LINGUIST List 15.738
Sun Feb 29 2004
Review: Pragmatics/Comp Ling: K�hnlein et al. (2003)
Editor for this issue: Naomi Ogasawara <naomilinguistlist.org>
What follows is a review or discussion note contributed to our Book
Discussion Forum. We expect discussions to be informal and
interactive; and the author of the book discussed is cordially invited
to join in.
If you are interested in leading a book discussion, look for books
announced on LINGUIST as "available for review." Then contact
Sheila Dooley Collberg at collberg
linguistlist.org.
Directory
Mart� Quixal, Perspectives on Dialogue in the New Millenium
Message 1: Perspectives on Dialogue in the New Millenium
Date: Sun, 29 Feb 2004 14:44:31 -0500 (EST)
From: Mart� Quixal <marti.quixalupf.edu>
Subject: Perspectives on Dialogue in the New Millenium
EDITOR: K�hnlein, Peter; Rieser, Hannes; Zeevat, Henk
TITLE: Perspectives on Dialogue in the New Millennium
SERIES: Pragmatics & Beyond New Series 114
PUBLISHER: John Benjamins
YEAR: 2003
Announced at http://linguistlist.org/issues/14/14-2815.html
Mart� Quixal Mart�nez, Grup de Ling��stica Computacional, Departament
de Traducci� i Interpretaci�, Universitat Pompeu Fabra
''Perspectives on Dialogue in the New Millennium'' is a selection of
19 papers out of the 30 papers that were presented during BI-DIALOG (a
meeting point for researchers trying to further understand dialog, as
well as generating applicational spin-offs), held in Bielefeld in
2001. According to the book's editors, they contain (in some cases
dramatic) changes with respect to the original ones. The book seems to
be appropriate for readers with a minimum background knowledge in both
theoretical semantics and pragmatics, and their computational
counterparts. It is far from being an introduction to dialog
(systems), neither from a theoretical standpoint, nor from an
implementational standpoint. However, it is for sure a representative
collection of papers that ''documents the busy development in the
field of research on dialog'' (in the editors' words).
The disciplines that can be more clearly distinguished as taking part
in the development of dialog theory and praxis are: theoretical and
computational semantics and pragmatics, artificial intelligence and
computer science, and a range of disciplines that could be gathered
under the label cognitive sciences (or to how humans conceive the
state or situation of things). All of them are represented in this
volume.
The first three papers are devoted to the study of strategies or
formalisms to account for the inter-relation among semantics,
pragmatics and the construction of knowledge:
The first one of them is written by A. Lascarides and N. Asher, and
is entitled ''Imperatives in dialogue''. Their main point is the
utilization of a dynamic discourse semantics in order to capture
discourse effects as a by- product of discourse update. The dynamicity
of semantics is assured by the semantic definition and formalisation
of rhetorical relations such as narration, (default)- consequence,
explanation, correction, etc. Their proposal is formalised in
Segmented Discourse Representation Theory (SDRT). In this paper, the
authors defend a modular approach for the modelling of reasoning,
which is ''an attempt to do justice to the complexity of interaction
between the information sources that contribute to interpretation''.
The second paper is written by J. Ginzburg, I. Sag and M. Purver, and
is entitled ''Integrating conversational move types in the grammar of
conversation''. The paper argues in favour of integrating
conversational move types (CMT) in grammatical analysis of
conversation --within the framework of Head-driven Phrase Structure
Grammar. Their inclusion of CMTs in grammar analysis is based on both
lexical and pragmatic facts --though they admit certain points are
still to be further clarified. The enlarging of the formalism consists
--among others-- in introducing a type ''illocutionary-rel'' as part
of the HPSG type hierarchy, and a set of derivation constraints.
The third paper is written by C. Sassen, and is entitled ''An
HPSG-based representation model for illocutionary acts in crisis
talk''. This paper focuses on explaining and implementing
illocutionary acts in spontaneous speech. The author extends the
semantics of HPSG in order to handle illocutionary acts in crisis
talk. Her extension includes a set of rules for handling the distance
between an utterance, its meaning, and its logical form. She enlarges
the SEM-attribute with pieces of information such as point,
presupposition(s), sincerity, strength, etc. Thus incorporating some
of the Searlean conversation principles.
The following eight papers are related to rather
(linguistic-)theoretical aspects of dialogue. Speech acts is the
connecting theme among the first four of them: three of them are
related to presuppositions and the last one is devoted to how certain
speech acts can be considered part of semantics. The second four
papers (of this group of eight) are devoted to miscellaneous aspects
related to the explanation of human dialogue: politeness as a
perlocutionary act, models of intentions, context dependent
interpretation and implicit dialogue acts, and linguistic
non-linguistic context.
