Algorithm for deriving the dialect vector for one location:

1. select the n geographically nearest other locations
   (n is neighbourhood size)
2. correction vector = sum for all n locations of [ geographical vector ]
   (geographical vector is vector from the current location to one of the n neighbourhood locations)
3. dialect vector = sum for all n locations of [ (geographical vector - correction vector / n) / dialect difference ]

neighbourhood size: 18
neighbourhood size: 40
neighbourhood size: 18	neighbourhood size: 40
neighbourhood size: 18
neighbourhood size: 18