Clinical narrative classification using discriminant word embeddings with ELM
Paula Lauren, Guangzhi Qu, Feng Zhang, Amaury Lendasse · 2016
Clinical texts are inherently complex due to the medical domain expertise required for content comprehension. In addition, the unstructured nature of these narratives poses a challenge for automatically extracting information. In natural language processing, the use of word embeddings are an effective approach to generate word representations (vectors) in a low dimensional space. In this paper we use a log-linear model (a type of neural language model) and Linear Discriminant Analysis with a kernel-based Extreme Learning Machine (ELM) to map the clinical texts to the medical code. Experimental results on clinical texts indicate improvement with ELM in comparison to SVM and neural network approaches.