Abstracts of HCRC Research papers, 1993


Jim Miller and Regina Weinert:
The Function of LIKE in Dialogue
The paper investigates the function of LIKE in two bodies of data, the Map Task dialogues and spontaneous conversations. LIKE is analysed as a highlighting or focusing device, rather than a randomly occurring item devoid of semantic or pragmatic significance and functioning as a mere filler. LIKE is analysed as a non-contrastive focuser. Two major functions are distinguished. Clause-final LIKE counters objections and assumptions and is found only in the spontaneous conversations. In other syntactic positions LIKE elucidates previous comments and is associated in the Map Task dialogues with the game moves of instructing, aligning and checking. Previous treatments of LIKE are shown to be inadequate.
(January 1993; 24 pages)
Ref. No: HCRC/RP-40 Price: UKL 1.10

Regina Weinert and Jim Miller:
Cleft constructions in spoken language
Previous accounts of clefts in terms of focus are too narrow, and the presentation of a novel, maximal instantiation of a variable becomes a background property. The focusing function of clefts is crucially related to thematization and deixis and not to the givenness of the WH clause. The focusing function is separate from the information carried by the cleft components. IT and reverse clefts function as much to specify the clefted constituent as the variable in the WH clause. The macro-textual functions of the clefts are determined by the nature of the deictic. Reverse clefts, with a TH deictic, highlight the immediately preceding discourse, delay its progress and consolidate the exchange of information. WH clefts, with an indefinite WH deictic, point forward and propel the discourse onwards. IT clefts are neutral with respect to direction and are preferred for the expression of contrast.
(February 1993; 34 pages)
Ref. No: HCRC/RP-41 Price: UKL 1.30

Robin Cooper:
A note on the relationship between linguistic theory and linguistic engineering
We indicate some aspects of the relationship between theoretical work on natural language and linguistic engineering. We are concerned with the kind of theoretical work being undertaken in the {\sc dyana project (``Dynamic Interpretation of Natural Language''), an {\sc esprit basic research action which began its second phase in October 1992. In particular we try to relate this work to stochastic approaches to linguistic engineering, which initially might seem diametrically opposed to the kind of work undertaken in {\sc dyana. We will argue, however, that the theoretical and stochastic approaches may be seen as making complementary contributions to the future needs of linguistic engineering.
(February 1993; 10 pages)
Ref. No: HCRC/RP-42 Price: UKL 0.70

Jean Carletta, Richard Caley and Stephen Isard:
A System Architecture for Simulating Time-Constrained Language Production
The goal of our research is to simulate the human production of language under time constraints. In this paper, we briefly discuss two behaviours arising from time pressure, {\it hesitation and {\it spontaneous self-repair, which we have identified from a corpus of human dialogues. We then go on to describe a system architecture, adapted from work in real-time systems and Levelt's model for a speaker, which we intend to use in building the simulator.
(February 1993; 10 pages)
Ref. No: HCRC/RP-43 Price: UKL 0.70

Enric Vallduvi:
Information packaging: A survey
Information packaging is the speaker's manipulation of sentence structure in order to present propositional content to the hearer in different ways according to his or her assumptions about the knowledge and attentional state of the latter. Different languages exploit word order and prosody in different ways to express information packaging. In order to establish the basic range of variation in the realization of information packaging across languages, a set of of informational primitives that are crosslinguistically sufficient and methodologically useful need to be identified. To this end, this report reviews several earlier approaches and introduces a remodelled informational description of the sentence that draws from and improves on these approaches and shows how these informational distinctions are structurally manifested in several languages, with an emphasis on Catalan, Dutch, and English. It also contains a discussion of the structural effects of other pragmatic domains (referential status, presupposition) and their interaction with information packaging, and of some related issues concerning the use of intonation.
(August 1993; 43 pages)
Ref. No: HCRC/RP-44 Price: UKL 1.60


J. A. Taylor:
An ATMS-based Belief Model for Dialogue Simultation
This paper describes a belief model that has been developed for use in a computational model of dialogue. The model combines a basis in logic with restrictions on the propositions and reasoning steps that can be supported, resulting in a system that supports the type of reasoning that people seem to carry out subconsciously during simple, task-oriented dialogues.
The notion of vivid acquaintance with concepts is explained as a shorthand for propositional information. The effects of utterances on agents' degrees of acquaintance and beliefs about one another are described, and a simulated dialogue from the Map Task is presented.
(October 1993; 20 pages)
Ref. No: HCRC/RP-45 Price: UKL 1.10


Jonathan Ginzburg:
Resolving Questions
Interrogative sentences have two prominent uses: they can serve to name {\em questions}, the descriptive contents of queries (`So tell me: who left yesterday?'), but also to describe, roughly, the true and complete {\em answer} to a query use (`She told me who left yesterday.'). The paper develops a theory of questions, formulated within a situation theoretic setting, that attempts to explain the source of this duality without appealing to an ambiguity, as most past accounts have. The key notion developed is that of a question's being {\bf resolved} or {\bf unresolved}.
Resolvedness turns out to have a wide range of applications: for instance, it provides a means for defining an evaluation criterion for the optimality of a response to a query; resolvedness provides a suitably context sensitive notion of `exhaustiveness', the proper characterisation of which has been a long standing issue in the semantics of interrogatives: for instance, what information an agent will describe as constituting `being aware where I am' will vary depending on whether the agent is getting off an airplane (`I am in Helsinki' will do fine.) or getting out of a taxi (`I am in Helsinki' will definitely not do.); It is shown here that various interrogative embedding predicates (`resolutive predicates' e.g. know,tell,discover) carry the presupposition that their complement is a {\em resolved} question and shown the existence of strong parallels with {\em factivity}, where predicates carry the presupposition that their declarative complement is a {\em true} proposition.
(October 1993; 57 pages)
Ref. No. HCRC/RP-46 Price: UKL 1.90