From TUNA attribute sets to Portuguese text
Daniel Bastos Pereira, Ivandré Paraboni · 2008
This document describes the development of a surface realisation component for the Portuguese language that takes advantage of the data and evaluation tools provided by the REG-2008 team. At this initial stage, our work uses simple n-gram statistics to produce descriptions in the Furniture domain, with little or no linguistic variation. Preliminary results suggest that, unlike the generation of English descriptions, contextual information may be required to account for Portuguese word order.