If you follow anyone on Twitter who does anything with the web, you’ve no doubt seen this float across your feed a few dozen times this week:
An SEO copywriter walks into a bar, grill, pub, public house, Irish, bartender, drinks, beer, wine, liquor…
Oh, how we laughed.
It does highlight one of the key problems of search on the web, though: that it’s fairly dumb. Even with Google’s now-famous pagerank system, a lot of searching must be done by a computer sorting through text: a computer that’s unaware that a bar is the same thing as a pub (made more complicated by a bar only being the same as a pub some of the time).
The SEO solution to this—trying to embed a thesaurus into the copy—is poor for two reasons. First, obviously, it’s ugly and reads poorly to humans, who are perfectly capable of understanding the groups of meaning that drive our languages (and who the sites are nominally for).
Second, however, it’s hitting the problem from the wrong end. The problem here isn’t broken websites, it’s broken search engines. It seems ludicrous even to approach a problem like this by trying to fix 100 million websites one at a time than to fix the handful of important search engines. As hard as natural language processing is, fixing a single point of failure is almost certainly easier than muddying up millions of them.
