High Precision Hypertext Navigation Based on NLP Automatic Extractions

The growing popularity of hypertext navigation systems and the availability of large documentary databases is leading to the design of navigation systems that allow to explore these masses of documents. In this article, we present a prototype hypertext system for navigating through the dispatch archives created from Agence France-Presse. Taking account of the volume of documents to be processed, we show how we use natural language processing to produce automatically the data required to construct the hypertext. We describe a data model hinged around the notions of actor, event and role. We show that this model can be used to structure information contained in news archives and to explore them