12

I am building an application that uses both GeoNames and Google Places API. The thing is, when I do a search nearby by a specific location (say lat: 47.16, lng: 27.56) on both of the services I do not know how to remove entities that appear both in the results from Google Places and the results from GeoNames(findNearby). I was thinking about using location (latitude and longitude) but it isn't accurate enough. Also, the name varies considerably so this wouldn't work either. Another idea that crossed my mind would be using the types (feature codes for GeoNames and type for Google Places), but there are a lot of types and obviously I can not do a cross reference manually. Any ideas?

Note: I want to use both of them as this is a school project and the requirements specify using more than one source of info.
Thanks.

scarleth ohara
  • 357
  • 2
  • 12
  • 3
    Two obvious ways: 1) put extra effort in linking names from two services, for example phrase canonicalization 2) compute distance between every place in A to every place in B, make a histogram, the left spike in histogram could correspond to the same places, then pick a threshold and merge places that are of distance less than the threshold. – Vihari Piratla Mar 02 '17 at 10:57

1 Answers1

1

I think that, unfortunately, the reason you haven't received an answer is that there is no answer that would fully satisfy the requirements.

Even with one provider, a single coordinate could be associated with multiple results. Imagine a large building in New York, for example, where dozens of companies each occupying a floor or part of a floor in the same building, and yet they would all be associated with the same (latitude, longitude) coordinate.

Now consider two sources. Source A says there's a doctor at that location (let's say on the 7th floor). Source B says there'S a doctor too. Can we assume they're the same doctor? Nope. It might be another doctor on another floor. Or it could be the same doctor. It's impossible to tell. The point is, you could try to use feature codes / types to reduce the number of hits by assuming similar locations are the same location, but it's still an assumption.

Anyway, good luck with your assignment from 3 years ago. It was a good idea nonetheless. :)

Eric McLachlan
  • 3,132
  • 2
  • 25
  • 37