Chad Fowler (www.chadfowler.com)
got me interested in searching when he published an article on Latent
Semantic Indexing (here).
Between us we dinged around with the simple part (building the vectors and
calculating cosine differences between articles). The interesting part,
using Singular Value Decomposition to cluster articles into
What’s Related groups is currently beyond my rather pathetic
math (somehow I seem to have managed to have either skipped or forgotten
any classes on matrices). However, in the meantime it turns out to be
rather easy to use the word vectors to implement straightforward searches
of the blog. So, while I’m waiting for someone who knows what
they’d doing to explain the SVD computation, I’ve made the
simple search facility available as a plug-in for the sidebar.




Comments