Presented is an algorithm (for learning a subclass of erasing regular pattern languages) which can be made to run with arbitrarily high probability of success on extended regular languages generated by patterns π of the form x 0 α 1 x 1 ... α m x m for unknown m but known c, from number of examples polynomial in m (and exponential in c), where α 1...α m are variables and where α 1,...,α m are each strings of constants or terminals of length c. This assumes that the algorithm randomly draws samples with natural and plausible assumptions on the distribution.
The more general looking case of extended regular patterns which alternate between a variable and fixed length constant strings, beginning and ending with either a variable or a constant string is similarly handled.