Medical Concept Normalization In Social Media Posts With Recurrent Neural Networks

Text mining of scientific libraries and social media has already proven itself as a reliable tool for<br>
drug repurposing and hypothesis generation. The task of mapping a disease mention to a concept<br>
in a controlled vocabulary, typically to the standard thesaurus in the Unified Medical Language<br>
System (UMLS), is known as medical concept normalization. This task is challenging due to the<br>
differences in medical terminology between health care professionals and social media texts coming<br>
from the lay public. To bridge this gap, we use sequence learning with recurrent neural networks<br>
and semantic representation of one- or multi-word expressions: we develop end-to-end architectures<br>
directly tailored to the task, including bidirectional Long Short-Term Memory and Gated Recurrent<br>
Units with an attention mechanism and additional semantic similarity features based on UMLS.<br>
Our evaluation over a standard benchmark shows that recurrent neural networks improve results<br>
over an effective baseline for classification based on convolutional neural networks. A qualitative<br>
examination of mentions discovered in a dataset of user reviews collected from popular online health<br>
information platforms as well as quantitative evaluation both show improvements in the semantic<br>
representation of health-related expressions in social media.