Sentence Bottleneck Autoencoders from Transformer Language Models

TL;DR

This paper introduces a sentence-level autoencoder built from a pretrained transformer, improving text representation quality for various NLP tasks with fewer parameters by adapting masked language modeling into a denoising objective.

Contribution

It presents a novel sentence autoencoder architecture that leverages frozen pretrained transformers and a denoising objective, enhancing representation quality and efficiency.

Findings

Outperforms previous text similarity and classification methods

Uses fewer parameters than large pretrained models

Achieves better results on style transfer and GLUE tasks

Abstract

Representation learning for text via pretraining a language model on a large corpus has become a standard starting point for building NLP systems. This approach stands in contrast to autoencoders, also trained on raw text, but with the objective of learning to encode each input as a vector that allows full reconstruction. Autoencoders are attractive because of their latent space structure and generative properties. We therefore explore the construction of a sentence-level autoencoder from a pretrained, frozen transformer language model. We adapt the masked language modeling objective as a generative, denoising one, while only training a sentence bottleneck and a single-layer modified transformer decoder. We demonstrate that the sentence representations discovered by our model achieve better quality than previous methods that extract representations from pretrained transformers on text similarity tasks, style transfer (an example of controlled generation), and single-sentence classification tasks in the GLUE benchmark, while using fewer parameters than large pretrained models.