Forum Moderators: open
These both both make me think about how engines might be theming sites, which is to say if they find a pattern associated with a basic search that is repeated over several other sites, they'll give it a subtheme. Could it be that we might describe themes, then, as sets of commonly associated search terms?
Also, WISEnut and Vivisimo categories can probably be used as a targeting tool, in the same way, say, as entering a one or two word search in the GoTo Suggestion Tool to come up with a broader list of searches with more modifiers tacked on. I'd guess that WISEnut and Vivisimo get their subcategories from internal databases that look pretty much like GoTo lists.
The difference, of course, is that the engines are showing what's been targeted, whereas the GoTo tool is showing what's been searched.
For "theming" in these subcategories, a question might be which these engines are looking at... what's been searched, or what sites they have listed? I'd guess the latter, but it's just an uneducated guess.
Yep, that is exactly what it is Robert. Think back to the SAT's:
ship is to sail, what fender is to car, what stake is to wood, as steak is to cow, as ...
The tricky part is when words have duplicate meanings. You can have a "log" file and you can have a "log" of wood.
By using the actual searches people perform, you can build up a database in a rapid amount of time on how words relate to one another (called a Term Vector Database).
I think you are very correct at looking at WiseNut's category search and Viivisimo's too. I think they are powerful tools to help you understand your sites "keyword" slot in life.
They are also great tools at helping you find things on your own site. Like this post. I just found it from Wisenut.
For example: keyword marketing [wisenut.com] shows categories for 'search engine' and 'meta tags' which from a lexical standpoint are not related
Also: Betsy Ross [wisenut.com] shows 'United States' and 'flag and banner' which are also not related to keywords used.
More testing to find out if its a combination of words used, lexical similarities, and sites listed.
One question this did bring to mind is, what are the little white plus boxes next to the category?? They are not always the largest number in the group, but usually.
To answer your question, the little plus boxes are like the expand directory boxes in Windows Explorer. They in fact display a little text flag, "expand subcategories," when you mouse over them.
I think these category listings all provide clues to how the engines theme sites. Sort of like grouping things in real life... if there are enough of them that match the same recognizable pattern, it perhaps makes sense to categorize them. We have Teoma topic groupings too...
Just noticed something on Teoma that I thought might be worth noting here... A group of sites I've been monitoring now offers up a topic grouping in Teoma for "Fireworks Splice Html." This phrase is an artifact that a sloppy developer left in the meta descriptions of several related sites, on enough pages that gets clustered as a sub-theme all its own.
This is a pretty clear illustration of something that's perhaps obvious to everyone by now... that the groupings don't have to be driven by meaning, just by a set of relationships that occur often enough that the algo notices. In effect this clustering is a kind of lateral search of what's existing on the sites. I suppose meaning could be programmed into it, but I'm not seeing it here.
