SEO – New Takes on Topic Modeling

by Karl Tatgenhorst on September 7, 2010

As a nerd, I am attracted to science, science with lots of words and maybe some graphs. SEOMoz has delivered spectacularly this week with an article on LDA (latent dirichlet allocation). Apparently, this topic is somewhat controversial in SEO circles, but the evidence seems compelling. SEOMoz did a great job of demonstrating the strong corollary evidence of LDAs affecting Google’s rankings.

The evidence that was shown related to topic modeling using LDAs. Topic modeling helps a search engine determine “relevance” of a result. Some cases are very easy to understand, such as a query word being present in one result versus not in another. Other cases were more illuminating, like co-occurrence. The example for co-occurrence was a search for Superman, the engine knows that “Clark Kent”, “daily planet” often occur along with that term and thus ranks them higher. Several other cases are explained and I don’t want to rephrase the whole article, I just wanted to draw attention to it. Topic modeling isn’t actually new to SEO, so why would SEOMoz do this post now? Oh, they have a new LDALabs tool! Will you be trying it out?

About the author

Karl Tatgenhorst wrote 31 articles on this blog.

  • http://www.ra-kotz.de/urheberrecht2.htm Irreks

    Heya all, could somebody if possible help me out? I arrived here the other day by way of yahoo but now can not find again the post I have been reading that time :(. And before somebody asks, no I am not blond simply not that excellent with computer systems hehe.
    Kiss :)

    In those who would need it, in my signature there is some interesting information about
    copyrightwatch.

  • http://macwholesale-makeup.weebly.com/#64543 guleinmac

    to images, forms, and so on., for you personally to reduce and paste for your net internet site ,karen millen dresses uk

Previous post:

Next post: