And this is actually quite a simple approach--which is interesting in and of itself. While there would be diminishing returns, there are a ton of other techniques you could use to make stronger inferences about similarity.