Corpus ID: 211069137

Educating Text Autoencoders: Latent Representation Guidance via Denoising

@article{Shen2019EducatingTA,
  title={Educating Text Autoencoders: Latent Representation Guidance via Denoising},
  author={Tianxiao Shen and Jonas Mueller and Regina Barzilay and Tommi S. Jaakkola},
  journal={arXiv: Learning},
  year={2019}
}
  • Tianxiao Shen, Jonas Mueller, +1 author Tommi S. Jaakkola
  • Published 2019
  • Mathematics, Computer Science
  • arXiv: Learning
  • Generative autoencoders offer a promising approach for controllable text generation by leveraging their latent sentence representations. However, current models struggle to maintain coherent latent spaces required to perform meaningful text manipulations via latent vector operations. Specifically, we demonstrate by example that neural encoders do not necessarily map similar sentences to nearby latent vectors. A theoretical explanation for this phenomenon establishes that high-capacity… CONTINUE READING

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 36 REFERENCES

    Adversarially Regularized Autoencoders

    VIEW 5 EXCERPTS
    HIGHLY INFLUENTIAL

    Denoising Adversarial Autoencoders

    Variational Lossy Autoencoder

    Generating Sentences from a Continuous Space

    VIEW 5 EXCERPTS
    HIGHLY INFLUENTIAL