Poet's Little Helper: A methodology for computer-based poetry generation. A case study for the Basque language

摘要

We present Poet's Little Helper (PLH), a tool that implements a methodology to generate poetry using minimal language-dependent information. The user only needs to provide a corpus with a set of sentences, a rhyme checker and a syllable-counter. From these building blocks, PLH produces: (1) an exploratory analysis of the suitability of the given corpus for poetry generation. (2) a novel and non-trivial poem grammatically correct under metrical and rhyming constraints. This poem also shows content that is coherent with a topic given by the user. The process of poetry generation is a cycle with three phases: lexical exploratory analysis, semantic exploratory analysis and poem generation. The goal is twofold: on the one hand PLH aims to be a useful open source poem-generator for many languages with minimal effort; on the other hand, analizes how the particularities of each corpus affect in the creation of poems. The presented PHL tool is offered in a public repository. The results of an experiment with a corpus of Basque texts is shown.