Rosalind - 10:31 pm on Jun 26, 2006 (gmt 0)

Are you arguing that there is no spam other than junk pages?

A good example of when spam pages are not necessarily junk is typospam. I've seen instances when the page lists a long, long paragraph of common misspellings of a search term at the bottom. This could be used on a page with otherwise good content. Typos could also be introduced subtly into alt text and so forth. So in theory at least, spam is not junk.

However you're far more likely to come across spammy techniques like this on a site of pure junk. This is exactly what I found at the bottom of a cloaked, keyword-stuffed page which using multiple subdomains and random nonsense text, the other day. I also find a lot of the referrer spam I find in my logs is on domain names that are obvious typos of common words.

What I want to know is, how feasible is it to put a spellchecker in your algo, to search for text with more than a certain % of typos?