Convolutional Gated Recurrent Units for Medical Relation Classification

TL;DR

This paper introduces a unified CNN-GRU model that effectively captures both local and long-term features for medical relation classification in clinical records, outperforming previous single-model approaches.

Contribution

The paper presents a novel architecture combining CNN and bidirectional GRU for improved medical relation classification using only word embeddings.

Findings

Model outperforms previous methods on clinical datasets

Effective integration of CNN and GRU captures phrase-level and long-term dependencies

Significant performance improvements demonstrated in experiments

Abstract

Convolutional neural network (CNN) and recurrent neural network (RNN) models have become the mainstream methods for relation classification. We propose a unified architecture, which exploits the advantages of CNN and RNN simultaneously, to identify medical relations in clinical records, with only word embedding features. Our model learns phrase-level features through a CNN layer, and these feature representations are directly fed into a bidirectional gated recurrent unit (GRU) layer to capture long-term feature dependencies. We evaluate our model on two clinical datasets, and experiments demonstrate that our model performs significantly better than previous single-model methods on both datasets.