Interesting anecdote: an alumni of my university was chatting with me recently about his work at Google, essentially he replaced what he described as a "large, complex machine learning pipeline" with a "simple bloom filter" for displaying product results at the top of a search page if it determines you have searched for a product.
For example, if you search for "iPhone 5S", the filter determines whether to show something like this http://i.imgur.com/Dp3y1Gi.png (not sure if sponsored makes a difference here, possibly a bad example).
Is EE shop sites you would have visited before the search?
> a "simple bloom filter" for displaying product results at the top of a search page if it determines you have searched for a product.
"have searched" is not clear to me. I was expecting Google would just spit out a JSON response with all the search query for page 1 (including sponsored ads). Can you please elaborate on this please?
> Is EE shop sites you would have visited before the search?
No. I think you might be parsing his statement wrong - it's not "have searched" as in searched for previously, it's "have searched for a PRODUCT" as in, your search is for a product (something purchasable in a store, preferably an online store) as opposed to something else.
The Bloom filter figures out if your search is for a product, and if it is, puts that box w/ sponsored links to stores that sell that product up at the top of your results.
For example, if you search for "iPhone 5S", the filter determines whether to show something like this http://i.imgur.com/Dp3y1Gi.png (not sure if sponsored makes a difference here, possibly a bad example).