Detalhes bibliográficos
| Ano de defesa: |
2018 |
| Autor(a) principal: |
Azevedo, Ana Cleyge Silva de
 |
| Orientador(a): |
Loula, Angelo Conrado |
| Banca de defesa: |
Não Informado pela instituição |
| Tipo de documento: |
Dissertação
|
| Tipo de acesso: |
Acesso aberto |
| Idioma: |
por |
| Instituição de defesa: |
Universidade Estadual de Feira de Santana
|
| Programa de Pós-Graduação: |
Mestrado em Computação Aplicada
|
| Departamento: |
DEPARTAMENTO DE CIÊNCIAS EXATAS
|
| País: |
Brasil
|
| Palavras-chave em Português: |
|
| Palavras-chave em Inglês: |
|
| Área do conhecimento CNPq: |
|
| Link de acesso: |
http://tede2.uefs.br:8080/handle/tede/1333
|
Resumo: |
The computational generation of poems has been studied internationally for decades being improved over time. The most recent contributions consider multiple linguistic levels in the generation of poems, particularly phonological parallels, metrical structures of versification, rhythmic patterns and rhyme. On the other side, the presence of metric structure was identified in literary prose and the computational system MIVES (Mining Verse Structure) is able to automatically find such strucutres in prose. Therefore, this work presents the PROPOE (Prose to Poem Automatic Generator), a system capable of receiving the metric sentence data from MIVES to generate poems, filtering sentences to build poems. PROPOE is implemented based on the greedy algorithm, capable of performing combinatorial optimization of verses in stanzas, considering rhythmic aspects such as rhyme and tonic positioning. The poems obtained are evaluated by the system considering multiple criteria to obtain a general score. Considering several prose works as a source of sentences, the results achieved by PROPOE are presented and analyzed in multiple examples. |
