Selected Papers by Michael Schiehlen
2007
- Michael Schiehlen.
``Global Learning of Context Weights from GermaNet'',
in: Proceedings of the Recent Advances in Natural Language Processing, Borovets, September 2007.
PDF
- Michael Schiehlen and Kristina Spranger.
``Global Learning of Labelled Dependency Trees'',
in: Proceedings of the CoNLL Shared Task
Session of EMNLP-CoNLL 2007, Praha, June, 2007.
PDF
.
- Michael Schiehlen.
``Deriving Context Weights from WordNet'',
in: Proceedings of the Workshop on
Lexical-Semantic and Ontological Resources at the Biennial Spring
Conference of the GLDV, Tübingen, April 2007.
talk in PDF
2006
- Michael Schiehlen.
``Categorical Grammar as a Basis for the Semantic Analysis of Coordination and Ellipsis'',
Habilitationsschrift, Stuttgart, September 2006.
AIMS-Report 12, 1.
- Michael Schiehlen and Kristina Spranger.
``Language Independent Probabilistic Context-Free Parsing Bolstered by Machine Learning'',
in: Proceedings of the Tenth Conference on Computational Natural Language Learning, New York City, June, 2006.
PDF,
6 pages.
- Michael Schiehlen and Kristina Spranger.
``The Mass Count Distinction: Acquisition and Disambiguation'',
in: Proceedings of the 5th International Conference on
Language Resources and Evaluation, Genoa, May, 2006.
PDF,
6 pages.
2005
2004
- Michael Schiehlen.
``Annotation Strategies for Probabilistic Parsing in German'',
in: Proceedings of the 20th International Conference
on Computational Linguistics, Geneva, August, 2004.
PDF,
7 pages, demo.
- Michael Schiehlen.
``Optimizing Algorithms for Pronoun Resolution'',
in: Proceedings of the 20th International Conference
on Computational Linguistics, Geneva, August, 2004.
PDF,
7 pages.
- Michael Schiehlen and Kristina Spranger.
``Automatic Methods to Supplement Broad-Coverage Subcategorization Lexicons'',
in: Proceedings of the 4th International Conference on
Language Resources and Evaluation, Lisbon, May, 2004.
PDF,
4 pages, talk in powerpoint
and pdf.
2003
- Michael Schiehlen.
``Combining Deep and Shallow Approaches in Parsing German'',
in: Proceedings of the 41st Annual Meeting of the Association for Computational Linguistics, Sapporo, July, 2003.
gzipped
Postscript, 8 pages, talk in star office and PDF.
- Michael Schiehlen.
``A Cascaded Finite-State Parser for German'',
in: Proceedings of the Research Note Sessions of the 10th
Conference of the European Chapter of the Association for Computational
Linguistics (EACL'03), Budapest, April, 2003.
gzipped
Postscript, 4 pages, talk in powerpoint and pdf.
2002
- Michael Schiehlen.
``Ellipsis Resolution with Scope Underspecification'',
in: Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics, Philadelphia PA, July, 2002.
gzipped Postscript, 8 pages.
The paper presents an approach to ellipsis resolution
in a framework of scope underspecification (Underspecified Discourse
Representation Theory).
It is argued that the approach improves on previous proposals to
integrate ellipsis resolution and scope underspecification
(Crouch 95, Egg, Koller and Niehren 01)
in that application processes like anaphora resolution do not
require full disambiguation
but can work directly on the underspecified representation.
Furthermore it is shown that the approach presented can cope with the
examples discussed by Dalrymple, Shieber and Pereira (1991)
as well as a problem noted recently by Erk and Koller (2001).
- Michael Schiehlen.
``Experiments in German Noun Chunking'',
in: Proceedings of the
19th International Conference
on Computational Linguistics, Taipei, Taiwan, August, 2002. gzipped
Postscript, 7 pages, talk in powerpoint and pdf.
The paper describes a method to process recursive noun phrases with
finite-state cascades. It is shown that chunking of recursive noun
phrases makes necessary a readjustment of the finite-states cascades
approach. In particular, the property of monotonicity must be given up.
Furthermore, the paper explores the influence of POS tags and online
agreement checking on the overall performance.
2001
- Michael Schiehlen.
``Semantik'',
in: Computerlinguistik und Sprachtechnologie,
© Spektrum-Verlag, 2001.
Einführung in Semantik, eine erste ausführlichere Version mit
Programmbeispielen in Prolog (gzipped Postscript, 73 Seiten).
- Peter Krause, Uwe Reyle and Michael Schiehlen.
``Spatial Inferences in a Localization Dialogue'',
in: Semantic and Pragmatic Issues in Discourse and
Dialogue, Experimenting with Current Dynamic Theories,
© Pergamon, October, 2001.
- Michael Schiehlen.
``Syntactic Underspecification'',
in: SFB-Abschlußbericht, IMS Stuttgart, May 2001.
