Disentangling Representations of Text by Masking Transformers

Abstract

Representations from large pretrained models such as BERT encode a range of features into monolithic vectors, affording strong predictive accuracy across a multitude of downstream tasks. In this paper we explore whether it is possible to learn disentangled representations by identifying existing subnetworks within pretrained models that encode distinct, complementary aspect representations. Concretely, we learn binary masks over transformer weights or hidden units to uncover subsets of features that correlate with a specific factor of variation; this eliminates the need to train a disentangled model from scratch for a particular task. We evaluate this method with respect to its ability to disentangle representations of sentiment from genre in movie reviews, "toxicity" from dialect in Tweets, and syntax from semantics. By combining masking with magnitude pruning we find that we can identify sparse subnetworks within BERT that strongly encode particular aspects (e.g., toxicity) while only weakly encoding others (e.g., race). Moreover, despite only learning masks, we find that disentanglement-via-masking performs as well as -- and often better than -- previously proposed methods based on variational autoencoders and adversarial training.

arXiv comment14 pages, 9 figures

References in corpus (9)

• Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation
• Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations
• Deep Convolutional Inverse Graphics Network
• Disentangling by Factorising
• Towards a Definition of Disentangled Representations
• Flexibly Fair Representation Learning by Disentanglement
• HUBERT Untangles BERT to Improve Transfer across NLP Tasks
• Structured Neural Topic Models for Reviews
• Disentangled Representations for Manipulation of Sentiment in Text