BDB Bertsolaritzaren datu-basea

In this paper we present the preliminary work of a Basque poetry generation system. Basically, we have extracted the POS-tag sequences from some verse corpora and calculated the probabil- ity of each sequence. For the generation process we have defined 3 different experiments: Based on a strophe from the corpora, we (a) replace each word with other according to its POS-tag and suffixes, (b) replace each noun and adjective with another equally inflected word and (c) re- place only nouns with semantically related ones (inflected). Finally we evaluate those strategies using a Turing Test-like evaluation.