2000
- Michael Schiehlen.
``Semantic Construction'',
in: Verbmobil: Foundations of Speech-to-Speech
Translation,
© Springer-Verlag, July 2000.
This article describes the concepts and the implementation of the
semantic construction module (SemCon) used in the Verbmobil system.
SemCon maps trees to Verbmobil Interface Terms (VITs).
A main focus lies on robustness and underspecification.
A minimalistic syntax-semantics interface is defined to support
modularity. Diverse repair strategies are discussed to enhance robustness.
With SemCon, it is possible to process large amounts of data and
build semantic representations for a good part of the Verbmobil corpus.
- Michael Schiehlen, Johan Bos, and Michael Dorna.
``Verbmobil Interface Terms (VITs)'',
in: Verbmobil: Foundations of Speech-to-Speech
Translation, ©
Springer-Verlag, July 2000.
This article describes the concepts and the contents of Verbmobil
Interface Terms (VITs). In VITs all linguistic information of an
utterance relevant for translation is represented. They are used to
provide an interface representation between several linguistic and
dialog components of the Verbmobil system. Information in VITs is
encoded in a record-like data structure. The fields are
variable-free lists of non-recursive terms, so-called 'flat' set
representations. They are filled with semantic, scopal, sortal,
morpho-syntactic, prosodic, and discourse information. A labelling
system is used to relate different kinds of information to each
other. A library package realizing an abstract data type implements
construction, access, update, check, print, etc. facilities for
VITs.
- Michael Schiehlen.
``Granularity in Tense Translation'',
in: Proceedings of the
18th International Conference on Computational Linguistics, Saarbrücken, July/August, 2000.
gzipped Postscript, 7 pages.
The paper describes a module for tense disambiguation
and focuses on the translation of German present into English.
The approach is semantic in nature and aims at integrating
linguistic insights directly.
It is argued that scope resolution of time adverbials is
essential to the disambiguation task.
The paper shows a way to at least partially disambiguate scope by
resorting to granularity calculations.
The described algorithm has been integrated into the Verbmobil system.
Finally, results are reported for a test run on a sizable set of input data.
- Michael Schiehlen.
``Robust Semantic Construction'',
in: Proceedings of the
18th International Conference on Computational Linguistics, Saarbrücken, July/August, 2000.
gzipped Postscript, 5 pages.
The paper discusses a novel approach to the transformation of tree
banks and output of statistical parsers into semantic representations.
It points out some of the hurdles in this endeavour and shows
solutions. After an overview of the system, the minimal requirements
at the syntax-semantics-interface are stated that parsers serving
the module have to meet. After that, the submodules are explained.
Finally some performance data are discussed.
- Michael Schiehlen.
``Robuste Semantikkonstruktion'',
in: Proceedings of KONVENS 2000 / SPRACHKOMMUNIKATION, Ilmenau, October, 2000.
1999
- Michael Schiehlen.
``Semantikkonstruktion''.
Dissertation an der Universität
Stuttgart (PhD thesis), Februar 1999. AIMS-Report
5, 3.;
gzipped Postscript, 283 pages.
Die Arbeit behandelt Aspekte der Semantikkonstruktion.
Im zweiten Kapitel wird die Aufgabe der Semantikkonstruktion
definiert. Einige wesentlichen Kriterien werden genannt, an denen die
Qualität einer Semantikkonstruktion gemessen werden kann.
Gebräuchlichen Ansätze zur Semantikkonstruktion werden kritisch
diskutiert. Damit wird das Fundament für den später vorgestellten
eigenen Ansatz gelegt.
Das dritte Kapitel ist eine Einführung in die Theorie der
Unterspezifizierten Diskursrepräsentationsstrukturen (UDRT).
Am Ende des Kapitels wird die Auflösung der Skopusambiguität
diskutiert und Wege werden aufgezeigt, wie die benötigte Information
aus Korpora beschafft werden kann.
Das vierte Kapitel stellt den Formalismus der Verbmobil Interface
Terms (VITs) vor. VITs sind UDRT-basierte semantische Repräsentationen,
die im VerbMobil-Projekt als Schnittstellenstrukturen verwendet werden.
Die in diesem Kapitel beschriebenen Festlegungen wurden zusammen mit
Kollegen im VerbMobil-Projekt getroffen. Hier ist besonders die
VIT-Taskforce zu nennen, in der
Johan Bos (Universität des Saarlandes),
Dr. Markus Egg (IBM Heidelberg),
Dr. Walter Kasper (DFKI Saarbrücken) und
Dan Flickinger (CSLI Stanford) mitgearbeitet haben. Wertvolle Anregungen
kamen auch von
Michael Dorna,
Dr. Martin Emele und
Dr. C.J. Rupp
(alle Universität Stuttgart).
