SMER: A Novel Medical Entity Recognition Technique based on ScispaCy(BC5CDR) and Med7

Named Entity Recognition (NER) is a stream of Natural language pro-cessing playing a pivotal role in the KDD process of data mining. The Last few years have seen a sudden surge in research about NER models, however, the scope of this research is often restricted to general text and speech. This paper thus talks about a specific stream the Medical NER. In the scope of this paper, a novel technique is proposed which utilizes a combination of two state-of-the-art NER models, Sciscapcy ( BC5CDR) and Med7 to achieve optimal results. Scis-pacy is a machine learning model trained on a sizeable biomedical corpus while Med7 is a rule-based system that specializes in detecting medical concepts by utilizing its large medical pattern vocabulary. The model is evaluated on bench-mark datasets and showcases better susceptibility to medical terms in a text. The model is able to distinguish between a wide range of medical entities including drug name, disease name, drug quantity, clinical procedures, and pathogen name. The paper highlights the potential of combining rule-based systems with machine learning models in developing robust models better suited for domain NER. By fetching strengths from both Scispacy(BC5CDR) and Med7, the paper proposes “SMER – ScispaCy – Med7 Entity Recogniser”, an adaptable solution that con-tributes to more accurate and comprehensive medical entity recognition model that can further be used in automatic medical record-keeping, patient medicine orders and facilitate medical practioners.