Expansion von Ereignis-Propositionen zur Visualisierung - Die Grundlagen der begrifflichen Analyse von ANTLIMA
GWAI 90 - German Workshop on AI, volume 251 of Informatik-Fachberichte, page 246-256. Berlin, Springer, (1990)Abstract
ANTLIMA, the listener model of the system SOCCER, anticipates the listeners' mental images of events communicated by the system. To that purpose, compound event descriptions have to be mapped onto sequences of elementary spatio-temporal relations. Although an event proposition like double pass between player 5 and player 7 looks quite simple, it refers to a large set of possible sequences of event phases. To recognize events, SOCCER already uses the knowledge of which phases have to occur in which temporal order to make something an instance of double pass. I will show that this knowledge can be used for generating images as well. Moreover, the use of this knowledge ``in reverse'' within ANTLIMA's conceptual analysis can give hints for improving SOCCER's event recognition: during event expansion, t-TyPoFs, i.e., temporal restrictions in the form of typicality distributions that describe the possible durations of event phases, are necessary to find the concrete durations of the phases under consideration. They can also be used during recognition to distinguish more or less typical instances of an event type.
Description
dblp