Linguistic Adequacy Research Articles

Morphological representation in PC-based text processing systems

This paper summarizes our experience in optimizing representational models of inflexional (accidence) morphological analysis for different languages. Some of the linguistic engineering questions addressed in the paper include the following: How far can linguistic adequacy be maintained in morphological components of marketable machine translation software on the IBM PC platform? What generalizations can be made contributing to the more uniform approach to morphology for languages of various structures? What are the parameters that can be adjusted quantitatively to allow for general model application to different languages? A generalized approach of treating postfixal languages is proposed, which is supplemented by suggestions on its modification for languages that are clearly not postfixal (e.g. Arabic).

Anaphores parallèles et techniques de résolution

C. Gardent, Université de Sarrebruck : « Anaphores parallèles et techniques de résolution » This paper argues that Higher-Order Unification (HOU) is preferable to overwriting techniques for dealing with parallel anaphors i.e. non-nominal anaphors whose resolution is constrained by parallelism. Computationally, we show that overwriting techniques have undesirable properties which make their use incompatible with unification-based grammars. Linguistically, we show that HOU permits a high level of linguistic adequacy in that it achieves a reasonably wide empirical covering (e.g. VP ellipsis, focus, Second Occurrence Expressions) while minimising both under- and over- generation. The HOU approach we propose is systematically compared with comparable existing theories.

Judging appropriacy: the implication of inter‐observer variance

ABSTRACT Four observers, two speech and language therapists and two others, were asked to judge a corpus of utterances obtained from a test of verbal expression in context (The Story Completion Test). The sample came from a group of people suffering from memory‐related communication problems. Five hundred and sixty utterances were judged and reported: (1) Using a formula for criteria of appropriacy, utterances were compared with the target construction; (2) Overall agreement score was calculated against the clinical judgement of the assessing therapist; (3) Inter‐observer variance resulted. The influence of several factors, including the manner of presentation, tester bias, shared knowledge, bridging inference and situation, are discussed. The implication of appropriate judgements in clinical contexts is considered and the question raised as to whether variance, judged more on extra‐linguistic inappropriacy, can mask linguistic adequacy. The potential of bridging inference in interpreting the expressive speech, in particular in patients with dementia, is proposed.

Legislation, Linguistic Adequacy and Public Policy

Legislation, Linguistic Adequacy and Public Policy

On the equivalence of models of language used in the fields of mechanical translation and information retrieval

The purpose of this paper is to compare a certain number of well known models used in the fields of Mechanical Translation (M.T.) and Information Retrieval (I.R.). Different surveys of this type exist (Bar-Hillel [1], Hays [2], Lecerf [3], Sestier and Dupuis [4]), where models have been compared from the point of view of practical and linguistic adequacy. We wish here to compare certain formal characteristics of these models, in fact to show that they are strictly equivalent to a well studied model. The notion of equivalence will be defined formally, the model to which the other models are equivalent is Chomsky's model of context-free languages (c.f. languages). The equivalences discussed here have not only an abstract character; several practical problems which arise naturally are clarified.

The adequacy of language1

he notion of linguistic adequacy (the adequacy of sentences to express or describe) is explicated in terms of a set theoretical model of the communication situation. Roughly: a message is adequate to the degree it answers the receiver's questions. Adequacy is distinguished from openness, in such a way that a message can be both completely adequate in a communication event and also “inexhaustibly open”;. Using this explication it is possible to translate and clarify several familiar philosophical theses concerning the adequacy of language. The view that “humanistic”; language differs from scientific language is clarified and criticized. Seven types of empirical test of the explication are discussed.