Calculating the Upper Bounds for Portuguese Automatic Text Summarization Using Genetic Algorithm

摘要

Over the last years, Automatic Text Summarization (ATS) has been considered as one of the main tasks in Natural Language Processing (NLP) that generates summaries in several languages (e.g., English, Portuguese, Spanish, etc.). One of the most significant advances in ATS is developed for Portuguese reflected with the proposals of various state-of-art methods. It is essential to know the performance of different state-of-the-art methods with respect to the upper bounds (Topline), lower bounds (Baseline-random), and other heuristics (Baseline-first). In recent works, the significance and upper bounds for Single-Document Summarization (SDS) and Multi-Document Summarization (MDS) using corpora from Document Understanding Conferences (DUC) were calculated. In this paper, a calculus of upper bounds for SDS in Portuguese using Genetic Algorithms (GA) is performed. Moreover, we present a comparison of some state-of-the-art methods with respect to the upper bounds, lower bounds, and heuristics to determinate their level of significance.