Correcting Errors Using the Framework of Argumentation: Towards Generating Argumentative Correction Propositions from Error Annotation Schemas
Marie Garnier, Arnaud Rykner, Patrick Saint‐Dizier · Institutional Repositories DataBase (IRDB) · 2009
This paper presents a first step towards the automatic generation of argumentative responses to accompany the corrections proposed by a correction and writing-aid system.This system focuses on pairs of languages (e.g.French speakers writing in English), and incorporates a strong didactic orientation.We show how, in case several corrections are available, error annotations can be used to design argumentations weighing the pros and cons of each correction.Argumentation is paired with decision theory in order to help the user pick out the most appropriate correction.Argumentative responses produced manually are used to create the generation schemas required to implement the automatic generation of such texts in the future.