Augmented Transition Networks

An ATN [Woods, W. A., ``Transition Network Grammars for Natural Language Analysis'', Communications of the ACM, Oct. 1970] is like a finite state transition network, but is augmented in three ways:

1. Arbitrary tests can be added to the arcs. A test must be satisfied for the arc to be traversed. This allows, for example, tests on agreement of a word and its modifier.

2. Structure-building actions can be added to the arcs. These actions may save information in registers to be used later by the parser, or to build the representation of the meaning of the sentence. Transformations, e.g., active/passive, can also be handled.

3. Phrase names, as well as part-of-speech names, may appear on arcs. This allows a grammar to be called as a subroutine.

The combination of these features gives the ATN the power of a Turing Machine, i.e., it can do anything a computer program can do.