Abstract:
The sheer volume of scienti c literature challenges researchers to identify and utilize the knowledge embedded in them and makes automated extraction and process ing necessary. Automated processing becomes especially signi cant when the extracted information is combined with the linked data resources represented with ontologies. The vast knowledge space represented in Linked Open Data sources provides numerous knowledge discovery opportunities through semantic searching and inference. This thesis aims to extract biomedical entity relations embedded in scienti c articles and semantically represent them in a machine-processable manner. For this purpose, we proposed an ontology named Biomedical Entities Evidence (BEE) that represents biomedical entity relationships as well as their provenance in scienti c articles. To express the approach's feasibility, we extracted chemical-protein multiclass relations and chemical-disease binary relations. These relations are represented based on BEE ontology. To demonstrate the bene ts of ontology-based semantic representation, we have implemented a semantic application prototype that utilizes several Linked Open Data sources and inferred data based on ontologies and custom rules. This prototype was used to evaluate the bene ts of the semantic representation by performing informa tion retrieval tasks of varying complexities.
