Keith Stenning and Mike Oaksford:
Rational Reasoning and Human Implementations of Logic
In this paper we draw some general conclusions about the relation
between logic and the analysis of human reasoning; that relation is
illumninated by considerations of implementation. We conclude that a
modern view of logic which is much more abstract than earlier views
makes a good partner for psychology because it focusses attention on
issues about how fragments of logics can be implemented within known
resource limitations, and these are the foremost issues with which a
psychology of reasoning should concern itself.
(March 1992; 40 pages)
Ref. No: HCRC/RP-28 Price: UKL 1.50
Anne H. Anderson, Miles Bader, Ellen Gurman Bard, Elizabeth Boyle,
Gwyneth Docherty, Simon Garrod, Stephen Isard, Jacqueline Kowtko, Jan
McAllister, Jim Miller, Catherine Sotillo, Henry Thompson and Regina
Weinert:
The HCRC Map Task Corpus
This paper describes a corpus of unscripted, task-oriented dialogues
which has been designed, digitally recorded, and transcribed to support
the study of spontaneous speech on many levels. The corpus uses the Map
Task (Brown, Anderson, Yule, and Shillcock, 1983) in which speakers must
collaborate verbally to reproduce on one participant's map a route
printed on the other's. In all, the corpus includes four conversations
from each of 64 young adults and manipulates the following variables:
familiarity of speakers, eye contact between speakers, matching between
landmarks on the participants' maps, opportunities for contrastive
stress, and phonological characteristics of landmark names. The
motivations for the design are set out and basic corpus statistics are
presented.
(March 1992; 22 pages)
Ref. No: HCRC/RP-29 Price: UKL 1.10
Fairouz Kamareddine and Ewan Klein:
Nominalization, Predication and Type Containment
In an attempt to accommodate natural language phenomena involving
nominalization and self-application, various researchers in formal
semantics have proposed abandoning the hierarchical type system which
Montague inherited from Russell, in favour of more flexible type
regimes. We briefly review the main extant proposals, and then develop
a new approach, based semantically on Aczel's notion of Frege structure,
which implements a version of {\em subsumption polymorphism.
Nominalization is achieved by virtue of the fact that the types of
predicative and propositional complements are contained in the type of
individuals. Russell's paradox is avoided by placing a type-constraint
on lambda abstraction, rather than by restricting comprehension.
(May 1992; 42 pages)
Ref. No: HCRC/RP-30 Price: UKL 1.60
Jacqueline C. Kowtko, Stephen D. Isard and Gwyneth M. Doherty:
Conversational Games Within Dialogue
An analysis of task oriented dialogue has been developed around
goal-directed exchanges labelled conversational games. Games account
for that aspect of discourse coherence that is manifested in
initiation-response-feedback patterns, and they do so by relating the
form of dialogue to underlying non-linguistic goals. The system applies
to real, spontaneous dialogue and may prove useful in the design of
computer dialogue systems. A study was conducted to determine agreement
among analysts, upon coding two sets of dialogues. One expert and three
novice analysts agreed on an average of 78\% of the moves assigned.
When certain consistent differences are taken into account, however, the
accuracy increases to 88\%.
(June 1992; 12 pages)
Ref. No: HCRC/RP-31 Price: UKL 0.70
Workshop on Computational Semantics
This document reports on a workshop held 5th-8th March, 1992 in which a
number of people met to discuss topics relating to the current state and
future development of the field of computational semantics. The
majority of the time was spent in working groups discussing the issues
reported on in this document. The intention is that this report should
stimulate furhter discussion in the field.
(July 1992; 29 pages)
Ref. No: HCRC/RP-32 Price: UKL 1.20
Patrick Blackburn:
Structures, Languages and Translations: the Structural Approach to
Feature Logic
This paper discusses an approach to feature logic called the structural
approach. The method consists in first viewing feature structures as
models (or relational structures) in the standard model theoretic sense,
and then going on to consider various languages that can be interpreted
on these structures and the way these languages are interrelated. It
will be shown that the structural approach offers a natural perspective
from which to relate various feature logics that have been proposed in
the literature, and in addition it will be argued that it suggests
strategies for devising interesting and tractable constraint languages.
