A coherence analysis module for SciPo: Providing suggestions for scientific abstracts written in Portuguese

Abstract

SciPo is a writing tool whose ultimate goal is to assist novice writers in producing scientific writing in Portuguese, focusing primarily on abstracts and introductions from computer science. In this paper, we describe the development of a coherence analysis module for SciPo, which aims to automatically detect semantic coherence aspects of abstracts and provide suggestions for improvement. At the core of this new module are classifiers that identify different semantic relationships among sentences within an abstract and hence indicate semantic aspects that add coherence to the abstract. Such classifiers are based on a set of features extracted automatically from the surface of the text and by Latent Semantic Analysis processing. All classifiers were evaluated intrinsically and their performance was higher than the baseline. We also resorted to actual users to evaluate our coherence analysis module, which has been incorporated into the SciPo system, and results demonstrate its potential to help users write scientific abstracts with a higher level of coherence. © 2012 The Brazilian Computer Society.