Markus Egg

The Constraint Language for Lambda Structures

This paper presents the Constraint Language for Lambda Structures(CLLS), a first-order language for semantic underspecification thatconservatively extends dominance constraints. It is interpreted overlambda structures, tree-like structures that encode λ-terms. Based onCLLS, we present an underspecified, uniform analysis of scope,ellipsis, anaphora, and their interactions. CLLS solves a variablecapturing problem that is omnipresent in scope underspecification andcan be processed efficiently.

Constraints over Lambda-Structures in Semantic Underspecification

We introduce a first-order language for semantic underspecification that we call Constraint Language for Lambda-Structures (CLLS). A λ-structure can be considered as a λ-term up to consistent renaming of bound variables (λ-equality); a constraint of CLLS is an underspecified description of a λ-structure. CLLS solves a capturing problem omnipresent in underspecified scope representations. CLLS features constraints for dominance, lambda binding, parallelism, and anaphoric links. Based on CLLS we present a simple, integrated, and underspecified treatment of scope, parallelism, and anaphora.

Beginning Novels and Finishing Hamburgers: Remarks on the Semantics of to begin

Verbs like begin may take either a VP or an NP complement, but their meaning is pretty similar in both cases, e.g. for begin, the start of an eventuality is at stake. Pustejovskys approach captures this similarity in terms of an invariant meaning of the verb, which entails a process of reinterpretation for the transitive variant of the verb. I will show that while the intuitions of this proposal are on the right track, its actual implementation suffers from a number of shortcomings. I will offer an analysis that preserves Pustejovskys intuition but avoids these shortcomings. My analysis is based on an appropriate underspecification formalism.

Wh-questions in Underspecified Minimal Recursion Semantics

In this paper, I present Underspecified Minimal Recursion Semantics (UMRS), a representation language that represents structural ambiguities in terms of underspecification. It is argued that this kind of approach allows for transparent semantic representations and a straightforward syntax-semantics interface. UMRS is a semantic metalanguage, whose expressions describe expressions of an object language and (possibly underspecified) dependences between them. The potential of UMRS will be illustrated by employing it as the semantic component of an HPSG description of wh-questions

A computational model of natural language communication

The work presented in this book is motivated by the goal of applying linguistic theorybuilding to the concrete needs of potential linguistic applications such as question answering, dialogue systems, andmachine translation. To pursue this goal, a translation of linguistic theory into a framework of “practical linguistics” is suggested. Database Semantics (DBS) is presented as a first step towards such a framework. It models the communication between cognitive agents, which can be used, for example, to implement the communicative abilities of a cognitive robot. DBS serves as a single underlying format for modeling communication in that it lends itself to an account of both language processing and language production (thinking is added as a separate component, which refers to inferencing on stored information, and activating content to be verbalized). As such an underlying format, it can be used to describe linguistic as well as extralinguistic content (to represent utterances and the context, respectively). Being explicitly designed for practical applications, DBS deliberately ignores linguistic phenomena considered irrelevant for these (e.g., quantifier scope). The structure of the book is as follows. It has threemain parts, which introduce DBS, outline the range of constructions covered by DBS so far, and specify fragments that can be processed or produced in the framework of DBS. There is also an appendix with two sections on the treatment of word-order variation in DBS and on the global architecture of DBS systems, and a glossary. The first part of the book starts with general principles of linguistic analysis that apply to DBS. These principles include incrementality (input is to be processed successively as it comes in, which yields an analysis for incomplete as well as complete chunks of input; the syntactic basis for this strategy is Left-Associative Grammar [Hausser 1992]), surface orientation (no empty categories), and a focus on communication (description formalisms must be able to handle turn-taking, i.e., language processing and production). After a sketch of the general theory of communication of which DBS is a part, DBS is presented in detail. It is implemented as a non-recursive data structure, that is, a list of feature structures called proplets (usually, one per word1) that are linked by coindexing the values of specific features.2 For example, subcategorizing elements (“functors”) have features whose values indicate their arguments and the other way around. In spite of its name, DBS does not offer a purely semantic representation of linguistic expressions. Although it does abstract away from purely syntactic phenomena suchThe work presented in this book is motivated by the goal of applying linguistic theorybuilding to the concrete needs of potential linguistic applications such as question answering, dialogue systems, andmachine translation. To pursue this goal, a translation of linguistic theory into a framework of “practical linguistics” is suggested. Database Semantics (DBS) is presented as a first step towards such a framework. It models the communication between cognitive agents, which can be used, for example, to implement the communicative abilities of a cognitive robot. DBS serves as a single underlying format for modeling communication in that it lends itself to an account of both language processing and language production (thinking is added as a separate component, which refers to inferencing on stored information, and activating content to be verbalized). As such an underlying format, it can be used to describe linguistic as well as extralinguistic content (to represent utterances and the context, respectively). Being explicitly designed for practical applications, DBS deliberately ignores linguistic phenomena considered irrelevant for these (e.g., quantifier scope). The structure of the book is as follows. It has threemain parts, which introduce DBS, outline the range of constructions covered by DBS so far, and specify fragments that can be processed or produced in the framework of DBS. There is also an appendix with two sections on the treatment of word-order variation in DBS and on the global architecture of DBS systems, and a glossary. The first part of the book starts with general principles of linguistic analysis that apply to DBS. These principles include incrementality (input is to be processed successively as it comes in, which yields an analysis for incomplete as well as complete chunks of input; the syntactic basis for this strategy is Left-Associative Grammar [Hausser 1992]), surface orientation (no empty categories), and a focus on communication (description formalisms must be able to handle turn-taking, i.e., language processing and production). After a sketch of the general theory of communication of which DBS is a part, DBS is presented in detail. It is implemented as a non-recursive data structure, that is, a list of feature structures called proplets (usually, one per word1) that are linked by coindexing the values of specific features.2 For example, subcategorizing elements (“functors”) have features whose values indicate their arguments and the other way around. In spite of its name, DBS does not offer a purely semantic representation of linguistic expressions. Although it does abstract away from purely syntactic phenomena such

