-
Home page
-
Contact
|
|
A Short History of HPSG in Tübingen
The HPSG-projects at the Linguistics Department began at the beginning
of 1992 with project B4, Constraints on Grammar for Efficient
Generation, by Erhard Hinrichs and Dale Gerdemann in the former Special
Research Program (SFB) 340. This project began with the intention of
exploring the problems of language generation from the paradigm of
head-driven processing. With the fortunate addition of Paul
King and collaborator Guido Minnen to our team, the project quickly
turned its attention to the examination of the logical foundations HPSG
as the most promising grammar formalism. This research also focused
on the creation of a suitable implementation platform for HPSG
grammars (which would later lead to the development of Troll and from
there, the final development of the ConTroll system), and the specification
and implementation of suitable grammar fragments and example analyses,
particularly in German, to collect experience with linguistic theory
formation and its computational application.
Parallel to these developments, Tilman Höhle offered an
introductory seminar based on HPSG to Syntax, in the German department,
during the summer of 1992. These introductory seminars had previously
taken Generalized Phrase Structure Grammar or the Government and
Binding Framework in the Chomskian tradition as their starting
point. This was a particularly fortunate time, as Carl Pollard, guest
professor of the B4 project, was then available and willing to take
over the introductory courses and report first hand on the latest
developments in research. In the same semester, Paul King held a
seminar on the basics of HPSG, which covered the topic of Speciate
Re-entrant Logic, discussed in his 1989 dissertation. In this course,
he expanded the aspects discussed in his dissertation by discussing
the meaning of grammars from a model theoretical perspective. Since
then, we have incorporated an introduction to HPSG, advanced courses
in the various problems of HPSG application to individual linguistic
phenomena, as well as advanced courses in HPSG logical foundations
and/or grammar implementation into the course programs of our
Linguistics Department.
The second phase of the HPSG-projects of the SFB 340, during the time
from 1995 to 1997 brought a new direction: there were now two, instead
of one, HPSG projects. Under its new name, From Constraints to Rules:
Efficient Compilation of HPSG Grammars, the objective of B4 was then
to establish an efficient implementation platform in the evolution of
the Troll-System, developed during the first term. It was intended
that the functionality of the new Troll system be oriented toward the
logical foundations of HPSG and at the same time linguistically
motivated grammars of theoretical interest. The principle task of the
new co-project B8, An HPSG-Syntax fragment for German was to be the
generation of an empirically and theoretically interesting syntactic
and semantic fragment of German, done in cooperation with a semantic
SFB-project, based at the University of Stuttgart. The objectives of
the fragment project were twofold. One objective was integrating, as
much as possible, all the findings of the SFB into one extensive,
consistent entirety. Additionally, the fragment project was also to
lay the groundwork for Project B4, by preparing concrete grammars of
German for B4's implementation platform. At the same time as these
endeavors, the projects of the Verbmobil consortium that were located
in Tübingen started to use HPSG. Within several reading groups, which
were often enhanced by visiting academics, exchange students and
Ph.D. candidates of the graduate college Integriertes
Linguistikstudium, a lively, productive environment emerged. These
discussions increasingly went beyond the set topics of the SFB 340
projects to address questions of empirical and theoretical linguistics
on the basis of the HPSG framework.
In the last stage of the SFB 340, which began in 1998, projects B4 and
B8 were combined into a single, bigger project. There was a switchover
from the discontinued ConTroll system to TRALE. Gerald Penn's
experiences with ALE inspired its merger with ConTroll to form
TRALE. This marriage of the systems gave us the best of both worlds. While
ConTroll was emphatically characterized by our orientation toward
theoretical linguistics, ALE availed to us efficient methods of
parsing. To re-enforce the objectives of the project, the syntax
fragment of German was re-written for TRALE and expanded by further
syntactic phenomena, which was prompted by theoretical linguistics.
During this period many dissertations were completed by many
colleagues within the SFB as well as by researchers related to the SFB
project. Against the backdrop of the collaboration of Paul King and
Kiril Simov, of the Bulgarian Academy of Sciences in Sofia, who
initiated the international doctoral program of CLaRK, Anna Kupsc and
Gergana Popova as well as others arrived to work as the next
generation of researchers on HPSG analyses of a number of languages.
By the time we closed the HPSG project of the SFB 340 at the end of 2000,
we had developed a collaborative research exchange with the University
of Essex, comparing and contrasting HPSG and Lexical Functional
Grammar. Within the scope of this collaboration, we organized six
international workshops and offered a two week ESSLLI summer school
course. An additional post SFB HPSG project was a collaboration with
the BulTreeBank-Project at the Bulgarian Academy of Sciences, on the
topic of the creation of an HPSG-treebank of Bulgarian. Two new,
extensive research projects were begun, namely the MiLCA project
Grammar Formalisms and Parsing and the A5-Project Distributional
Idiosyncrasies of the new Tübingen SFB 441, in the autumn of 2001 and
January 2002.
For the next exciting episode of HPSG in Tübingen, our webpages of
running projects will keep you posted. |