Biomedical Named Entity Recognition from Malaria Literature using BioBERT

The unprecedented growth of unstructured text data in the healthcare domain makes it difficult to find and extract relevant information manually. Biomedical named entity recognition deals with the process of automatically identifying entities of clinical significance, such as symptoms, treatment, names of drugs, and organisms, from unstructured text documents such as electronic health records and radiology reports. This is the first but essential step for building many text-understanding applications such as conversational agents and information retrieval. Pre-trained language or linguistic models have recently gained popularity in natural language processing due to their inherent ability to manage contexts better. The proposed approach fine-tunes the pre-trained BioBERT model, which is one of the current state-of-the-art biomedical language models, on a large set of Malaria text to extract named entities. This work implements and compares different machine learning algorithms on different feature extraction techniques to establish the usefulness of the proposed approach. The method proposed is found to be outperforming our chosen baselines and shows better precision and f-measure on our extensive performance comparison experiments.