The DANDELION Project: Abstracts
Discourse Structure
Lavid, Julia.
Generic Structure Potential:
A Functional Characterization of Global Discourse Structures.
The aim of this deliverable is to focus on the role of the so-called "generic" structures for text comprehension and production. The paper reviews the different approaches to the study of these structures proposed in the literature, and attempts to delimit their discourse functions and their patterns of lexicogrammatical realization. An important part of these considerations is the interplay of these generic structures, which are largely determined by genre-specific conventions, and local coherence links. In this sense, this study illustrates, with practical examples, the similarities and differences between these two different approaches to text organization, and suggests the way in which they can be integrated and used complementarily in a text planning system.
[R111b.ps.Z, size: 577 kB].
This deliverable considers the problem of providing `coherence relation' analyses for natural texts. A critical review of the state of the art reveals a number of problems which require attention before any particular mode of analysis can be adopted. These problems are addressed on the basis of empirical analyses of texts in order to arrive at some improved guidelines for what is to be interpreted as a `coherence relation' and for establishing an improved specification of coherence relations in general. We illustrate both some of the benefits of the specification for linguistic analysis and
some possibly problematic areas for computational specification.
[R112.ps.Z, size: 463 kB].
I present an analysis of the connective while and the way it functions in context, based on a study of a set of while-clauses in two different text types. I propose to view while as one connective with a basic meaning and a number of different surface realizations. I show that different interpretations of while are subject to constraints coming from different directions, and that in determining how a particular interpretation arises attributes of the sentence as well as its context need to be taken into account. In the proposed approach, syntactic structure, semantic meaning and discourse function are closely intertwined.
[R113a.ps.Z, size: 116 kB].
This paper contains a proposal for analysing connectives. The causal and contrastive connectives studied appear to have several interpretations, corresponding to their primarily contributing to the ideational structure, to the rhetorical structure or to the attentional structure of the discourse. An analysis of the usage of eight connectives, omdat, want, zodat, opdat, doordat, maar, terwijl and hoewel, resulting in a set of conditions on context for each individual interpretation is presented. A representation mode for connectives is proposed, expressing a perspective on their relationship to coherence relations.
[R113b.ps.Z, size: 182 kB].
Staging
Lavid, Julia.
Thematic Development in texts.
The concepts of Theme and Rheme, Given and New in Linguistics, and the notion of Focus in Computational Linguistics, have been the object of much research in both fields where different authors have used the terms sometimes interchangeably and with such ranges of overlap and confusion that make their study a hard task for the researcher. This deliverable is an attempt to clarify these notions by mapping out their functionalities and extracting their commonalities in such a way that they are useful for both the linguistic and the computational communities. In particular, the paper attempts to provide operational redefinitions for these notions and explores the constraints imposed by several contextual factors on the discourse phenomena.
[R121.ps.Z, size: 130 kB].
Lavid, Julia.
Theme, Discourse Topic and Information Structuring.
The present deliverable is an investigation into the role of various phenomena responsible for the structuring and development of information in discourse and the contextual factors which constrain their deployment. These phenomena, studied and redefined in Deliverable R1.2.1, include Theme, Topic, Focus and Focus-shift rules. This paper redefines and tests the hypothesis presented in our R1.2.1, according to which certain features of the communicative situation - i.e. those that determine the text type - prespecify the selection of particular lexico-grammatical and textual realization options. The result of the empirical analysis presented here is a specification of the factors governing information structuring, empirically tested on a variety of text type examples. The paper also outlines a dynamic prespecification of the process of information development with a view towards computational implementation.
[R122b.ps.Z, size: 157 kB].
Temporal Structure
The central idea behind this deliverable is that text type has an
effect on the functioning of aspectual constructions in discourse.
To show this, we compare the discourse function of aspectual information
in two different text types in English, viz. narrative fictional texts
and newspaper reports. We outline the general characteristics of each
text type, and the way these affect the use of events and states at
the discourse level. We also consider in more detail the past perfect
in the two text types: we show how its use is subject to a number of
general discourse principles, and illustrate how these principles
interact with the structure of the two text types.
