Applying Argumentative Zoning in an Automatic Critiquer of Academic Writing

This paper presents an automatic critiquer of Computer Science Abstracts in Portuguese, which formulates critiques and/or suggestions of improvement based on automatic argumentative structure recognition. The recognition is performed by an statistical classifier, similar to Teufel and Moens’s Argumentative Zoning (AZ) [1], but ported to work on Portuguese abstracts. The critiques and suggestions made by the system come from a set of fixed critiquing rules based on corpus observations and guidelines for good writing from the literature. Here we describe the overall system and report on the AZ porting exercise, its intrinsic evaluation and application in the critiquer.