At the moment, there is a great mismatch between the amount of data available in electronic (machine-readable) form and the amount of data that can actually be processed or used by computers. This is because most information is stored as text, which is easy for humans to process but too unstructured for use by computer algorithms.

In computer science, ontologies represent knowledge in different domains, and are designed for the purpose of knowledge sharing and reuse. Ontologies enable computers to reason about knowledge and do many tasks automatically. However, it is time-consuming to build ontologies by hand. Also, different tasks may require different ontologies.

In this thesis, I describe two approaches to automatic ontology extraction: one that operates on structured Web pages and one that operates on short text segments. I present example ontologies automatically extracted from real data, and describe and evaluate several applications of these ontologies: Web page content labeling, data extraction, multimedia generation and web service matching. I show that automatic ontology extraction is feasible and can be used to quickly acquire information about new domains.