Article I just read about Microsoft's search relevancy:
Microsoft's Google-killer arrives with a 'whuh?'
By Andrew Orlowski in San Francisco
Published Friday 12th November 2004 12:01 GMT
Google's executives might be sleeping a little easier this weekend after Microsoft unveiled its much-hyped new search engine. It's fast, slick, and comes with a raft of interesting new features: confounding some expectations as surely as it confirms others. In short, Microsoft has produced a search engine that's better in almost every way than Google, except for one: its search results are terrible. But let's start with the good stuff.
Incredibly, MSN Beta Search trumps Google for speed: it's an order of magnitude faster. Anyone who doubted that Microsoft could deliver a large scale distributed cluster, and that's probably most of you, will be surprised at the nippy performance (although the true test comes when the system has to scale under heavy loads, of course).
Microsoft has also made building complicated queries much more attractive than its rivals. Click on the "Search Builder" option and you get five additional fields which you can add, one at a time, the fifth being three gauges for altering the search term's topicality, popularity, and semantic accuracy. This puts all its rivals to shame, and makes Google's Advanced Search page look about as appealing as an Assembly Language manual. Microsoft's new engine also has a rough caching service modeled on Google's cache, but without the keywords highlighted in colors: one of Google's most subtle and indelibly useful UI features.
Microsoft has also been busy in other departments. It attempts to produce a natural language answer to something it thinks is a particular kind of question. What's the capital of England? Gives the answer: London, for example. It didn't fare so well with the question "How many mickle in a muckle?", but it's a start.
But MSN Beta Search falls down badly where it really matters: in delivering results with any relevancy. Like Google, it struggles to distinguish between a source query and an effect query. Searching for "John Leyden"+"blaster worm" and "John Lettice"+"Windows" returned a lot of prattle, but hardly any original articles. When a search is so specific, you're reasonably expected to receive source articles, you might think, rather than what people are saying about them. And this illustrates a fundamental blindspot that both search engine designers, and web-happy techno utopians both exhibit: they mistake the web for the world.
