QUOTATION EXTRACTION FOR PORTUGUESE
William Paulo Ducca Fernandes · 2012
Quotation extraction consists of identifying quotations a nd their au- thors. In this work, we present a Quotation Extraction syste m for Portuguese that is based on Entropy Guided Transformation Learning, a s upervised Ma- chine Learning algorithm. This is the first system that uses aMachine Learning approach for Portuguese. In order to train and evaluate the p system, we build the GLOBOQUOTES corpus, with news extracted from theGLOBO.COM portal. Our system obtains anF�=1 score of 79.02% for the subtask of associ- ating a quotation to its author. For the whole Quotation Extr action task, the observed F�=1 score value is 66.03%. These findings indicate that the overa ll extraction quality is highly dependant on the quotation ide ntification subtask.