Using Machine Learning to Collect and Facilitate Remote Access to Biomedical Databases: Development of the Biomedical Database Inventory

Background Currently, existing biomedical literature repositories do not commonly provide users with specific means to locate and remotely access databases. Objective To address this issue, we developed the Biomedical Database Inventory (BiDI), a repository linking databases automatically extracted from scientific literature. BiDI provides an index of data resources path them seamlessly. Methods We designed ensemble deep learning methods extract database mentions. train system, annotated set 1242 articles that included mentions publications. Such was used along transfer techniques natural language processing models targeted at publication detection. Results The system obtained F1 score 0.929 on detection, showing high precision recall values. When applying model PubMed Central databases, identified over 10,000 unique also weblinks reported discarded irrelevant links. For extraction weblinks, achieved cross-validated 0.908. show two use cases: one related “omics” other COVID-19 pandemic. Conclusions enables internet facilitates data-driven research initiatives. is openly available online will be regularly updated automatic text pipeline. approach can reused create different types (ie, others).