Morphology, analogy and machine translation

Pirrelli, V

Abstract

following straightforward questions: "What domain of linguistic knowledge is dealt with
in this work?", "How are these phenomena approached from a formal standpoint?" and
"What for?". They thus define the three main axes around which the present
interdisciplinary investigation revolves: main topic, methodology and intended purpose or
range of application. Topic apart, methodology and purpose are strongly interconnected,
in much the same way as the form of an object is related to its function. So, the present
research is orientated towards the representation of morphological phenomena as rather
abstract formal linguistic objects, that are expected to function as a kind of exchange
format aimed at interfacing two or more languages in a "transfer-based" Machine
Translation System. As to the nature and content of these representations, many
alternative proposals, put forward in both theoretical and computational linguistic circles,
are carefully considered and extensively discussed. In particular, I take advantage of
insights coming from "Word and Paradigm Morphology" and "Functional Grammar". In
the end, an original computational framework is presented in somewhat detail, that
crucially hinges on the notion of linguistic Analogy, as an effective, formal procedure for
extending linguistic generalizations from known cases to unknown ones. Within this
framework the set of morphological rules and the Lexicon are not implemented as
separate grammatical compartments, as in most computational models I know of, but they
are really part and parcel of the same self-modelling network of lexical redundancies. This
move makes the whole computational machinery efficient and cost-effective, while
providing a convenient and elegant solution to a number of non-trivial theoretical
paradoxes raised in the relevant literature. Such a model has already been subjected to
the test of a computational implementation, and some results of its application to aspects
of Italian Morphology are detailed in the final part of this work.

Citation

Pirrelli, V. Morphology, analogy and machine translation. (Thesis). University of Salford

• Thesis Type: Thesis
• Deposit Date: Jul 23, 2021
• Award Date: Jan 1, 1993