|
Abstract:
|
During the lifecycle of drug development, much of detailed information is embedded in narrative texts, such as clinical trial protocols, drug labels, adverse events reports, which makes it difficult for downstream computerized applications. Natural language processing (NLP) provides automated methods to extract structured information from such unstructured data sources, and it has received great attention in the biomedical domain. In this paper, we will describe our recent efforts on applying deep learning-based NLP approaches to unlock information from diverse narrative documents generated during drug development, including parsing eligibility criteria sections of clinical trial protocols, standardizing adverse events from drug labels, as well as extracting information from adverse event reports. In addition, we will discuss the lessons learned from those projects and future directions of NLP in drug development.
|
