Compositionality and Syntactic Generalizations
Publication date
1993-11-12
Authors
Odijk, J.E.J.M.
Editors
Advisors
Riemsdijk, H.C. van
Landsbergen, S.P.J.
Supervisors
DOI
Document Type
Dissertation
Metadata
Show full item recordCollections
License
Abstract
The central problem of this dissertation is the question how syntactic generalizations
can be adequately captured in a compositional framework.
This problem will be investigated within the controlled M-Grammar formalism.
I will describe how a number of complex syntactic constructions have been dealt
with in this formalism, which has been used in developing the Rosetta machine
translation system. In particular, I will show that these syntactic constructions
have been dealt with in a syntactically adequate manner in a framework which
is compositional in nature, and where consequently the grammar has a strong
semantic bias.
The syntactic generalizations that I am mainly interested in here relate to the
fact that many constructions can be described most adequately by a conglomerate
of construction-independent rules. This construction-independence of syntactic
rules, and the relation between syntax and semantics will be discussed in
detail.
In addition, the research has been carried out in the context of research into machine translation, i.e. application-oriented research. Application-oriented research
differs from purely theoretical research in a number of respects, some of which will be discussed in the thesis.
The general conclusions of this study are (1) that the grammatical formalism
used, controlled M-grammar, supplies, due to its compositional nature, a
firm framework to deal with certain phenomena, especially when they relate fairly
directly to semantics (e.g. predicate-argument relations);(2) that the framework
makes it possible to incorporate analyses in which constructions are created by
a conglomerate of construction-independent rules; (3) that it is possible to incorporate and extend syntactically adequate descriptions based on insights from
theoretical linguistics into this compositional framework in a fairly direct manner.
It will, however, also become clear that many improvements of the framework,
or of specific linguistic analyses within it, are still possible.