Convolutional neural networks for chemical-disease relation extraction are improved with character-based word embeddings

TL;DR

This paper demonstrates that integrating character-based word embeddings into CNN and LSTM models significantly enhances chemical-disease relation extraction, achieving state-of-the-art results on the BioCreative-V CDR corpus.

Contribution

It introduces the use of character-based word representations into neural relation extraction models, improving their performance over previous approaches.

Findings

Character-based embeddings improve relation extraction accuracy

Models achieve state-of-the-art results on BioCreative-V CDR corpus

Both CNN and LSTM architectures benefit from character-level information

Abstract

We investigate the incorporation of character-based word representations into a standard CNN-based relation extraction model. We experiment with two common neural architectures, CNN and LSTM, to learn word vector representations from character embeddings. Through a task on the BioCreative-V CDR corpus, extracting relationships between chemicals and diseases, we show that models exploiting the character-based word representations improve on models that do not use this information, obtaining state-of-the-art result relative to previous neural approaches.