Deep Learning for Text Style Transfer: A Survey

A complete and comprehensive survey paper of Text Style Transfer. It depicts a full picture with well-defined hierarchy of TST:

  • Methods on Parallel Data
    • Multi-Tasking
      • auxiliary tasks (classifier, reconstruction, MT, GEC)
    • Inference Techniques
      • avoid copying too many words from the source sentence, e.g., constrained decoding
    • Data Augmentation
      • MT
  • Methods on Non-Parallel Data (the majority TST methods assume only non-parallel monostyle corpora)
    • Disentanglement
      • Encoder-Decoder Training Method
        • AE, VAE, GAN
      • Latent Representation Manipulation
        • style code contains no semantics and vise versa
      • Training Objectives
        • Style-Oriented Losses
          • The target attribute is fully and exclusively controlled by \mathbf{a} (and not \mathbf{z})
        • Content-Oriented Losses
          • The attribute-independent information is fully and exclusively captured by \mathbf{z} (and not \mathbf{a})
    • Prototype Editing
      • Explicit replace attribute markers
    • Pseudo-Parallel Corpus Construction
      • Retrieval-Based Corpora Construction
        • Find sentence pairs through IR
      • Generation-Based Corpora Construction
        • Train two separate TSTs for both directions and use them to generate silver pairs
  • Research Agenda
  • 5: Transformative: This paper is likely to change our field. It should be considered for a best paper award.
  • 4.5: Exciting: It changed my thinking on this topic. I would fight for it to be accepted.
  • 4: Strong: I learned a lot from it. I would like to see it accepted.
  • 3.5: Leaning positive: It can be accepted more or less in its current form. However, the work it describes is not particularly exciting and/or inspiring, so it will not be a big loss if people don’t see it in this conference.
  • 3: Ambivalent: It has merits (e.g., it reports state-of-the-art results, the idea is nice), but there are key weaknesses (e.g., I didn’t learn much from it, evaluation is not convincing, it describes incremental work). I believe it can significantly benefit from another round of revision, but I won’t object to accepting it if my co-reviewers are willing to champion it.
  • 2.5: Leaning negative: I am leaning towards rejection, but I can be persuaded if my co-reviewers think otherwise.
  • 2: Mediocre: I would rather not see it in the conference.
  • 1.5: Weak: I am pretty confident that it should be rejected.
  • 1: Poor: I would fight to have it rejected.

0 投票者