[R131-part1.ps.Z, size: 127 kB].
Part II of this report focusses on differences between English and
Dutch in the use of Pluperfect. As in Part I, the investigation is not
restricted to one text type, but covers narrative texts and news reports.
Distributional differences in the use of Pluperfect are explained in
terms of the interplay between functional text properties, specifically
the role of states and events in the two genres, and language-specific
characteristics in the interpretation of Perfect forms,
in particular the Pluperfect, in English and Dutch.
[R131-part2.ps.Z, size: 76 kB].
This paper presents an analysis of a set of Dutch temporal
connectives (TCs). It attempts to explain both the semantic unity of certain
TCs (within the so-called TTT framework) and the influence of contextual
properties on their interpretation. Both temporal and not primarily temporal
interpretations of TCs are analyzed. The results of the analysis are presented along the lines of Systemic Functional Grammar, conforming to the purpose and format of the Dandelion project.
[R132a.ps.Z, size: 235 kB].
[R132a-fp.ps.Z, size: 24 kB].
N.B. this paper consists of two files: the paper itself (R132a.ps.Z) and its front page (R132a-fp.ps.Z).
Hitzeman, Janet.
Text Type and the Position of a Temporal Adverbial within the Sentence.
In this document, I argue that a sentence with a certain type of temporal adverbial is ambiguous, and that one reading is lost when the adverbial appears in sentence-initial position. Sentence (1a), for example, has a reading in which there was some three-year period in the past during which Mary lived in Amsterdam and a reading in which Mary has lived in Amsterdam for the three years preceding speech time:
- a. Mary has lived in Amsterdam for three years.
b. For three years Mary has lived in Amsterdam.
Sentence (1b) has only the reading in which Mary lives in Amsterdam at speech time and has done so for the preceding three years. The reading that remains when the adverbial is in sentence-initial position is more specific about the time at which the event occurs, and therefore one would expect to see more initial-position adverbials in a narrative text, where the order of events is important. In testing this hypothesis on the ECI corpus, it was found that it is not the narrative/non-narrative distinction that results in a significant difference in initial-position adverbial usage; instead, narratives with a large amount of flashback material have significantly more initial position adverbials, indicating that in order to accurately predict adverbial position a subclassification of the category "narrative" based on the amount of flashback material is needed.
[R132b.ps.Z, size: 106 kB].
Nominal Anaphora
This paper presents a functional analysis of nominal anaphors (NAs) in
discourse. We survey the area of NP-anaphora starting from the
traditional view on anaphoric expressions, according to which
NP-anaphors form an accessibility hierarchy and NAs provide anaphoric
access to discourse referents with low accessibility. This
`identificational view' is then challenged with empirical evidence
showing various non-identificational discourse functions of definite,
indefinite, and demonstrative NAs. NAs are shown to participate not
only in the management of reference, but also in other semantic and
pragmatic discourse functions at local and global levels. They express
interactive relations between writer and reader, they contribute to the
semantic representation of discourse, and they express structural
thematic relationships in discourse.
[R141.ps.Z, size: 128 kB].
Hustinx, Lettica and Wietske Vonk.
Psycholinguistic experimentation on the qualificational
function of demonstrative noun anaphora.
This deliverable presents an empirical validation of an interpretation
proposal for demonstrative noun anaphors. A demonstrative noun anaphor was
proposed to fulfil more than the merely identificational function or its
definite counterpart. This study focussed on the additional function of the demonstrative noun anaphor of the type
`classify'. This type of demonstrative anaphor, consisting of a
superordinate anaphor (e.g., `this president') that refers to its
subordinate antecedent (e.g., `Clinton'), was proposed to activate other
class members of the superordinate concept (e.g., `Kennedy', `Reagan'). The
classify function of the demonstrative noun anaphor was experimentally
investigated in a number of experiments by comparing the activation evoked
by the demonstrative anaphor (e.g., `this president') to that evoked by the
definite anaphor (e.g., `the president'). Demonstrative and definite noun
anaphors were embedded in texts in controlled experimental settings. The
hypothesized classify effect of the demonstrative anaphors was
corroborated.
