Malaria continues to impose a heavy global health burden, and researchers need to rapidly extract molecular mechanisms, epidemiology, and potential therapies from an ever‑growing body of literature. Manual curation cannot keep up with the scale and speed required, motivating the development of automated biomedical information extraction.\ \ This paper introduces MiNER, a fine‑tuned BioBERT‑based entity recognition framework designed to extract clinically relevant entities from malaria‑focused scientific articles. We first assemble a corpus of several thousand malaria papers, apply sentence segmentation, noise removal, and standardization, and then manually annotate entities such as diseases, drugs, genes, and symptoms according to medical guidelines, creating a high‑quality supervised dataset.\ \ At the model level, BioBERT encodes the text into context‑aware vectors, followed by a linear classification head for sequence labeling. Supervised fine‑tuning on the annotated data enables the model to capture malaria‑specific semantic patterns.\ \ In experiments, we compare different encoders (BioBERT, BioClinicalBERT, SciBERT) and traditional machine‑learning baselines (CRF, SVM) across precision, recall, and F1. MiNER consistently outperforms all baselines, achieving roughly a 7% increase in precision and a 9% increase in recall.\ \ To encourage reproducibility, we release the fully annotated dataset and training scripts. These resources can be directly leveraged for more advanced relation extraction or cross‑lingual modeling.\ \ Review: The study demonstrates that fine‑tuning BioBERT on a disease‑specific corpus yields highly effective entity recognition, providing valuable data and tools for the biomedical NLP community.