Towards Extracting Structured Drug Information from Raw Texts using Deep Learning
Stefan Cristian Hantig, Radu Răzvan Slăvescu, Kinga Cristina Slăvescu · 2020
This paper presents a technique that extracts relevant information from a medical prescription. In order to solve this problem we have approached the problem from Name Entity Recognition (NER) perspective, which aims to label each word with a specific tag. The NER problem is solved using Deep Learning (DL) techniques combined with Conditional Random Fields (CRF). The final model is tested using 5-fold cross-validation technique, and it achieves around 91% F1-measure. The training and testing is done on our corpus, which was manually annotated.