Abstract
PoeTryMe is an intelligent platform for the automatic generation of poetry, with a versatile architecture that provides a high level of customisation. Before generating a poem, the user can define many parameters, including the poetry form, the line templates, the available semantic knowledge, the initial seed words that define the generation domain, the transmitted sentiment, and the generation strategy. In the end, PoeTryMe can explain the choice of words and lines. The core constituents of PoeTryMe are: (i) a semantic network, with words and semantic relations; (ii) a grammar with line templates which are semantic relation renderings; (iii) a sentence generator that uses the previous; (iv) a tool for perfoming syllable related operations; (v) a contextualizer for explaining the choices of words and lines. PoeTryMe is developed in the Center of Informatics and Systems of the University of Coimbra, Portugal, mainly by Hugo Gonçalo Oliveira, and followed the research on Tra-la-Lyrics, a system that generated Portuguese text, based on rhythm. The first prototype of PoeTryMe dates from 2009, but it has been the target of several developments and instantiations over the last years. It has been used, mainly, to generate poetry in Portuguese but, in 2014, it was adapted to Spanish and, in 2015, to English. It has also been used to produce song lyrics, poetry (vaguely) inspired in Twitter, or based on facts referred in textual documents.
Follow @poetartificialArchitecture
