Detecting Negated and Uncertain Information in Biomedical and Review Texts

The thesis proposed here intends to assist Natural Language Processing tasks through the negation and speculation detection. We are focusing on the biomedical and review domain in which it has been proven that the treatment of these language forms helps to improve the performance of the main task. In the biomedical domain, the existence of a corpus annotated for negation, speculation and their scope has made it possible for the development of a machine learning system to automatically detect these language forms. Although the performance for clinical documents is high, we need to continue working on it to improve the efficiency of the system for scientific papers. On the other hand, in the review domain, the absence of an annotated corpus with this kind of information has led us to carry out the annotation for negation, speculation and their scope of a set of reviews. The next step in this direction will be to adapt it to this domain for the system developed by the biomedical area.