Generování textů českých coververzí anglických písní
| Thesis title in Czech: | Generování textů českých coververzí anglických písní |
|---|---|
| Thesis title in English: | Generation of Czech Lyrics to Cover Songs |
| Key words: | zpracování přirozeného jazyka|generování textu|literární NLP|strojový překlad |
| English key words: | natural language processing|text generation|literary NLP|machine translation |
| Academic year of topic announcement: | 2023/2024 |
| Thesis type: | Bachelor's thesis |
| Thesis language: | |
| Department: | Institute of Formal and Applied Linguistics (32-UFAL) |
| Supervisor: | Mgr. Rudolf Rosa, Ph.D. |
| Author: | Barbora Štěpánková - assigned and confirmed by the Study Dept. |
| Date of registration: | 03.04.2024 |
| Date of assignment: | 04.04.2024 |
| Confirmed by Study dept. on: | 04.04.2024 |
| Guidelines |
| The goal of the work is to explore automatic generation of Czech lyrics for songs in English language, i.e. lyrics for Czech cover versions of English songs.
Necessary components of the work include: - Gathering data. Optimal data for the work are pairs of English language song lyrics and their existing Czech cover versions. As such data are probably rather scarce, it is expected that other kinds of data will be useful as well, at least monolingual Czech and English song lyrics. The student will gather appropriate data and process them (filtering, format conversion, possible automated translation, etc.) so that they are usable for the work. Existing corpora may be used if available. - Analysing the problem. The topic seems to be quite under-researched, it is therefore not obvious which aspects of songs and lyrics are most important when creating a cover version (e.g. rhythm, number of syllables, rhyming, meaning, etc.) and what approaches are most promising for generating the output. The student will research existing works, as well as conduct own research by analyzing existing Czech covers together with their English originals. - Implementing lyrics processing tools. Based on the analyses, it is expected that some tools for processing lyrics will need to be implemented, such as syllable splitting or rhyming analyzer. Existing tools may be used if available. - Designing and implementing a lyrics generator. The ultimate goal is to implement a tool that will take English lyrics on the input and produce corresponding Czech cover lyrics on the output. It is expected that language modeling techniques and/or machine translation techniques will be used. As the problem is under-researched, it is not expected that the results will be perfect; even a negative result is possible, i.e. finding that generating high-quality cover lyrics is not possible with the available data and tools and with the selected approach. However, in any case, the results should be manually evaluated and analyzed, including a discussion of the results with respect to the goals of the work. |
| References |
| GUO, Fenfei, et al. Automatic Song Translation for Tonal Languages. In: Findings of the Association for Computational Linguistics: ACL 2022. 2022. p. 729-743.
WATANABE, Kento, et al. A melody-conditioned lyrics language model. In: Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers). 2018. p. 163-172. OU, Longshen, et al. Songs Across Borders: Singable and Controllable Neural Lyric Translation. In: Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023. p. 447-467. |