Updated September 7, 2010Doc Sheldon
The Wizard of Moz, Rand Fishkin, posted yesterday on the topic of Latent Dirichlet Allocation (LDA), on the SEOmoz blog. It’s stirred a lot of conversation, which was probably at least part of the intent. Most of the comments I’ve seen fall clearly on either the “Cool!”, or the “Gimme a freakin’ BREAK!” side, with relatively few taking the position of “Could be handy… I’m not sure.”
Of course, in the SEO – SEM community, you don’t encounter a lot of “not sure” opinions. Theories abound, and are often passionately defended. It’s been my observation that the level of passion is often inversely proportional to the amount of supporting evidence. That isn’t true of all, of course. The people whose opinions I most respect are usually those that are the most objective and open-minded. One such is Andy Beard. He responded with some observations on the SEOmoz LDA tool, on his blog, SEOmoz LDA Tool – Just 3 Points, in his typically fair manner. It addresses some concerns that I share, and makes one think, as it should.
My Sensei at the SEO Dojo, David Harry, is another person whose writing I tend to give a second read. His post, The New LSI for 3rd Generation SEOs? makes some excellent points. And given the almost nuclear passion that David holds on the topic of search (and particularly anything even remotely related to LSI), I gotta tip my hat… he presented a fair, even-handed analysis, without using the term “bullshit” even once.
Must have been tempting, though.
My opinion, at this point, is that the moz LDA tool might give us some help in discovering a relevancy problem, but it falls short of the more subtle implications in Rand’s article. I also share the concern with others, that this may simply turn into more “snake oil” for wannabe SEOs to peddle.