Understanding novel language

In this article we treat in some detail the problem of designing mechanisms that will allow us to deal with two types of novel language: (1) text requiring scheme learning; and (2) the understanding of novel metaphorical use of verbs. Schema learning is addressed by four types of processes: schema composition, secondary effect elevation, schema alteration, and volitionalization. The processing of novel metaphors depends on a decompositional analysis of verbs into "event shape diagrams," along with a matching process that uses semantic marker-like information, to construct novel meaning structures. The examples we describe have been chosen to be types that occur commonly, so that rules that we need to understand them can also be used to understand a much wider range of novel language.