“Data -Driven” Ontologies for an Information Extraction System from Polish Mammography Reports

The paper describes the ontology development for an IE (Information Extraction) application for Polish mammography reports, experiences and lessons learned, and the evaluation of the system. Information extraction requires prior knowledge on data structures we would like to identify. When information being searched for is as complicated as this contained in mammography reports, a theoretical approach of using the predefined domain knowledge is required. For our research goal, extraction of possibly all precise information from mammography reports, ontology has been chosen as platform of knowledge representation. During the work two ontologies have been developed, the first based mainly on BI-RADS [5], the second adjusted to the task of information extraction. The paper is structured as follows: section 2 relates our experiences concerning the reuse of existing ontologies, sections 3 and 4 present respectively initial mammographic ontology and modified model adapted for the task of information extraction, section 5 presents the IE system and results of information extractions, section 6 concludes the paper.