Underspecified changes: a dynamic, probabilistic frame theory for verbs

The verb ‘rise’ can be used both with property-denoting nouns like ‘temperature’but also with NPs like ‘a Titan’ or ‘China’. Whereas in the former case the change triggeredby a rising event is directly related to the subject (its current value increases), this does nothold for ‘a titan’ or ‘China’. In this case it is a property of these objects, say their height ortheir political power, which increases in value. Furthermore, ‘rise’ does not target a particularproperty as the examples above show. This data has led Cooper (2010) to the conclusion thatit is presumably not possible (i) “to extract a single general meaning of words which covers allthe particular meanings of the word in context”, and (ii) “to determine once and for all the setof particular contextually determined meanings of a word”. In this article we present a solutionto the two problems raised by ‘rise’ in a frame theory. ‘Rise’ is analyzed as a scalar verb whichdoes not lexicalize a complete scale in its meaning. Rather, it shows underspecification relativeto the dimension (property) parameter of a scale. The set of admissible properties is determinedby a constraint on the value ranges of properties. If the property is not uniquely determinedby the subject, the comprehender uses probabilistic reasoning based on world knowledge anddiscourse information to defeasibly infer the most likely candidates from this set (2nd problem).The first problem is solved not by simply introducing objects into the representation of adiscourse but instead by pairs consisting of an object and an associated frame component whichcollects the object information contributed by the discourse. Changes triggered by events likethe one denoted by ‘rise’ are modelled as update operations on the frame component while theobject component is left unchanged.Keywords: lexical semantics, scalar changes, frame theory, probabilities.