From Semantic Properties to Surface Text: the Generation of Domain Object Descriptions
Ivandré Paraboni, Diego Jesus de Lucena, Daniel Bastos Pereira · INTELIGENCIA ARTIFICIAL · 2010
"At both semantic and syntactic levels, the generation of referring expressions (REG) involves far more than simply producing 'correct' output strings and, accordingly, remains central to the study and development of Natural Language Generation (NLG) systems. In particular, REG algorithms have to pay regard to humanlikeness, an issue that lies at the very heart of the classic denition of Articial Intelligence as, e.g., motivated by the Turing test. In this work we present an end-to-end approach to REG that takes humanlikeness into account, addressing both the issues of semantic content determination and surface realisation as natural language."