The Formalisation of the Internal Structure of Events in Language Texts

Theories involving descriptions of mental or physical events in language can be described as either the external structure of events, which model the prepositional relationships between events through logical frameworks; theories of events as change, which model events as changes to situations by decomposing the meaning of verb phrases into semantic primitives; and the internal structure of events, which attempt to classify types of events based on the internal temporal properties of a verb phrase.
The external structure of events can be seen in temporal reasoning systems in Artificial Intelligence which develop logical or axiomatic representations of temporal prepositional phrases that indicate relationships between events without considering the internal temporal properties of a reported event. As a result, incompatible point and interval based approaches for actions in time have been derived to represent events.
The identification of the internal structure of events in the Linguistic Philosophy tradition, can support temporal reasoning systems by identifying the temporal nature of a reported event and indicating the most suitable temporal representation for that event. However, these approaches rely on intuitive human judgements to identify the temporal properties of a verb phrase and have two fundamental limitations; the informal approach used to identify an event is subject to human interpretation, allowing events to be classified differently depending on the reader; and the identification of event type, based on the verb phrase alone, ignores other linguistic information that affects the event classification, resulting in an incomplete analysis of a reported event.
This thesis demonstrates that a formal method for identifying the type of event reported in a sentence is possible, based on deriving rules that incrementally set the state of binary features representing temporal information conveyed by linguistic elements in that sentence. The final combination of these binary features indicate the type of event described in a sentence according to an extended classification of events based on Vendler’s analysis, thereby avoiding the problems noted in the Linguistic Philosophy approaches.