An Analysis of Sentence Boundary Detection Systems for English and Portuguese Documents

In this paper we present a study comparing the performance of different systems found in the literature that perform the task of automatic text segmentation in sentences for English documents. We also show the difficulties found to adapt these systems to make them work with Portuguese documents and the results obtained after the adaptation. We analyzed two systems that use a machine learning approach: MxTerminator and Satz, and a customized system based on fixed rules expressed by Regular Expressions. The results achieved by the Satz system were surprisingly positive for Portuguese documents.