to sort a bag of 
words and determine its relevance to 
other things in the space.  Because every 
enterprise has lots of word bags, it is a 
trick worth knowing and it isn't hard to 
implement, so if Bill's Guys or Jonathan's 
Guys or Your Guys want to build one and use 
it, it's doable, not exotic.  I'll bet Tim Bray 
can knock it out in an afternoon and put it 
on his Atom feeds the next day, then put a page 
on his server for searching his blogs all 
working inside a week or less.  Heck, there 
is a PERL script out there to do it for free.

VSM comes under the heading of stupid pet 
tricks that work remarkably well.  Old tech, yes, 
but again, that's a good thing.  If you want 
the bleeding edge, you move on to quantum logic 
systems which have a lot of promise but likely 
require quantum computers (from what I read but 
I am no expert).  

What I can say is that when the hardware 
is ready, the mathematicians are already there.

len


From: David Lyon [mailto:david.lyon@c...]

Of course, xml needs a good classification system to
be truly effective in a business context.

but I guess as Ken says, it's all wrapped into the
operating system these days. (Pity though it is an
undocumented feature)

• Prev by Date: Re: XML construction using Java/Perl/C/C++
• Next by Date: Re: InfoWorld Article: SOA meets the real world
• Previous by thread: Re: More on Vector Models
• Next by thread: FW: [soa-rm] SOA RM in the media - Infoworld
• Index(es):
  • Date
  • Thread