This research project develops an ontology-based technique to exploit the information contained in free-text surgical pathology reports for breast cancer patients. A novel ontology for the domain is designed and several tools for information extraction and reasoning are developed, supported by machine learning algorithms aiding the identification of the relevant information within the documents. The research shows that information extraction from surgical pathology reports can be significantly enhanced by machine learning pre-processing, which will select the appropriate extraction technique for the report layout and filter out irrelevant portions of text. Also, such a system can be coupled with clearly defined, formal semantic models of both the reality, which will support the information extraction tasks, and of coding systems, which will enable to automatically assign clinical codes with complex rules. As a whole, it can alleviate the burden for cancer registry staff, researchers or clinicians of reading pathology reports, calculating cancer staging codes' and entering information on a database. The main benefits of this research will result in cost savings and in the augmented completeness and accuracy of both routine cancer registrations and study-specific cancer data collection for cancer registries. The outcomes of this research will also be appreciated by the management of pathology laboratories. Increasing their awareness of the reports' use in automated contexts will hopefully induce relevant modifications in the writing styles of the documents or, even better, encourage the adoption of structured collection of information for, at least, the essential data items used for cancer epidemiology.