[R142a.ps.Z, size: 71 kB].
In this paper a discourse functional analysis of the use of marked
indefinite noun phrases is presented. It is based on English and Dutch text
fragments of narrative as well as expository texts. The aim is to show that
marked indefinite expressions are used not only to identify discourse
referents, but also to convey additional meaning. The analysis is based on
the assumption that the use of referring expressions is guided by the
pragmatic principle of cooperative interaction between reader or writer and
recipient. Following this principle, the form of a referring expression is
in line with the recipient's knowledge about the referent at the moment of
comprehension of the utterance. For marked referential expressions,
however, this principle is violated. It will be argued that the use and
interpretation of a marked indefinite expression is derived from the
original function of the indefinite form when it is used in an unmarked
way.
[R142b.ps.Z, size: 80 kB].
Grammar Resources
Calder, Jo (editor).
Update of Baseline Linguistic Resources in Dandelion.
In this report, we focus on necessary extensions and refinements of the linguistic resources made available to the project by month 18. In a second step, we identify for the PENMAN and KOMET systems where, relative to the projects needs, extensions and refinements seem desirable. This constitutes the main part of the current report as most of the specification and implementation work undertaken so far has been done in that system. Thereafter, we discuss the grammar developed by Bob Carpenter and Gerald Penn, of CMU, which is likely to form the basis for developments in the grammatical and discourse descriptions of English and Dutch. In an appendix, we shall comment in some detail on the grammar development shell Pleuk made available from Edinburgh - while this is not strictly speaking a linguistic resource, it does represent a possible basis for the implementation of the workbench to be pursued in the work package 2, and this report seems an appropriate forum in which to discuss advantages and disadvantages of following this route.
[R211.ps.Z, size: 85 kB].
Verschuur, Leon.
A Constraint-Based Grammar of Dutch Temporal Connectives.
This papers formalises the essence of Oversteegen's work on Dutch temporal connectives in a HPSG variant of constraint-based grammar. It does so by developing typed feature structure representations of tense and aspect, proposing a way to combine them into lexical representations of verbs, and developing a method of checking consistency of all temporal information in a connective subsentence-main sentence configuration. Checking consistency is done by combining the representations of the core meanings within a context feature. Deriving the temporal relations holding between the denoted events depends on the aspectual types of sub- and main clause. By notating the various possibilities in disjuncts, constrained by event type, the actual relations are derived. Also steps are taken to account for some rhetorical relations such as consequentiality and causation.
[R231a.ps.Z, size: 171 kB].
Verschuur, Leon.
A Discourse Grammar for Classifying Demonstrative Nominal Anaphora.
A formal semantic explanation is given for the fact that superordinate demonstrative noun anaphors can classify their referents, an effect which has been corroborated by experimental investigations. A completed implementation in Definite Clause Grammar is described, while the possibilities for a Typed Feature Structure implementation are indicated.
[R231b.ps.Z, size: 170 kB].
This document outlines central elements of a `discourse grammar for German'. It is structured around two complementary perspectives on the topic: one proceeds from a global perspective and starts from a text as a whole with its text-type specific properties which find their characteristic reflections on local, sentence-level discourse organization. The second perspective takes the local, sentence-level options of discourse organization as a starting point in order to specify an interface to their global, textual motivations. The aim is to integrate the two views in a common architecture.
[R232.ps.Z, size: 469 kB].
Hitzeman, Janet.
A Constraint-Based Grammar of English Temporal Connectives.
This document describes a set of feature structure representations for English temporal connectives. Their lexical entries are described and motivated in detail, along with a type hierarchy that will be used in the computational HPSG/DRT system we are building to parse texts containing these temporal connectives. This work is based on the TTT proposal for Dutch temporal connectives described in [Oversteegen, 1989, Oversteegen, 1993] and on Martin's [1992] system representing the ideational (external) meaning of temporal relations in English.
[R233.ps.Z, size: 131 kB].
Return to the main page...
Return to the previous page...
Last update: January 23, 1996.
Klaas Jan Rondhuis