allow for cross-comparison of names

[zkamvar]

Joe smith

smith joe

should be compared

[zkamvar]

Elaboration on this:

The current implementation of nameDists() assumes a definite order and compares the strings that way. By contrast, the current implementation of locationDists() does not assume order. Instead, it calculates a distance between all combination of words and takes the lowest value.

Upon thinking about the way names can be implemented, it's not all that different than the way locations are implemented. You can have multiple names for one person, they could have different spellings, and they could be in different orders, so locationDists() would be a generalized nameDists()

Unfortunately, there is a weird problem in the way locationDists() matches things. It would match disparate things like the state of Georgia and the country of Georgia with a distance of 0.

I realized that a way to solve this is to create a matrix with these distances, and take the path with the lowest summed value:

For example comparing the name a aa aaa with baa b ba wha, the distance matrix would look like this:

	a	aa	aaa
baa	2	1	1
b	1	2	3
ba	1	1	2
wha	2	2	2

The smallest path (b, ba, baa) has a total distance of 3, and that would be the resulting score. The upside is that we would be able to narrow down the false positives, but the downside is that finding the smallest path is something on the order of n^2.

Thoughts?