Neural networks used to improve relevance in web search


Did you know that the most recent version of the MSN Search engine uses neural networks to improve search relevance? The ultimate goal is to have near-perfect relevance. The algorithms were developed in collaboration with Microsoft Research, with help from Chris Burges and others.


I used to program neural networks in college, but I haven’t read anything about them in the last years. I believe that it’s a fantastic research area, for example in speech recognition, or in OCR software. Neural networks shine especially in classification problems when there is a high level of incertitude (or noise) in the input data. However, I never thought to apply this thing to a search engine. (Hmm.. maybe I should pick up a modern course in neural networks and bring myself up to date).

Skip to main content