A spreading activation approach to text illustration

摘要

In this paper we present a new approach to implement intelligent multimedia interfaces. Its central elements are a media-independent formal representation of the presented knowledge and media-specific realization statements. Reference hypotheses for media objects are established automatically. Subsequently, the reference hypotheses are validated and weighted by a spreading activation algorithm. Moreover, the spreading activation algorithm determines those entities of the formal representation which are relevant considering an interactively modified multimedia presentation. Finally, the information on the relevance and on the co-reference of media objects are exploited to guarantee the coherency of the multimedia presentation.This paper is focused on the automatic generation of illustrations which are adjusted to the content of interactively selected text segments. Initial results of two application systems, the TEXTILLUSTRATOR and VISDOK, are presented.