Text Generation
Résumé
The Hyperdocsy ESPRIT Exploratory Action (11/1990-09/1991) studied the automatic production of technical documentation. The documentation of the project includes three main parts: Domain Modelling Analysis (task 1), Text Generation (task 2) and Outline of a Future Hyperdocsy System (task 3). This document is the report on task 2, Text Generation. It is organized in two main parts. The first part deals with corpus analysis, with a view to identify the obstacles to be dealt with when generating text. Different aspect of the corpus are analyzed: topics and rhetorical structure, communicative organization, syntax, semantics, lexicon. The second part deals with the evaluation of existing text generation techniques. It compares two models: Unification Categorial Grammar (UCG) and the Meaning-Text Theory (MTT), and proposes general criteria for evaluating text generation systems.
Origine : Fichiers produits par l'(les) auteur(s)