The increasing availability of whole human genomes due to the improvements in high-throughput sequencing technologies makes the interpretation and annotation of data at a whole-genome scale more and more feasible. However, these tasks critically depend on the availability of knowledge already stored in databases or published in the scientific literature. This scenario requires new reliable and integrative information extraction systems to be available to the biomedical community. In this work we present a hybrid approach for mining the wealth of knowledge stored in the scientific literature. This approach is based on the use of efficient text-mining tools in combination with highly accurate collaborative human curation. BioNotate-2.0 is an open source, modular friendly tool which implements a collaborative semi-automatic annotation platform in which human and automated annotations are efficiently combined. BioNotate-2.0 also builds upon the Semantic Web, facilitating the dissemination of annotated facts into other resources and pipelines. BioNotate-2.0 allows any interested user to run his own annotation efforts or to contribute to an existing annotation project. To access and contribute to the BioNotate-2.0 annotation platform, please check: http://genome2.ugr.es/bionotate2/. BioNotate source code is available at: http://sourceforge.net/projects/bionotate/.