Dynamic linguistic resources: a recipe for miscommunication
Robin Cooper (University of Gothenburg, SE)
I am a linguistic semanticist by training who has been involved in computational linguistics for many years. My current work takes ideas from type theory and relates them to perception of objects and events and the coordination of action. My main focus is on trying to use this as a basis for accounts of linguistic communication and miscommunication. In this talk I will attempt to show some of the rich possibilities available for miscommunication on this kind of view.
The logician Richard Montague, founder of what came to be known as Montague Grammar, wrote a paper called ‘English as a Formal Language’. His claim was that there was no significant difference between the artificial languages of logicians and natural languages like English and Swedish. In this talk I will suggest that we need to refine this view.
The main intuition is that we regard natural languages not as formal languages but as collections of resources (or “toolboxes”) which can be used to construct local domain specific languages (possibly formal in Montague’s sense) for particular situations of use. We will discuss three areas which seem to support this view:
A frequent assumption in computational and corpus linguistics as well as theoretical linguistics is that words are associated with a fairly small set of meanings, statically defined in a lexical resource. This view is challenged by work in the psychology of language where dialogue participants are regarded as creating meaning on the fly for the purposes of particular dialogues. We argue that a view of lexical meaning in flux is important for natural language analysis. The only constraint on constructing a new meaning for a word is that it should be related to an old meaning for a word in such a way that your interlocutor can figure out the new meaning. This sounds like a recipe for disaster but we are surprisingly good at getting close enough to the right meaning – or at least not noticing when things have gone wrong.
Quotation (work in progress with Jonathan Ginzburg)
In direct quotation we frequently quote people speaking in a way that is different from the way we would normally speak ourselves (e.g. a child who has not yet acquired an adult grammar, a speaker of a dialect other than our own, a quotation in a foreign language). From the formal language perspective these are puzzling examples which suggest that such quotations are extra-grammatical. From the toolbox perspective we can consider them as representing a combination of different linguistic resources into an ad hoc grammar for a given occasion. Direct quotation involves demonstrating a previous utterance event by imitating a certain aspect of it to a certain degree of similarity. This sounds like a recipe for disaster. How do we know which aspect and which degree of similarity is involved?
Enthymematic reasoning (work in progress by Ellen Breitholtz)
In a soon to be completed PhD thesis, Breitholtz formulates a version of Aristotle’s notions of topos and enthymeme to represent the kind of defeasible inference rules which speakers seem to rely on in dialogue. As Ducrot has pointed out the topoi (or inference rules) that we have available are not necessarily consistent with each other. That is, they do not add up to a single monolithic consistent system. We try to maintain a form of consistency within a single dialogue, however. This suggests a kind of pragmatics based on the toolbox idea where topoi are seen as rhetorical resources that can be used to construct reasoning systems local to given dialogues or domains. This sounds like a recipe for disaster. The space of topoi (or inference patterns) that are available is huge and largely inexplicitly represented in dialogue. And yet we are surprisingly good, at least at believing we have understood what topoi our interlocutor is basing her argument on. Is it just an illusion of communication amidst a mass of miscommunication?