Abstracts of HCRC Research papers, 1990

Matthew W. Crocker:
Principle-Based Sentence Processing: A Cross Linguistic Account
This paper presents a model of sentence processing based directly on the ``principles and parameters'' model grammar. The model distinguishes four processing modules, each for a particular informational type, namely: 1) phrase-structure 2) thematic structure 3) chains, and 4) coindexation relationships. In the context of this model, we postulate that the phrase-structure processor prefers attachment of lexical items into argument positions (or, A-positions). This ``A-Attachment'' strategy makes interesting predictions for languages in which arguments precede their heads, such as Dutch. Specifically, we provide an account for some attachment preferences in Dutch as observed by Frazier and then extend our analysis to explain some garden path phenomena in German.
(March 1990; 15 pages)
Ref. No: HCRC/RP-1 Price: UKL 1.00

Robin Cooper:
Information in the Early Stages of Language Acquisition
In the early stages of language acquisition children seem to use single words or a small number of words to describe the kind of situations that adults would describe with complete sentences. We shall sketch a theory of information states and communication which will account for the fact that children can use short utterances to report complex situations they perceive---or at least suggest how they might appear to be doing so to an adult. We shall suggest that a theory of information states and communication can give a more explanatory account of early language acquisition than formal semantics and can provide a theoretical foundation for the claim in the psycholinguistic literature that ``semantics'' is central to the genesis of language in children. Furthermore we shall suggest that an examination of the restricted language used by children and the perception of their utterances by adults can provide telling arguments for the situated nature of human language processing.
(April 1990; 11 pages)
Ref. No: HCRC/RP-2 Price: UKL 0.70

Jon Oberlander and Peter Dayan:
Altered States and Virtual Beliefs
Functionalism avoids a potentially fatal infinite regress by realising the low level phenomena of mind in mere Turing machines rather than via undischarged homunculae. Adopting a narrow, logical view of such mechanisms has led to two sorts of problems: exponentially intractable models of human inferential competence; and a wide divergence between the supposed competence and observable human performance. These problems have produced a retreat to instrumentalism about folk psychological concepts. We argue that abandoning hope in this way is premature. The problems arise from two quarters: an impoverished notion of competence models that divorce them too far from performance models; and too narrow a view of possible functional mechanisms. We explore the consequences of retaining a traditional view of inference, while adopting a new mechanism for memory. We motivate the latter by developing a seemingly unnatural picture of von Neumann machines, taking their finite memory seriously. Together, the two repairs suffice to make instrumentalism avoidable.
(April 1990; 10 pages)
Ref. No: HCRC/RP-3 Price: UKL 0.70

Jerry Seligman:
Perspectives in Situation Theory
Situation Theory supplements the logician's diet of individuals, properties and propositions with a number of new metaphysical categories. Amongst these are situations, facts (or infons) and constraints. In this paper we add a new one (perspectives) to capture the different ways situations are classified by facts they support and the constraints that hold in them. Importantly, there are many perspectives on the same situation.
Perspective can be used to explain the `background conditions' appealed to in situation theoretic accounts of conditionals and also account for differing intuitions about the flow of information, corresponding to `logical' and `informational theoretical' (Dretskian) approaches. In the final section we provide a construction of some of the old metaphysical categories (individuals and properties) from perspectives and certain maps between perspectives called shifts. This offers both a novel approach to characterising individuation and {\em predication and the possibility of a more parsimonious metaphysics.
(April 1990; 39 pages)
Ref. No: HCRC/RP-4 Price: UKL 1.50

Judy Delin:
Accounting for Cleft Constructions in Discourse: A Multi-Layered Approach
Existing accounts of why cleft constructions are used in discourse---such as those that claim that clefts serve to signal `focus', or that clefts indicate the position of information of a particular discourse status---are demonstrated to be inadequate when analysed in the light of a corpus of naturally-occurring data. This paper points out the problems with some prominent claims about clefts, and presents an alternative account, suggesting that the decision to use clefts is based on a combination of factors at the levels of syntax, semantics and pragmatics.
(April 1990; 33 pages)
Ref. No: HCRC/RP-5 Price: UKL 1.30

Keith Stenning:
Modelling Memory for Models
We contrast two theoretical interpretations of mental models. Johnson-Laird and his associates (eg. Johnson-Laird \& Steedman 1978, Johnson-Laird 1983) interpret models as a medium of mental representation. Alternatively, models may be treated as abstract objects which are important because they specify what is represented. This paper takes the second view of mental models and shows how a theory of distributed representation of binding in memory, elaborated elsewhere (Stenning, Shepherd \& Levy 1988, Stenning \& Levy 1988 , explains why people adopt distinctive reasoning strategies.
(May 1990; 15 pages)
Ref. No: HCRC/RP-6 Price: UKL 1.00

