# Source The dataset was developed by [figure-eight](https://www.figure-eight.com/dataset/medical-sentence-summary-and-relation-extraction/) and the full version is freely downloadable there as well as indications for how to make similar datasets using their platform. # About This dataset contains 3,984 medical sentences extracted from PubMed abstracts and relationships between discrete medical terms were annotated. This dataset focuses primarily on “treat” and “cause” relationships, with 1,043 sentences containing treatment relations and 1,787 containing causal ones. Human-in-the-loop annotators were given two different terms (such as “Lewy Body Dementia” and “Well-formed Visual Hallucinations”) and were asked to mark the relationship between those terms (in this case “Lewy Body Dementia causes Well-Formed Visual Hallucinations). This corpus has been referenced in the following papers: - Anca Dumitrache, Lora Aroyo, Chris Welty: CrowdTruth Measures for Language Ambiguity: The Case of Medical Relation Extraction. LD4IE at ISWC 2015. - Anca Dumitrache, Lora Aroyo, Chris Welty: Achieving Expert-Level Annotation Quality with CrowdTruth: The Case of Medical Relation Extraction. BDM2I at ISWC 2015.