What is the technology behind bing? Its own version of map-reduce algorithm or something else?

Question

Bing's search hits are quite impressive, has Microsoft not let anyone in on behind the scenes of their search technology? Tried http://www.discoverbing.com but couldn't find the answer to my question.

Answer 1

Microsoft historically has used a Neural Network Ranking Function as its ranking function. The Neural Network combines the hundreds of variables that a Url will have associated with it related to ranking. Paper They would typically score more than 100 docs using a detailed ranker. Each query node will need to score its top documents in isolation and return them to the aggregator. Ranking is actually very complex and scoring algorithms are typically multileveled.

For compute jobs, factor generation etc... Microsoft Search uses SCOPE which I believe is built on top of Dryad but does not use DryadLINQ. SCOPE is basically a SQL language on top of a cluster.

Actually Microsoft is far more open about their technology in search than Google is. Microsoft Research Asia and Microsoft Research Silicon Valley

Answer 2

The is second-hand information, but I understand they use inverted indexes (indices?) for finding the top 100 or so results, and then they use a set of neural networks to narrow it down several times to the top 10, top 3, and then to find the first one.

They do this because they reason the first hit is what makes a user thing the search engine works or not. If you search for CNN and you don't get CNN.com as the first hit, users think the engine doesn't work.

Again, this is second-hand knowledge. I heard this from a friend who worked at MS for a while on their search team.