Acceso usuarios
The goal ofthis paper, devoted to automatically generated texts, is twofold. First,and more importantly,we aim at presenting a methodology allowing to describe and assess the 'quality'of these texts, as well as uncover some of their specificities compared to non-generated texts. Second, we want to identify a series of relevant features related to both the lexicon and the macro-and micro-structuring of these texts.Based on the application of our methodology to a small workcorpus of 100 articles produced in French by the CAC40 software in the field of finance, we conclude that, taken individually, each generated text shows a sufficiently rich internal variation to be perceived as natural.