(August 1992; 28 pages)
Ref. No: HCRC/RP-33 Price: UKL 1.20
Matthew Walter Crocker:
A Logical Model of Competence and Performance in the Human Sentence
Processor
This work is concerned with the way in which principal theories of
syntax and modular theories of mind may participate in an incremental
model of human linguistic performance. Central to current linguistic
theory is the distinction between competence, what we know about
language, and performance, how we use that knowledge. While current
theories of grammar suggest a highly modular, abstract, language
universal characterisation of linguistic competence, traditional models
of performance have postulated parsing strategies based on
construction-oriented, phrase-structure grammars. In contrast, we
construct a principled theory of performance on the basis of
cross-linguistic evidence, with the aim of shedding greater light on the
relationship between grammar and processing.
On the basis of Fodor's Modularity Hypothesis and a range of empiracal
evidence, we assume the existence of a distinct syntactic processor
within the human sentence processor. We further hypothesize {\it The
principle of Incremental Comprehension, entailing that the sentence
processing strive to achieve maximal incremental comprehension as each
word in a sentence is encountered. To achieve incremental comprehension
at the various levels of linguistics processing (syntactic, semantic,
etc.) we therefore predict that modules operate concurrently. We then
suggest that the modularity paradigm is one to be exploited whenever
possible, precisely because it permits distributed processing within a
particular domain, thereby improving real-time performance. This maxim
of modularity, combined with the natural partitioning of syntactic
information into several informationally encapsulated representation
types leads us to posit four sub-modules within the syntactic processor:
(1) phrase structure, (2) chain structure, (3) conference, and (4)
thematic structure. We then present several processing strategies which
are claimed to be operative within the phrase structure module and
demonstrate how the proposed architecture successfully accounts for a
variety of empiracal phenomena, across several languages both the
cross-linguistic application and possibly innate status of the model. We
further argue that the processing strategies follow from the more
general Principle of Incremental Comprehension.
To demonstrate the operation of the proposed model, and illustrate its
principled basis, we employ the Algorithm = Logic + Control paradigm of
logic programming. In particular, we present a 'deductive'
implementation wherein each module is realised as a specialised
meta-interpreter which constructs its own representation with respect to
the relevant principles of grammar. We then demonstrate how these
interpreters may be `coroutined' so as to model the concurrency of the
performance theory. Finally, we see that the implemented model
successfully accounts for subsets of both English and German, without
any variation of the control strategy. We take this as further support
for the existence of an innate, unparameterised sentence processing
mechanism, and discuss some possible implications of this consclusion.
(September 1992; 200 pages)
Ref. No: HCRC/RP-34 Price: UKL 7.70
Richard C Shillcock and Ellen Gurman Bard:
Modularity and the processing of closed-class words
The process of lexical access has been shown to be immune from the
effects of syntactic context; it is informationally encapsulated. Thus a
listener automatically accesses even the syntactically inappropriate
meanings of an ambiguous word like "rose". Two cross-modal priming
experiments are presented which demonstrate that the lexical access
module is permeable to syntactic information in the case of those
closed-class words like "can" and "would" which are homophonous with
open-class words: when "would" is heard in context, subjects do not
access "wood". Subsequent analyses, along with experiments testing the
discriminability of the open- and closed-class tokens and measuring the
strength of the syntactic context, suggest that the results are not the
result of word frequency effects, but genuinely reflect the deployment
of syntactic information. The results are discussed in terms of their
implications for the modularity hypothesis and the storage and access of
closed-class words.
(September 1992; 17 pages)
Ref. No: HCRC/RP-35 Price: UKL 1.00
Ellen Gurman Bard and Richard C Shillcock:
Competitor Effects during Lexical Access: Chasing Zipf's Tail
The difficulty of recognizing a spoken word appears to be related to at
least three different characteristics of the set of similar words or
"lexical competitors": the size of the set of competitors (Luce, 1986,
Frauenfelder and Peeters, 1990), the aggregate frequency of all
competitors (Luce, 1986) and the frequency of the most frequent
competitor (Brown, 1987; Grainger @i[et al.] 1988; Marslen-Wilson,
1990). The present paper reconciles these claims by showing that the
three measures of competition intercorrelate because of the shape of the
distribution of word frequencies in a typical set of lexical
competitors. Like the whole lexicon (Zipf, 1935), sets of lexical
competitors contain many rare and very few extremely frequent
words. Both the word-initial cohorts formed in an exhaustive
partitioning of the phonemically transcribed words of the MRC
Psycholinguistics Database (Coltheart, 1981) and a set of orthographic
neighbours (Andrews, 1989) are shown to have distributions of word
frequency expected in random samples from a population with extreme
positive skew. These facts stand in contradiction to Landauer and
Streeter's (1973) claims that different phonological elements of English
are associated non-randomly with different parts of the word-frequency
continuum. The same facts permit simplification of lexical access
models.
