add semantic labeling w.r.t. counting a subset of letters mod 2

Basic idea: if each rule keeps parity of no. of occ. of letter a, then a : not, others : id is a model, and we can label a by 0 or 1.

Extensions (indepedent)

• also label the other letters
• look for subset A in Sigma such that (a:A) : not is a model
• user other modulus