PharmKE: Knowledge Extraction Platform for Pharmaceutical Texts Using Transfer Learning

The challenge of recognizing named entities in a given text has been very dynamic field recent years. This is due to the advances neural network architectures, increase computing power and availability diverse labeled datasets, which deliver pre-trained, highly accurate models. These tasks are generally focused on tagging common entities, but domain-specific use-cases require custom not part pre-trained can be solved by either fine-tuning models, or training main lies obtaining reliable test manual labeling would tedious task. In this paper we present PharmKE, analysis platform pharmaceutical domain, applies deep learning through several stages for thorough semantic articles. It performs classification using state-of-the-art transfer thoroughly integrates results obtained proposed methodology. methodology used create accurately then train models entity tasks, centered domain. compared fine-tuned BERT BioBERT trained same dataset. Additionally, PharmKE from recognition resolve co-references analyze relations every sentence, thus setting up baseline additional such as question answering fact extraction. recognized also expand knowledge graph generated DBpedia Spotlight text.