Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Educating Text Autoencoders: Latent Representation Guidance via Denoising

About

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 autoencoders can learn an arbitrary mapping between sequences and associated latent representations. To remedy this issue, we augment adversarial autoencoders with a denoising objective where original sentences are reconstructed from perturbed versions (referred to as DAAE). We prove that this simple modification guides the latent space geometry of the resulting model by encouraging the encoder to map similar texts to similar latent representations. In empirical comparisons with various types of autoencoders, our model provides the best trade-off between generation quality and reconstruction capacity. Moreover, the improved geometry of the DAAE latent space enables zero-shot text style transfer via simple latent vector arithmetic.

Tianxiao Shen, Jonas Mueller, Regina Barzilay, Tommi Jaakkola• 2019

Related benchmarks

TaskDatasetResultRank
AutoencodingMathematical expressions EVAL (test)
BLEU38
22
Sentence Interpolation SmoothnessARGO randomly sampled 200 sentence pairs
Average IS0.055
22
Language modellingExplanatory sentences
BLEU22
19
Language modellingMathematical expression EVAL (test)
Exact Match0.00e+0
19
AutoencodingExplanatory sentences (test)
BLEU22
13
AutoencodingMathematical expressions VAR (test)
BLEU48
11
AutoencodingMathematical expressions EASY (test)
BLEU Score35
11
AutoencodingMathematical expressions LEN (test)
BLEU0.49
11
Explanatory Inference RetrievalWorldTree 1.0 (test)
MAP (t=1)13.16
7
Showing 9 of 9 rows

Other info

Follow for update