Issues in the representation of real texts: the design of KRISP
Natural language processing and knowledge representation
Improving the accessibility of line graphs in multimodal documents
SLPAT '11 Proceedings of the Second Workshop on Speech and Language Processing for Assistive Technologies
| Hi-index | 0.00 |
A significant factor in the complexity of the compressed, complex prose style used by journalists in short, targeted commercial reports (Who's News, joint ventures, earnings reports, etc.) is the fact that many of the phrases are semantically incomplete, i.e. their interpretation is dependent on information in other parts of the sentence or the in discourse context. We propose that the complexity that such partially saturated referents contribute to the overall process of semantic interpretation can be characterized by two factors we will call displacement and unpacking. This complexity source can be quantified by counting the distance, in nodes, between each phrase that has a locally incomplete interpretation and the phrase(s) that supply the terms that complete them. In this paper we will define this phenomenon and illustrate its impact on interpretation by examining short texts excerpted from the Tipster corpus and other online sources.