Bing’s Frédéric Dubut said on Twitter and in a quote at Search Engine Journal that machine learning helps with scaling generalizing, which is how humans would manually rank search results. So using machine learning to rank search results is a good way to get Bing’s search results to look more human tailored and thus better.

Frédéric Dubut said:

A major goal of our Bing ranking team is to build an algorithm that would rank documents in the same order as humans would as they are following the guidelines. You can only do so at the scale of the web by generalizing your ranking algorithm as much as possible.

It turns out that modern machine learning is very good at generalizing, so you can expect our core ranking algorithm to get closer to that ideal Intelligent Search product view that we hold internally and which we try to capture in our own guidelines.

I am sure a ton of you will love to chew on this quote for some time.

Forum discussion at Twitter.

Original Source