Two-Head Finite-State Acceptors With Translucent Letters

Finite-state acceptors are studied that have two heads that read the input from opposite sides. In addition, a set of translucent letters is associated with each state. It is shown that these two-head automata are strictly more expressive than the model with a single head, but that they still only accept languages that have a semi-linear Parikh image. In fact, we obtain a characterization for the class of linear context-free trace languages in terms of a specific class of two-head finite-state acceptors with translucent letters.