Agile Semantic Annotation of Scientific Texts at the Biomedical Scenario

Semantic annotation enables the inclusion of semantic information (ontology concepts), readable by software agents, into the most varied types of documents available on the Web. This operation facilitates search and access to relevant information. In the biomedical domain, semantic annotation is not an easy task, since the content to be annotated can be associated with a large set of texts and large ontologies. This work describes a scientific scenario, where a corpus of 500 selected scientific papers was annotated using three distinct ontologies, with focus on the research of gene essentiality. Then it presents an ontology modularization approach (UMA Project) that aims at improving the performance on scientific text automatic annotation, using a structure composed of concepts extracted from two ontologies. Finally, it describes an experiment that used the UMA Project tool to automatically annotate the same Corpus, and discusses the gains in terms of performance.