tedster - 6:06 pm on Sep 1, 2007 (gmt 0)

Is google a semantic search engine now?

Google has been folding in various kinds of semantic indexing since they purchased Applied Semantics in 2003 [webmasterworld.com]. The effects of semantic factors in the SERPs are becoming more and more important to understand IMO -- at least in a broad, top-level way. One of the Information Retrieval (IR) approaches that uses semantics is co-occurence [webmasterworld.com].

In a pure text-match algorithm, when you want to rank for a given keyword, you focus on including just those words in your copy and anchor text. From a semantic point of view, this approach can generate an unnatural page -- because it does not include any of the related words that would naturally be expected to "co-occur" with the target phrase. For example, if you want to rank for "garbage collection", then with a co-occurence algo you would also expect to see on the page at least some words like "truck, waste, refuse, route, sanitation, regulations, recycling, compactor" and so on.

In some cases I've worked with, significant ranking improvements came after we took the restraints off the copy and folded in more of the naturally related vocabulary. It's something to think about, and it's a lot less risky than removing the actual words you want to rank for.

Caution - don't overdo this. One of the signs of an automated scraper page is that, by collecting snippets from many different sites, they can end up with TOO MANY different co-occuring words. I'm pretty sure that Google looks for this unnatural signal - high co-occurence - as much as it looks for zero co-occurence. So this approach is really a "non-technique" technique, and there's a big fat sweet spot.

In basic terms, you might just say "write good copy that communicates effectively to your users."