WHO: András Kornai
WHAT: Lexical semantics by machines
WHEN: Friday 14 October, 1:00PM-2:30PM
WHERE: 32-D831
Abstract:
A simple information-theoretic argument shows that about 90% of the
information in a sentence is carried by the choice of words, and only
10% depends on function-argument structure. In light of this, it is
somewhat surprising that 90% of formal semantics (basically, all of
Montague Grammar and its descendants) deals with the compositional
aspects of meaning, with lexical semantics relegated to the fringes of
formal work. In this talk we discuss how a classical piece of algebra,
Eilenberg’s (1974) theory of machines, can be used to provide a
nontrivial formal theory of word meaning, and how the resulting model
interfaces with compositional tecto- and phenogrammar.
References: (* = suggested reading)
S. Eilenberg (1974) Automata, Languages, and Machines. Academic Press
[This book is excellent background for those interested in the
algebraic theory of finite state automata, transducers, and machines
that this work puts to use for semantics.]
*A. Kornai (2010a) The treatment of ordinary quantification in English
proper. Hungarian Review of Philosophy 2010 54/4 150-162
[This is a summary of what is wrong, in this author’s view, with MG and
related theories of formal semantics. Formal issues are discussed,
but no heavy machinery is used.]
*A. Kornai (2010b) The algebra of lexical semantics. In C. Ebert, G.
Jaeger, J. Michaelis (eds) Proc. 11th Mathematics of Language workshop
(MOL11) Springer LNCS 6149 174-199
[This is the intro to the new approach.]
*A. Kornai (2011) Eliminating ditransitives. To appear in M. Egg, P. de
Groote, M-J Nederhof, F. Richter (eds) Selected Papers from the 15th
and 16th Formal Grammar Conferences, Springer LNAI, in press.
[One area where the algebraic approach shows some promise.]