The fourth paper (in absolute figures) is written by R. van der Sandt,
and is entitled ''Denial and presupposition''. The main goal of this
paper is to shed some light in the formalisation and conceptualisation
of presupposition denials. The author distinguishes between denials
and negations as two different concepts. Denial is opposed to
assertion, and negative sentences are confronted with positive
ones. Both denials and assertions may be achieved (as speech act
concepts) by any of the two types of sentences (positive or
negative). He presents two possible accounts of presupposition denial,
which in his view differ conceptually in the fact that one of them
(the echo-analysis) ''attributes the effect [of presupposition
denials] to the discourse function of negatory force'' while the other
one (the anaphoric theory of presupposition) supports a treatment of
''presuppositional expressions as anaphoric expressions''. The
modularity and further generalization power of the first makes him
adopt. In addition, he admits that ''a formal implementation of his
proposal [in DRT] would require a non-trivial extension of the DRT
apparatus.
The fifth paper is written by J. Spenader, and is entitled ''Between
binding and accommodation''. The main goal of this paper is to analyse
whether there is a need to distinguish binding, bridging and
accommodation as three different operations. For such purpose, the
author reviews several of the proposals made in that respect (grouping
them in lexical or encyclopaedic based approaches and functional based
approaches) and a series of (bridging) annotation studies. Eventually,
she presents some work conducted in the annotation of certain
discourse strategies on spoken language data, looking for a
complementation of the annotation work done for written text, as well
as for validation of the presented theories. After going over a
number of interesting and picky examples, she concludes that
''bridging examples are very different from binding and
accommodation''. However, she proposes that bridging should be newly
defined, for which she herself gives a (provisional) definition,
specially in order to better capture aspects such as under- and over-
generation of anchors or links when looking for concrete samples of
bridging relations.
The sixth paper is written by A. Capone, and is entitled ''Theories of
presuppositions and presuppositional clitics''. This paper is divided
in two parts. The first one is a miscellaneous overview of some
aspects regarding presupposition theory, where some of the proposed
theories for presupposition projection, and its main features and some
problems are mentioned. The second part, not consistently related to
the first one (in my view), handles some aspects about
presuppositional clitics, by which is meant the (im)possibility of
using clitics in certain discourse contexts (examples from Italian).
The seventh paper is written by E. Oishi, and is entitled ''Semantic
meaning and four types of speech act''. The author considers the
question regarding truth-conditional meaning and non-truth-conditional
meaning. In contrast to previous theories, she proposes that by
extending the scope of semantic meaning straightforward truth-
conditional meaning can be regarded as a kind of conventional
meaning. Inspired in Austin (1953), the author hypothesizes the
existence of four different speech acts (stating, placing, casting and
instancing) which help her explaining some differences between both
truth- conditional meaning and non-truth-conditional meaning remaining
at the level of semantics.
The eighth paper is written by M. Terkourafi, and is entitled
''Generalised and particularised implicatures of linguistic
politeness''. It basically dwells on the several linguistic strategies
used in order to communicate and understand politeness. The author
analyses how implicatures of politeness can be calculated in several
contexts, and eventually concludes that particularised implicatures
can be drawn at several stages in the inferential process, but this is
not the case for generalised (politeness) implicatures that arise from
conventionalised language uses.
The ninth paper is written by W. Mann, and is entitled ''Models of
intentions in language''. The goal of this paper is to encourage
researchers in the field of intentions modelling to set aside the idea
that one model of intentions suffices for explaining the number of
observable communicative situations (labelled by its author as the
Single Comprehensive Model Fallacy). In contrast, the author supports
the elaboration of partial models, although he acknowledges the need
for determining aspects such as sets of attributes of intentions, or
devices in order to establish which models would more advisable to be
used in certain communicative contexts.
The tenth paper is written by J. Kreutel and C. Matheson, and is
entitled ''Context-dependent interpretation and implicit dialogue
acts''. The paper theorizes and presents an algorithm for dealing with
implicit acceptance acts in dialogue modelling. Their algorithm
(formalised and implemented as an extension of the work done in the
TRINDI project) assigns context dependent dialogue acts using certain
update rules, context-(in)dependent interpretation rules, and context
accommodation rules for handling information states. Their claim is
that dialogue participants do act even in situations where they do not
seem to be intending, whereas the representation of obligations
provides means for determining characteristic states in the course of
dialogue.
The eleventh paper is written by K. Fischer, and is entitled ''Notes
on analysing context''. This paper focuses on ''how speaker make use
of contextual factors in human- robot communication''. It briefly
overviews motivation and previous studies regarding the need of
handling context in dialogue situations. Moreover, it presents the
analysis of an experiment by which 15 different subjects were asked to
interact with a robot, where the goal was to make the robot move to
certain places within the experimentation room. The author looks for a
matching between a theory previously proposed by Clark (1996) and the
observations arising from the experimental situation, and eventually
draws some interesting conclusions regarding the corroboration of
Clark's classification.
Although some of the previous systems where framed within the
implementation of human-machine dialogue systems, none of them was a
presentation in itself of such a system. The following five papers are
centred on the development of human-machine dialogue systems.
The twelfth paper is written by A. Knoll, and is entitled ''A basic
system for multimodal robot instruction''. The paper briefly sketches
some theoretical aspects regarding the incorporation of natural
language in architectures for human-robot interaction (front-end
approach vs. communicator or incremental approach). In addition, it
presents an experiment in which human subjects are asked to
communicate with a robot so that the machine finally builds up a
series of elements of a toy construction set. Interestingly the
system incorporates multimodal interaction, which results, according
the results, in a ''very natural way'' of communication.
