Send us a comment.
3.1 Recursive transition networks 3.2 Modelling recursion in English grammar 3.3 Traversing RTNs 3.4 RTNs in LISP 3.5 Pushdown transducers 3.6 Advantages and limitations of RTNs 3.7 Augmented transition networks 3.8 Developing ATNs 3.9 Implementation of ATNs 3.10 Some reflections on ATNs
We saw in the last chapter how FSTNs and FSTs can capture patterns of language and be used for responding to and producing language. We also implied, however, that FSTNs cannot capture, or cannot capture elegantly, some aspects of natural languages that we would like to be able to handle in NLP. For instance, FSTNs are not mathematically adequate for the description of certain kinds of embedded structures. There are also questions of notational adequacy, to which we return in this chapter, as we investigate how the network metaphor can be taken further to allow for the description of recursive structures in natural languages. Recursive transition networks and pushdown transducers are introduced as the recursive analogue of FSTNs and FSTs. The development of network-based machines for language processing has culminated in the design of various kinds of augmented transition networks, which essentially provide us with specialized programming languages for writing language processing applications.
Send us a comment.