Efficient Processing of Underspecified Discourse Representations

Underspecification-based algorithms for processing partially disambiguated discourse structure must cope with extremely high numbers of readings. Based on previous work on dominance graphs and weighted tree grammars, we provide the first possibility for computing an underspecified discourse description and a best discourse representation efficiently enough to process even the longest discourses in the RST Discourse Treebank.

Anti-Ikonizität an der Syntax-Semantik-Schnittstelle

Abstract The syntax-semantics interface is iconic in that it maps syntactic asymmetries (in particular, unilateral c-command) onto semantic asymmetries (scope relations). But many modification structures seem to violate this iconicity: here the modifier has (optionally or obligatorily) semantic scope over only a part of the expression that it modifies syntactically. First I will show that some well-known cases of syntax-semantics mismatch are instances of this phenomenon. Then I will specify an extremely flexible syntax-semantics interface to handle the apparent anti-iconicity. This interface crucially relies on the expressive power of a suitable underspecification formalism. With the interface one can derive the semantic representations of the problematic examples from surface-oriented syntactic structures without giving up the iconicity between syntax and semantics.Apparent anti-iconicity eventually emerges as scope underspecification between a modifier and part of the expression that it modifies. The analysis is applied to German and Turkish data.

Subcategorisation Acquisition from Raw Text for a Free Word-Order Language

We describe a state-of-the-art automatic system that can acquire subcategorisation frames from raw text for a free word-order language. We use it to construct a subcategorisation lexicon of German verbs from a large Web page corpus. With an automatic verb classification paradigm we evaluate our subcategorisation lexicon against a previous classification of German verbs; the lexicon produced by our system performs better than the best previous results.

Steuerung der Inferenz in der Diskursverarbeitung

SummarySemantic interpretation is an essential component of natural language understanding, which draws on extremely efficient language-based inference techniques. Such techniques are still lacking in computational systems for natural language processing. We have investigated specialized representation formalisms and suitable inference techniques that meet some of these desiderata. We have developed higher-order inference procedures to accurately represent linguistic ambiguities in terms of underspecification, and show how these procedures can be guided by information from other linguistic strata.ZusammenfassungSemantische Auswertung ist in der menschlichen Sprachverarbeitung unabdingbar und wird mittels sehr effizienter Inferenztechniken ausgeführt. In der maschinellen Sprachverarbeitung dagegen fehlen vergleichbare Verfahren. Wir haben spezialisierte Repräsentationsformalismen und auch geeignete Inferenzmethoden untersucht, die hier eine Verbesserung darstellen. Es wurden höherstufige Inferenzprozeduren zur paßgenauen Darstellung sprachlicher Mehrdeutigkeiten durch Unterspezifikation erarbeitet und Möglichkeiten aufgezeigt, diese durch Information aus anderen sprachlichen Ebenen zu steuern.

