Integrating word embedding neural networks with pubmed abstracts to extract keyword proximity of chronic diseases

Chronic diseases are a leading cause of morbidity and mortality worldwide. They are common enough to affect large numbers of patients, and the chronic nature makes them costly to both patients and healthcare providers. Diagnosis of many chronic diseases is challenged by variability in their clinical manifestations. Although chronic diseases bear a set of structured terminology aiming to standardize nomenclature of the presentation and outcomes of the disease, in practice there is a wide spectrum of terminology associated with these diseases across different venues such as clinical notes, biomedical literature, and health-related social media. Among these sources, the scientific articles published in the biomedical literature usually follow principled approaches to terminology and are thus especially valuable for extracting diseases keywords. Given the fact that it is very costly and time-consuming to manually extract disease terminology from a large column of scientific articles, we aim to utilize artificial neural network strategies to automatically extract vocabularies associated with a set of chronic diseases. Our finding indicates the feasibility of developing word embedding neural nets for autonomous keyword extraction and abstraction of chronic diseases.