Anschließend wird der implementierten Ansatz zur Semantikkonstruktion
beschrieben, zunächst abstrakt (Kapitel 5), dann auf einer
implementierungsnäheren Ebene (Kapitel 6).
Im siebten Kapitel findet sich eine Erweiterung der
UDRT zur Ellipsenauflösung und Parallelismusbestimmung auf
unterspezifizierten Strukturen.
Das achte Kapitel schließlich geht auf Graduierungskonstruktionen
ein. Diese Konstruktionen sind sowohl syntaktisch als auch semantisch
von Interesse. Ihre syntaktisch-morphologischen Idiosynkrasie
ermöglicht es, mit einfachen Mitteln zu reichem Datenmaterial zu
kommen, aus dem die lexikalischen Beschränkungen gelernt werden können.
Das Kapitel stellt eine semantische Theorie für
Graduierungskonstruktionen auf der Grundlage einer Axiomatisierung der
Meßtheorie vor.
Ein kurzer Ausblick beschließt die Arbeit.
1998
- Michael Schiehlen.
``Learning Tense Translation from Bilingual Corpora'',
in:
Proceedings of the
17th International Conference on Computational Linguistics,
Copenhagen, August, 1998.
gzipped Postscript, 5 pages.
This paper studies and evaluates disambiguation strategies for the
translation of tense between German and English, using
a bilingual corpus of appointment scheduling dialogues.
It describes a scheme to detect complex verb predicates
based on verb form subcategorization and grammatical knowledge.
The extracted verb and tense information is presented and the
role of different context factors is discussed.
1997
- Michael Schiehlen.
``Disambiguation of Underspecified Discourse Representation
Structures under Anaphoric Constraints'',
in:
Proceedings of the
Second International Workshop on Computational Semantics,
Tilburg, The Netherlands, January, 1997.
gzipped Postscript, 13 pages.
The paper explores the possibility of performing anaphora resolution
and presupposition accommodation in a situation where scope
relations are only partially known.
For underspecification of scope ambiguity, the approach builds on
Reyle's formalism of Underspecified Discourse Representation Structures
(UDRSs, Reyle 1993) which is extended to cover arbitrary
constraints from anaphora resolution.
Wellformedness conditions are formulated for UDRSs to restrict scope
readings as required by anaphora resolution.
Based on these conditions, a disambiguation algorithm is presented
for scope resolution and presupposition accommodation, which has
also been implemented.
1996
- Michael Schiehlen.
``Semantic Construction from Parse Forests'',
in:
Proceedings of the
16th International Conference on Computational Linguistics,
Copenhagen, August, 1996.
gzipped Postscript, 6 pages.
The paper describes a system which uses packed parser output directly
to build semantic representations. More specifically, the system
takes as input Packed Shared Forests in the sense of Tomita
(1985) and produces packed Underspecified Discourse
Representation Structures.
The algorithm visits every node in the Parse Forest only a bounded
number of times, so that a significant increase in efficiency is
registered for ambiguous sentences.
- Johan Bos, Michael Schiehlen, Markus Egg.
``Definition of the Abstract Semantic Classes for the
Verbmobil, Forschungsprototyp 1.0.'',
Universität des Saarlandes, IBM Heidelberg, Universitaet Stuttgart.
August 1996. 7 Seiten.
Verbmobil-Report 165.
ISSN 1434-8845
1995
- B.Buschbeck-Wolf, M.Butt, M.Damowa, M.Dorna, K.Eberle, M.Emele, R.Nübel, S.Reinhard, B.Ripplinger, M.Schiehlen, H.Schmid.
``Transfer in the Verbmobil Demonstrator.''
IAI Saarbrücken, IBM Deutschland Informationssysteme GmbH, Universität
Stuttgart, E.-K.-Universität Tübingen. August 1995. 133 Seiten.
Verbmobil-Report 147.
ISSN 1434-8845
1993
- Michael Schiehlen.
``Localising Barriers Theory'',
in:
Proceedings of the 6th Conference of the European Chapter of
the Association for Computational Linguistics,
Utrecht, The Netherlands, April 1993.
gzipped Postscript, 7 pages.
In recent years the development of adequate restrictions on
movement phenomena has been at the centre of the stage in
theoretical linguistics. However, computational interest
in such advances has been fairly small, largely because
suitable mechanisms were not on hand. We propose a logical
device to handle constraints on movement. We motivate it
by analysing the approaches of Chomsky (1986) and Rizzi (1990)
in its terms and indicate two ways to use the device in
an incremental parser.
- Michael Schiehlen.
``Eine Axiomatisierung der Skopusbestimmung zur
Constraintbasierten Erzeugung der Logischen Form'',
Diplomarbeit (Master's Thesis), Faculty of Computer Science,
University of Stuttgart, October 1993.
gzipped Postscript, 81 pages.
mike@ims.uni-stuttgart.de / 96-02-07