We examine cases in which the basic meaning of a word is stretched in context due to the meaning of the co-occurring words, as in open the wine, where the meaning of wine is stretched to that of its container by the co-occurrent verb open. We focus on verb-argument combinations, in which the stretching is triggered by the verb, and targets one of its arguments, as in the example above. We claim that stretching is both constrained and graded, and we validate these assumptions with two types of data: corpus evidence and human judgements. We show that constraints on stretching are not necessarily motivated cognitively, and argue that this represents evidence that there exists a context-independent lexical meaning distinct from conceptual knowledge and pragmatic inference. The overall goal of our contribution is to provide an evidence-based theoretical framework for a variety of approaches (logical, probabilistic), which aim at modelling language taking into account the context sensitivity of word meaning. We focus on the Italian language but our findings and results are valid crosslinguistically.
Stretching the Meaning of Words: Constraints and Gradedness in Semantic Coercion
Jezek E.
2017-01-01
Abstract
We examine cases in which the basic meaning of a word is stretched in context due to the meaning of the co-occurring words, as in open the wine, where the meaning of wine is stretched to that of its container by the co-occurrent verb open. We focus on verb-argument combinations, in which the stretching is triggered by the verb, and targets one of its arguments, as in the example above. We claim that stretching is both constrained and graded, and we validate these assumptions with two types of data: corpus evidence and human judgements. We show that constraints on stretching are not necessarily motivated cognitively, and argue that this represents evidence that there exists a context-independent lexical meaning distinct from conceptual knowledge and pragmatic inference. The overall goal of our contribution is to provide an evidence-based theoretical framework for a variety of approaches (logical, probabilistic), which aim at modelling language taking into account the context sensitivity of word meaning. We focus on the Italian language but our findings and results are valid crosslinguistically.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
