In this paper, we shall deal with a systemudfor text generation and understanding calledudGETA_RUN (GEneration of Text and Analysisudfor Reference Understanding). Dialogues in ourudsystem may be regarded as a sequence ofudquestion-answering actions with the aim toudretrieve as much information as possible from audtext. In a certain sense the system acts as a veryudintelligent information retrieval system whichudallows to recover the contents of a text on theudbasis of what is actually being linguisticallyudexpressed. No implicit information is recoverableudfrom the system, however.udThe system works by encoding theudlinguistic contents of a text fully analyzed in theudDiscourse Model where all entities, propertiesudand relations are recorded. Further semanticudproperties and relations are then transferred into audKnowledge Representation Module runningudunder KL-One to be used for consistency checkudwhenever the dialogue is started up.
展开▼
