Learning Disentangled Representations of Texts with Application to Biomedical Abstracts

TL;DR

This paper introduces a novel adversarial method for learning disentangled text representations, specifically applied to biomedical abstracts, enabling better interpretability and aspect-specific retrieval.

Contribution

The paper presents a new adversarial approach to induce disentangled embeddings of texts, focusing on clinical trial abstracts to separate key aspects like populations, interventions, and outcomes.

Findings

Learned representations encode clinically salient aspects

Effective aspect-specific retrieval demonstrated

Method generalizes to other multi-aspect corpora

Abstract

We propose a method for learning disentangled representations of texts that code for distinct and complementary aspects, with the aim of affording efficient model transfer and interpretability. To induce disentangled embeddings, we propose an adversarial objective based on the (dis)similarity between triplets of documents with respect to specific aspects. Our motivating application is embedding biomedical abstracts describing clinical trials in a manner that disentangles the populations, interventions, and outcomes in a given trial. We show that our method learns representations that encode these clinically salient aspects, and that these can be effectively used to perform aspect-specific retrieval. We demonstrate that the approach generalizes beyond our motivating application in experiments on two multi-aspect review corpora.