When running a search, regular expressions are generated with alternatives based on the given tokens. At a basic level, alternatives include equivalent forms of the same token. When uncertainty is greater than 0, similar tokens are also added as alternatives.
The similar tokens are defined by a similarity graph. The set of tokens returned as being similar are all those within a distance equal to the uncertainty from the target node when performing a breadth-first-search.
The similarity graph is shown below.
