At the University of Tübingen, HPSG research has traditionally focused
on foundational work in linguistics, on the mathematics of grammar
formalisms, and on grammar implementation. Over the years, the
central leading research ideas have been most closely associated with
theoretical linguistics. Such a guiding principle stands in opposition
to an approach in HPSG that is predominantly interested in efficient
computational processing. Of course, taking a great interest in
theoretical linguistics doesn't mean that other areas of linguistics
must be excluded, but it does notably influence the form that
linguistic analysis takes, the choice of mathematical-logical tools,
and the style of the grammar implementations developed within the
program. Our approach is based on the expectation that our work will
result in a deeper understanding of basic principles which will, in
turn, contribute to the development of more adequate and more
efficient language processing systems.
HPSG research at the University of Tübingen is currently being
conducted in quite a number of different areas. In the area of
semantics, the focus is the continuing development of Lexical Resource
Semantics (LRS). LRS is a semantics formalism originally specified in
the feature logic of HPSG. It serves to integrate classic
linguistic ideas of semantic meaning representation in the tradition
of Richard Montague with techniques of underspecified semantics. The
purpose is to obtain meaning assignments to linguistic signs that are
both empirically adequate and computationally promising. Our goal is
then the phenomenon-oriented extension of existing semantic fragments
of various languages, as well as the further development of the LRS
module of our grammar implementation platform TRALE. The DFG funded
German research network
CoGETI
is a platform for some aspects of this research.
The TRALE system, which is used for the implementation of various
sized HPSG fragments in both research and the classroom, is
continuously being developed in collaboration with Gerald Penn of the
University of Toronto. These efforts are directed towards attaining
grammar implementations which are as true to linguistic theory as
possible with the greatest possible processing efficiency. The effort
made to implement grammars, as far as possible, in the form that is
logically specified in theoretical linguistics, is last but not least
connected to our strong emphasis on e-Learning, which has stood at the
forefront in the last few years for the development of teaching
materials and educational software.
In connection with e-Learning, we have created comprehensive course
materials that closely combine questions concerning the logical
principles of HPSG, the linguistic basis of the mathematical concepts
used there, the computational implementation of purely logically
specified grammars, and genuine linguistic analysis problems. Using
the key concept of feature structures as an interface between these
various research areas, the main goal is to help students obtain a
deeper understanding of the interdisciplinary character of HPSG,
thereby encouraging a creative approach towards the various components
of HPSG as a linguistic theory. This is done by using appropriate
software to help students visualize, in an interactive as well as
graphic way, many aspects of the logical specifications of
grammars. Our course materials and software pertaining to grammar
formalisms and parsing are being actively perfected and are enjoying
increasing integration into current research areas.
An interface between corpus linguistic methods, grammar development
and current empirical problems can be found in the SFB project
entitled
Distributional Idiosyncrasies: Corpus Linguistic
Acquisition and Grammar Theoretic Analysis. This project originally
began as research concerning 'cranberry' words, a sub-category of idiom
and phraseology research. This research project has since been
expanded to include the analysis of negative polarity items as an
additional form of idiosyncratic context restrictions on lexical
elements. The semantic nature of this phenomenon allows for
interaction between this research field and the area of LRS, as well
as grammar implementation in TRALE.
In addition to this necessarily brief overview of some of our
research topics, HPSG research interests in Tübingen also
include a series of further topics, some of which have appeared in
recent publications. If you are interested, please refer to
the
Tübingen HPSG Online Bibliography, where you can find
all of the most recent publications.
