I recall that in Google’s early days they held a competition where they published a corpus of thousands of crawled webpages and invited developers to do somethi

I recall that in Google’s early days they held a competition where they published a corpus of thousands of crawled webpages and invited developers to do something novel with the data. The winner used pattern matching to identify addresses, then geocoded them using the US Census Bureau’s TIGER database. Thus it was possible to associate fairly accurate locations to a page, which enabled local searches and a number of other products. Someone please correct me or fill in details I’ve forgotten.