Unsupervised Discovery of Compound Entities for Relationship Extraction

In this paper we investigate unsupervised population of a biomedical ontology via information extraction from biomedical literature. Relationships in text seldom connect simple entities. We therefore focus on identifying compound entities rather than mentions of simple entities. We present a method based on rules over grammatical dependency structures for unsupervised segmentation of sentences into compound entities and relationships. We complement the rule-based approach with a statistical component that prunes structures with low information content, thereby reducing false positives in the prediction of compound entities, their constituents and relationships. The extraction is manually evaluated with respect to the UMLS Semantic Network by analyzing the conformance of the extracted triples with the corresponding UMLS relationship type definitions.