PSINK: Automatically Populating a Preclinical Spinal Cord Injury Knowledge Base to Support Clinical Translation
In this BMBF-funded project, we apply ontology-based information extraction from textual data to large collections of scientific publications reporting on pre-clinical studies in the context of spinal cord injury (SCI). Our goal is to create the first comprehensive SCI knowledge base in order to foster translation from pre-clinicial research into therapeutic concepts that are likely to be effective in human patients. Our approach to information extraction is based on probabilistic graphical models. We apply techniques from active learning and distant supervision in order to reduce manual annotation and curation efforts to a minimum.