Semantic information elicitation from unstructured medical records

Semantic elicitation of relevant information entities from semi- and unstructured documents is an important problem in many application fields. This paper describes HiLXa system implementing a very powerful semantic approach to information extraction from semi- and unstructured documents obtained combining knowledge representation formalisms, like ontology languages, and two-dimensional languages exploiting a two-dimensional spatial representation of documents. The HiLX system constitutes a new generation technology capable of capturing and eliciting relevant information regarding a specific domain. It is founded on OntoDLP, an extension of disjunctive logic programming for ontology representation and reasoning. In the HiLX system the semantics of the information to be extracted is represented by using OntoDLP ontologies and the extraction patterns are expressed by means of regular and two-dimensional expressions. By converting the extraction patterns to OntoDLP reasoning modules, the HiLX system can actually extract information from HTML pages as well as from flat text documents using the same patterns. In this paper the extraction of clinical information and events, regarding patients, diseases, therapies and drugs, from electronic textual medical records is shown. Extracted information are represented in XML and can be stored in structured form using relational database or ad-hoc ontologies to enable further analysis.