Does Microsoft’s Bing have Google running scared?
There was some interesting commentary from some of my Microsoft buddies (names removed for anonymity) about this article and Bing vs. Google.
This article would be true in only one case – if Live Search team somehow managed to beat Google on NDCG (normalized discounted cumulative gain – a measure of relevance). Historically they could not, particularly on longer queries. If they pulled this off, that would be quite an achievement, since LS algorithms are purely machine learned, and Google reportedly relies on hand tweaked algorithms. This could mean that GOOG is running out of relevance headroom with their approach, whereas Live could take full advantage of the Moore’s law and deploy more sophisticated ML algorithms over time.
That said, I doubt this is the case. Historically, Google has been cranking relevance up a notch every time YHOO or MSFT would deploy improved algorithms, just enough to beat them both by a narrow margin. 🙂
Looks like a slow news day at CNet.
Relevance is important and i’m sure the bing guys will keep making it better. but the “blue ocean strategy” microsoft adopted was to come at it from other interesting angles as well that google was oblivious about.. the user experience is SO much better with bing..
IMO the only big win that I can see is snippet expansion (this is why Bing is currently my default search engine). This, however, is something that would take Google maybe a couple of weeks to replicate if they wanted to.
Here’s why this is a great feature. The fundamental problem in search is that data labelers (i.e. the guys who label training / evaluation data) make a determination about relevance based on the content of the document, but customer satisfaction will mostly depend on the quality of the snippets, since that’s what the user perceives as the main relevance characteristic. Good snippets are hard to generate, since you don’t really want to go into the hairy (and AI complete, if you want to do it well) topic of document summarization. So what you end up having is good relevance and shitty customer sat, since the users can not determine if the page is relevant or not by looking at search results.
Bing’s new UI sidesteps the problem by providing more snippet text on hover, which is very smart. Of course, I’d prefer to have an intelligently generated snippet, but at this point I’ll take whatever I can get.
D: if you can be converted, then half the battle has been won already 🙂 the others will be much easier to convert..
I’m actually fairly easy to “convert”, since I’ve seen it all, from Windows all the way to Linux and everywhere in between and oftentimes know it for a fact that the grass is _not_ greener on the other side of the fence.
Just don’t be worse than the competition, and I’ll use the product every time I can. Win 7, Office 2007, SQL 2008 (including Analysis Services), Exchange, Bing, Visual Studio, .NET Runtime and Framework – those are just a few Microsoft products that have my complete approval.
That said, I always was very much anti-Koolaid, and while I will be the first to praise folks for a superior product, Koolaid by itself has no effect on me whatsoever.
So don’t expect me to become a fan of Zune, WinMo or MSN (outside of Bing) anytime soon. A lot of work needs to be done there.
Outside of Microsoft this applies to Apple TV, Amazon Kindle and all other eBook readers, single-button mouse (though not to single or zero-button trackpad – those are convenient), iPod Shuffle, Java (and 90%+ of open-source Java frameworks), plastic MacBook, all smartphones other than iPhone and Pre, etc., etc.