• 대충 정리해보기
  • 실험 벤치마킹용
  • Dataset, evaluation metrics 중심으로

https://github.com/OpenGVLab/LLaMA-Adapter

Make any large language model a better poet

• Infilling

  • **Enabling Language Models to Fill in the Blanks (ACL 2020)**

• Ormazabal et al., PoeLM

  https://arxiv.org/abs/2205.12206

  • control codes to encode both the number of syllables in a line and the ending syllable

• Chakrabarty et al., CoPoet

  https://arxiv.org/abs/2210.13669

  • Instruction-tuned poetrywriting assistant
  • leverages the power of various T5 models

• Most Language Models can be Poets too: An AI Writing Assistant and Constrained Text Generation Studio

• Haven kim, Singable Lyric Translation

  • A Computational Evaluation Framework for Singable Lyric Translation
  • K-pop Lyric Translation: Dataset, Analysis, and Neural Modeling

• in Kento watanabe et al., 2023: Text-to-lyrics generation with image-based semantics and reduced risk of plagiarism

  • melody constrained lyric generation
    • A melody-conditioned lyrics language model, NAACL 2018
    • A syllable-structured, contextually-based conditionally generation of Chinese lyrics, PRICAI, 2019
    • Melody-conditioned lyrics generation with SeqGANs, ISM, 2020
    • Automated generation of Chinese lyrics based on melody emotions, 2021
    • AI-Lyricist: Generating music and vocabulary constrained lyrics, ACMMM, 2021
    • SongMASS: Automatic song writing with pre-training and alignment constraint, AAAI, 2021
  • Say what? collaborative pop lyric generation using multitask transfer learning, HAI, 2021
    • Ram et al., T5 model based single-line lyrics generation from several lines of input lyrics
    • control rhyme and syllable count of generated lyrcs by adding special tokens at the end of the input sentence
  • QiuNiu: A chinese lyrics generation system with passage-level input
    • Capture the semeantic content of the input text
    • They collected lyrics and passage-level data (such as short novels and essays) separately and utilized unsupervised machine translation framework
    • They prepared two encoder-decoders, one for lyric text and one for passage-level text. then, aligned the latent representation space of these two enc-dec to build a text-to-lyrics encoder-decoder.

• UniLG: A Unified Structure-aware Framework for Lyrics Generation (ACL 2023)

• Unsupervised Melody-to-Lyric Generation (ACL 2023)

• LOAF-M2L: Joint Learning of Wording and Formatting for Singable Melody-to-Lyric Generation