(September 1992; 44 pages)
Ref. No: HCRC/RP-36 Price: UKL 1.60
Jonathan Ginzburg:
Propositional and Non-Propositional Attitudes
The term propositional attitudes, has achieved legitimacy under
the assumption that the objects of such activities as claiming,
believing, reporting or discovering are individuated by
means of a single class of entities, the propositions, a class
of descriptively potent entities of which truth and falsity can be
predicated, corresponding roughly to their descriptive accuracy. This
paper argues that no logically coherent class of entities can
simultaneously individuate the potential fillers of the `cognitive'
argument role of all propositional attitude predicates (PAP's). Thus,
it is shown that there exist at least three syntactically distinct
classes of expressions, interrogative clauses, verbal gerunds and noun
phrases, all of which have uses which denote veridical descriptive
entities. And yet a coherent class of PAP's, including the predicate
true, cannot felicitously predicate of (uses of) any of these
expressions types. It is proposed that this class of predicates can
be (partially) characterized in terms of a transparency to predication
of truth which their nominally presented arguments exhibit. An
account of these data in terms of pragmatic presupposition is
considered and argued to be inadequate. This leads to an account
reminiscent of but distinct from that propounded at a certain period
by Russell. The proposal involves recognizing that two classes of
entities that can fill `cognitive' argument roles: propositions and
states-of-affairs. On the account proposed, which builds particularly
on insights of Austin, Vendler and work in Situation Theory.
States-of-affairs are entities which can but need not be {\em
factual. Truth, however, is not applicable to this class of entities,
but rather is a property of propositions, conceived of as
predications that a state-of-affairs accurately describes a given part
of the world. The paper contains a semantics for declarative
and interrogative sentences.
(October 1992; 47 pages)
Ref. No: HCRC/RP-37 Price: UKL 1.70
Jon Barwise and Robin Cooper:
Extended Kamp Notation A Graphical Notation for Situation Theory
This work is part of a project to consolidate recent work on situation
theory into a workable version and to solve a represenational problem
with situation theory. Our goal here is to design a notation for
representing situation-theoretic objects that uses insights from
Kamp's discourse representation structures.
(November 1992; 27 pages)
Ref. No: HCRC/RP-38 Price: UKL 1.20
Alistair Knott and Robert Dale:
Using Linguistic Phenomena to Motivate a Set of Rhetorical Relations
The notion that a text is coherent in virtue of the `relations' which
hold between the elements of that text has become fairly common
currency, both in the study of discourse coherence and in the field of
text generation. The set of relations proposed in Rhetorical Structure
Theory (Mann and Thompson [1987]) has had particular influence in both
of these fields. But the widespread adoption of `relational'
terminology belies a certain amount of confusion about the relational
constructs themselves: no two theorists use exactly the same set of
relations; and often there seems no motivation for introducing a new
relation beyond considerations of descriptive adequacy or engineering
expedience.
To alleviate this confusion, it is useful to think of relations not just
as constructs with descriptive or operational utility, but as
constructs with psychological reality, modelling real cognitive
processes in readers and writers.
This conception of rhetorical relations suggests a methodology for
delineating a set of relations to work with. Evidence that a relation
is actually used by speakers of a language can be obtained by
looking at the language itself---in particular by looking at the range
of cue phrases the language provides for signalling relations. It is to
be expected that simple methods will have evolved for signalling the
relations we find most useful.
This paper presents a bottom-up methodology for determining a set of
relations on the basis of the cue phrases which can be used to
mark them in text. This methodology has the advantage of starting from
concrete linguistic data, rather than from controversial assumptions
about notions like `intention' and `semantics'.
(December 1992; 30 pages)
Ref. No: HCRC/RP-39 Price: UKL 1.20