Computational analysis and synthesis of song lyrics
| Thesis title in Czech: | Automatická analýza a syntéza písňových textů |
|---|---|
| Thesis title in English: | Computational analysis and synthesis of song lyrics |
| Key words: | texty písní|automatická evaluce|detekce rýmů|generování textů|GPT-2 |
| English key words: | song lyrics|automatic evaluation|rhyme detection|lyrics generation|GPT-2 |
| Academic year of topic announcement: | 2020/2021 |
| Thesis type: | diploma thesis |
| Thesis language: | angličtina |
| Department: | Institute of Formal and Applied Linguistics (32-UFAL) |
| Supervisor: | Mgr. Martin Popel, Ph.D. |
| Author: | hidden - assigned and confirmed by the Study Dept. |
| Date of registration: | 25.09.2020 |
| Date of assignment: | 25.09.2020 |
| Confirmed by Study dept. on: | 16.03.2021 |
| Date and time of defence: | 02.09.2021 09:00 |
| Date of electronic submission: | 22.07.2021 |
| Date of submission of printed version: | 22.07.2021 |
| Date of proceeded defence: | 02.09.2021 |
| Opponents: | Mgr. Rudolf Rosa, Ph.D. |
| Guidelines |
| Among natural language generation tasks, an interesting niche is poetry, including song lyrics of various genres. Poetry, especially when set to music, has relatively strong formal constraints beyond fluency: a verse structure, rhythm, and rhyme on the lower level, and a structure of repetition at a higher level of description. At the same time it can be closer to spoken than written language in lexical choices and tolerance of grammatical irregularities. These factors make poetry perhaps difficult to generate, but at the same time if the constraints on form are observed, there is considerable leeway in terms of content.
The goal of this thesis is to explore generating popular song lyrics of various genres with focus on form. An equally significant part of the thesis is designing adequate automated evaluation methods. While the golden standard of analysis remains human evaluation, it is not feasible to conduct a survey every time the generation system changes during development -- automated methods are therefore necessary. Poetic form, which can be objectively measured, can play an important role in automated evaluation as well. |
| References |
| Delmonte, R., & Prati, A. M. (2014, April). Sparsar: An expressive poetry reader. In Proceedings of the Demonstrations at the 14th Conference of the European Chapter of the Association for Computational Linguistics (pp. 73-76).
Malmi, Eric, et al. "Dopelearning: A computational approach to rap lyrics generation." Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. 2016. Ghazvininejad, Marjan, et al. "Hafez: an interactive poetry generation system." Proceedings of ACL 2017, System Demonstrations. 2017. Zhang, Xingxing, and Mirella Lapata. "Chinese poetry generation with recurrent neural networks." Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP). 2014. |