A Computational Model of Natural Language Communication Roland Hausser (Friedrich-Alexander-Universität Erlangen-Nürnberg) Springer, 2006, xii+365 pp; hardbound, ISBN 3-540-35476-X/978-3-540-35476-5, €69.50

The work presented in this book is motivated by the goal of applying linguistic theorybuilding to the concrete needs of potential linguistic applications such as question answering, dialogue systems, andmachine translation. To pursue this goal, a translation of linguistic theory into a framework of “practical linguistics” is suggested. Database Semantics (DBS) is presented as a first step towards such a framework. It models the communication between cognitive agents, which can be used, for example, to implement the communicative abilities of a cognitive robot. DBS serves as a single underlying format for modeling communication in that it lends itself to an account of both language processing and language production (thinking is added as a separate component, which refers to inferencing on stored information, and activating content to be verbalized). As such an underlying format, it can be used to describe linguistic as well as extralinguistic content (to represent utterances and the context, respectively). Being explicitly designed for practical applications, DBS deliberately ignores linguistic phenomena considered irrelevant for these (e.g., quantifier scope). The structure of the book is as follows. It has threemain parts, which introduce DBS, outline the range of constructions covered by DBS so far, and specify fragments that can be processed or produced in the framework of DBS. There is also an appendix with two sections on the treatment of word-order variation in DBS and on the global architecture of DBS systems, and a glossary. The first part of the book starts with general principles of linguistic analysis that apply to DBS. These principles include incrementality (input is to be processed successively as it comes in, which yields an analysis for incomplete as well as complete chunks of input; the syntactic basis for this strategy is Left-Associative Grammar [Hausser 1992]), surface orientation (no empty categories), and a focus on communication (description formalisms must be able to handle turn-taking, i.e., language processing and production). After a sketch of the general theory of communication of which DBS is a part, DBS is presented in detail. It is implemented as a non-recursive data structure, that is, a list of feature structures called proplets (usually, one per word1) that are linked by coindexing the values of specific features.2 For example, subcategorizing elements (“functors”) have features whose values indicate their arguments and the other way around. In spite of its name, DBS does not offer a purely semantic representation of linguistic expressions. Although it does abstract away from purely syntactic phenomena suchThe work presented in this book is motivated by the goal of applying linguistic theorybuilding to the concrete needs of potential linguistic applications such as question answering, dialogue systems, andmachine translation. To pursue this goal, a translation of linguistic theory into a framework of “practical linguistics” is suggested. Database Semantics (DBS) is presented as a first step towards such a framework. It models the communication between cognitive agents, which can be used, for example, to implement the communicative abilities of a cognitive robot. DBS serves as a single underlying format for modeling communication in that it lends itself to an account of both language processing and language production (thinking is added as a separate component, which refers to inferencing on stored information, and activating content to be verbalized). As such an underlying format, it can be used to describe linguistic as well as extralinguistic content (to represent utterances and the context, respectively). Being explicitly designed for practical applications, DBS deliberately ignores linguistic phenomena considered irrelevant for these (e.g., quantifier scope). The structure of the book is as follows. It has threemain parts, which introduce DBS, outline the range of constructions covered by DBS so far, and specify fragments that can be processed or produced in the framework of DBS. There is also an appendix with two sections on the treatment of word-order variation in DBS and on the global architecture of DBS systems, and a glossary. The first part of the book starts with general principles of linguistic analysis that apply to DBS. These principles include incrementality (input is to be processed successively as it comes in, which yields an analysis for incomplete as well as complete chunks of input; the syntactic basis for this strategy is Left-Associative Grammar [Hausser 1992]), surface orientation (no empty categories), and a focus on communication (description formalisms must be able to handle turn-taking, i.e., language processing and production). After a sketch of the general theory of communication of which DBS is a part, DBS is presented in detail. It is implemented as a non-recursive data structure, that is, a list of feature structures called proplets (usually, one per word1) that are linked by coindexing the values of specific features.2 For example, subcategorizing elements (“functors”) have features whose values indicate their arguments and the other way around. In spite of its name, DBS does not offer a purely semantic representation of linguistic expressions. Although it does abstract away from purely syntactic phenomena such