This report focuses on text mining in the biomedical domain for the generation of semantic annotations based on a formal model which is ontology. We start by exposing the generic methodology for the generation of annotations from texts. Then, we present a state of the art on different knowledge extraction techniques used on biomedical texts. We propose our approach based on Semantic Web Technologies and Natural Language Processing (NLP): it relies on formal ontologies to generate semantic annotations on scientific articles and on other knowledge sources (databases, experiment sheets). This approach can be extended to other do-mains requiring experiments and massive data analyses. Finally, we conclude with a discussion about our work and we present some learnt lessons.