Thurs
May 1 |
Marcus Kracht
UCLA
"Working with Referent Systems"
"
|
For some time I have been developing a theory that allows to do semantic
analysis by means of Referent Systems (due to Kees Vermeulen). The main
idea is that the type hierarchy is flattened, and morphological properties are
used side-by-side with syntactic properties. In principle this should do away
with most of the type fiddling known from categorial grammar. The emphasis has
since shifted (in my case) from developing the theory to developing a program
that implements it. I shall outline the theory and then give a demo of the
program.
|
| 5:00PM
2122 Campbell Hall
| |
|
----------------------------------------
|
----------------------------------------
|
|
| |
Thurs
Apr 24 |
Timothy Tangherlini
UCLA Scandinavian Section
"Toward a Morphological Analyzer for Old
Icelandic: Challenges and Possible Solutions"
|
We propose to develop an Old Icelandic morphological analyzer using the
functional programming language Haskell. Instead of trying to gain
incremental increases in the accuracy of our existing Old Icelandic
morphological analyzer by debugging and refining increasingly complex
Perl code, we propose to develop an analyzer that returns 100% accuracy
for known lemmata, that incorporates a simple mechanism for adding
lemmata, and that allows for quick correction of improperly or
over-generated forms. Given its architecture and implementation, the
morphological analyzer we propose to develop will be easy to debug, will
allow for straight forward additions to the underlying lexical set, and
will be written in a manner that can be easily grasped by non-computer
programmers. A morphological analyzer developed according to these
parameters also allows for a high degree of extensibility, making it
easier for non-programmers to develop accurate morphological analysis
tools for other languages. Some of the main challenges in developing
this analyzer are related to the morphonological complexity of Old
Icelandic. These challenges include umlaut, breaking and syncope. Other,
second order challenges include grammatical disambiguation, as well as
accounting for by-forms and orthographic change.
|
| 5:00PM
2122 Campbell Hall
| |
Thurs
April 17 |
Hans-Martin Gärtner
Zentrum für Allgemeine Sprachwissenschaft, Typologie und Universalienforschung (ZAS), Berlin
"Low Risk Quantifiers"
"
|
Low risk quantifiers (LRQs) are quantifiers for which an opponent has no
superior falsification strategy in a GTS-style verification game. LRQs are
shown to closely approximate the class of DP-quantifiers allowed in a
presentational construction of German involving prosodically and
information-structurally integrated V2 clauses. The notion of "risk" will be
linked to a speaker strategy in competitive argumentation.
|
| 5:00PM
2122 Campbell Hall
| |
Thurs
Apr 3 |
Greg Kobele
Humboldt University
" Ellipsis in minimalist grammars"
|
Abstract: I show how to extend minimalist grammars so as to be able to
assign meanings to elliptical sentences. Adding an operation of
deletion under identity poses two problems. First, what is to count
as identity (syntactic, semantic, etc)? Second, once a notion of
identity is agreed upon, how does this new mechanism affect the sound-
meaning pairs licensed by the grammar? This latter question is
tricky, as deletion under identity makes the structure of derivations
much more complicated. Luckily however, measurably so. I show that
minimalist derivations with deletion under identity can be described
using first order logic extended with a binary deterministic
transitive closure operator. This is an immediate consequence of
various other facts, most interesting among which is that minimalist
derivations without identity can be described using FOL with a *unary*
deterministic transitive closure operator (improving on previous
characterizations in terms of monadic second order logic). The upshot
of all this is that we obtain a logical characterization of the
complexity of the sound meaning correspondence minimalist grammars
with deletion under identity can define, which is independent of the
deletion analysis of ellipsis (i.e. it is true of the `ellipsis as a
pro-form' approach, too). This is cool, because we can use
continuations, or whatever else ensures ship buoyancy, to deal with
ellipsis, secure in the knowledge that we know exactly what we are
claiming about natural language.
|
| 5:00PM
2122 Campbell Hall
| |
|