The thirteenth paper is written by O. Lemon, A. Bracy, A. Gruenstein
and S. Peters, and is entitled ''An information state approach in a
multi-modal dialogue system for human- robot conversation''. This
paper discusses several aspects regarding multimodal human-robot
conversations. In addition, it presents the architecture of the Open
Agent Architecture, which incorporates from a speech recognition
system to an interactive map display over speech synthesis, linguistic
parsing and language generation modules. An interesting aspect of the
system is that by means of updating information states it manages to a
more flexible strategy for modelling and processing conversations.
The fourteenth paper is written by B. Ludwig, and is entitled
''Dialogue understanding in dynamic domains''. This paper presents a
strategy for computing dialogue situations from a pragmatics first
perspective within dynamic applications. This research work relies on
the idea that dialogue can be determined ''by dynamically changing
content of the belief structures of the dialogue participants. The
implementation bases on a set of minimal set of communicative acts and
a number of orthogonal conditions for updating belief structures in
order to compute dialogue progress.
The fifteenth paper is written by R. Copper, S. Ericsson, Staffan
Larson and I. Lewin, and is entitled ''An information state update
approach to collaborative negotiation''. The paper presents a strategy
for dealing with collaborative negotiation within a GoDiS
implementation. The authors start by analyzing a previous proposal
--Sidner (1994),-- and try to overcome some of its shortcomings. Their
proposal reduces considerably the number of dialogue moves needed to
handle the (sub-)domain presented in the paper --only four central
primitive moves are needed. In addition, their proposal is flexible
with respect to the degree of optimism (or pessimism) with which
participant's contributions have to be interpreted. Last but not
least, they introduce a distinction between negotiation of
alternatives and negotiation of uptake, for which they point out a
feasible solution.
The sixteenth paper is written by D. Schlangen, A. Lascarides and
A. Copestake, and is entitled ''Resolving Underspecification using
Discourse Information''. The paper aims at investigating the
interaction between compositional semantics, goals, and discourse
structure in task-oriented dialogues, namely in the domain of fixing
appointments. Their main thesis is that information can flow either
from resolving the semantic underspecification to computing the
rhetorical relation, or vice versa. The proposal is in the framework
of Segmented Discourse Representation Theory and is implemented using
LKB.
The last three papers centre on theoretical aspects of dialogue
formalisation. The first one of them centres on the interpretation of
coordinated sentences, whereas the two following ones are centred on
the use of specific discourse markers both in German and English.
The seventeenth paper is written by A. Benz, and is entitled ''On
coordinating interpretations -- optimality theory and rational
interaction''. The author tries to clarify the relation between a
model approaching anaphora resolution within a theory of rational
interaction and bidirectional Optimality Theory. It defends that
anaphora resolution can be seen as a translation problem where a set
of ''original'' sentences must be handled with a set of translating
formulas. The result of such an operation must be a balance between a
best form chosen by the speaker and a most preferred meaning chosen by
the listener. The whole process is conducted and determined by the
communicative situation and the dialogue state.
The eighteenth paper is written by E. Karagjosova, and is entitled
''Modal particles and the common ground''. The paper centres on the
analysis of the meaning of certain German modal particles. The author
tries to distinguish the amount of inherent meaning in such modal
particles from that amount of meaning that they acquire in context.
Finally the author suggests that the contribution of the modal
particles to utterance meaning can be captured by framework that
considers the basic meaning of the particle, the illocution of the
utterance the modal particle occurs in, and the function of the
utterance in discourse.
The nineteenth paper is written by T. Tenbrink & F. Schilder, and is
entitled ''(Non-)Temporal concepts conveyed by before, after, and then
in dialogue''. The papers aims at analyzing the use of before, after
and then as indicators of temporal order. They review previous
analyses, and explore some corpora in order to check real uses. They
eventually conclude that there is clear presence of time sense in the
use of such words, either precedence or proximality. Moreover, they
prove that discourse relations are required for determining the
semantics of those words. Finally, they conclude that the presence of
a time reference shall have effects on their interpretation.
REFERENCES
Austin, J.L. (1953) How to talk -- some simple ways. Proceedings of
the Aristotelian Society. [Reprinted in: J.O. Urmson & G.J. Warnock
(Eds.), Philosophical papers (pp. 134-135). Oxford: Oxford University
Press.
Clark (1996) Using language. Cambridge University Press.
Sidner (1994) An artificial discourse language for collborative
negotiation. In Proceedings of the fourteenth National Conference of
the American Association for Artificail Intelligence (AAAI-94) (pp.
814-819)
ABOUT THE REVIEWER
Mart� Quixal is PhD student in the program Cognitive Science and
Language at Universitat Pompeu Fabra in Barcelona. He works at the
GLiCom research group http://mutis.upf.es/glicom in the development
of robust low-level linguistics-based parsers. He is currently working
in the implementation of NLP-enhanced error detection tools for second
language learners, and is interested in introducing partial semantic
or pragmatic information in texts in order to detect errors beyond the
morphosyntactic level.