Nonconcatenative finite-state morphology

Abstract

this paper to explore the formal properties of finite-state transducers, a brief excursion may be in order at this point to forestall a possible objection to the claim that a parallel configuration of transducers can be combined into a single one. On the face of it, this cannot generally be so because there is generally no finite-state transducer that will accept the intersection of the sets of tape pairs accepted by an arbitrary set of transducers. It is, for example, easy to design a transducer that will map a string of x's onto the same number of y's followed by an arbitrary number of z's. It is equally easy to design one that maps a string of x's onto the same number of z's preceded by an arbitrary number of x's. The intersection of these two sets contains just those pairs with some number of x's on one tape, and that same number of y's followed by the same number of z's on the other tape. The set of second tapes therefore contains a context-free language which it is clearly not within the power of any finite-state device to generate

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 141 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.