Alex Lascarides:
The Progressive and the Imperfective Paradox
Formal semantics constitutes the framework of the research presented here, and the aim is to provide a solution to the imperfective paradox; i.e. explain why ``Max was running'' entails ``Max ran'', but ``Max was running home'' does not entail ``Max ran home''. This paper is divided into two parts. In Part I we evaluate what I will call the Eventual Outcome Strategy for solving the imperfective paradox. This strategy is commonly used (Dowty 1979, Hinrichs 1983, Cooper 1985), and is highly intuitively motivated. I will show, however, that the formulations of the intuitions give rise to conflicts and tensions when it comes to explaining the natural language data. In Part II we offer a new approach to tackle the imperfective paradox that overcomes the problems with the Eventual Outcome Strategy.
(June 1990; 38 pages)
Ref. No: HCRC/RP-7 Price: UKL 1.40

Alex Lascarides:
Knowledge, Causality and Temporal Representation
In this paper, a formal semantic framework is developed in order to account for the temporal semantics of text. The theory is able to represent and reason about both semantic issues, which are independent of world knowledge ({\sc wk), and pragmatic issues, which are not, within a single logical framework. The theory will allow a text's semantic entailments to differ from its pragmatic ones, even though they are all derived within the same logic. I demonstrate that this feature of the theory gives rise to solutions to several puzzles concerning the temporal structure of text.
(June 1990; 39 pages)
Ref. No: HCRC/RP-8 Price: UKL 1.50

Keith Stenning:
What are We to Make of Computers, and Computers to Make of Us? Reflections on Winograd and Flores' Understanding Computers and Cognition
Winograd and Flores' `Understanding Computers and Cognition' proposes that the rationalist tradition in AI must be replaced by a hermeneutic approach. Associating the rationalist tradition with the goal of building a human mind, the authors propose that a hermeneutic approach must adopt the goal of constructing prostheses which magnify the human mind.
This paper argues that what AI needs is not so much a hermeneutic approach as a better appreciation of biology and psychology. These disciplines are responsible for what we know of organisms adapted to environments, and the relationship between knowledge and context. This point of view is illustrated by appeal the distinctive characteristics of human episodic memory.
(August 1990; 16 pages)
Ref. No: HCRC/RP-9 \sellingprice{pounds 1.00

Matthew W. Crocker:
Incrementality and Modularity in a Principle-Based Model of Sentence Processing
Central to the ``principles and parameters'' paradigm of modern transformational grammar is a modular set of axiomatic principles which are stated in terms of several distinct representational forms; namely phrase structure, chains, coindexation, and thematic structure. In previous research, we have hypothesized the existence of four distince processing modules; one for each informational aspect of the grammar. The modules are intended to act concurrently, with a limited flow of communication between them.
This paper describes a ``deductive'' implementation of the processing model, in which each processing module is a specialised meta-interpreter over the relevant components of the grammar. Within this framework, we will show how processing strategies can be implemented simply as specialised `selection' rules within the individual meta-interpreters. In addition, the four processors are ``coroutined'' so as to model the concurrency intended by the theory of processing.
(September 1990; 22 pages)
Ref. No: HCRC/RP-10 Price: UKL 1.10

HCRC Research Overview 1990
(November 1990; 117 pages)
Ref. No: HCRC/RP-12 Price: UKL 0.60

Robin Cooper:
Three Lectures on Situation Theoretic Grammar
This report presents extensive notes for three lectures on Situation Theoretic Grammar that were prepared for the APPIA Advanced School on Natural Language Processing, which took place in Guarda, Portugal in October 1990. Topics covered include motivation for the situation theoretic approach to natural language processing, infons, parametric objects, environments, generalised quantifiers and quantifier scope.
(November 1990; 37 pages)
Ref. No: HCRC/RP-13 Price: UKL 1.40

Judith Ramsey \& Keith Oatley:
Designing Minimal Models from Scratch
We designed from scratch a minimal manual of the kind proposed by Caroll (1990) for Unix e-mail, using a set of six user measurement instruments in the different phases of iterative design and testing. The design process involved acquiring from a set of experts information about e-mail use that would be helpful to novices, and acquiring from novices information that guided subsequent versions of the manual. The seventh version of the minimal manual was tested against a commercial manual, in a comparative performance experiment with 30 naive subjects. It had approximately 13\% of the pages of the commercial manual; it resulted in 30\% faster learning and more effective use of the e-mail system overall, and significantly better performance on individual subtasks; including the recovering from errors. Significantly more users were satisfied with it than the conventional manual. A serendipitous finding from the performance experiment was that males made more attempts at error recovery than females, but were not more successful in recovering from errors. Carroll's general principles of manual design for minimal manuals were found to be a good basis for design, and we suggest these guidelines are suitable for the design of such manuals from scratch.
(November 1990; 41 pages)
Ref. No: HCRC/RP-14 Price: UKL 1.60