Detecting negative and speculative information is essential in most biomedical text-mining tasks where these language forms are used to express impressions, hypotheses, or explanations of experimental results. Our research is focused on developing a system based on machine-learning techniques that identifies negation and speculation signals and their scope in clinical texts. The proposed system works in two consecutive phases: first, a classifier decides whether each token in a sentence is a negation/speculation signal or not. Then another classifier determines, at sentence level, the tokens which are affected by the signals previously identified. The system was trained and evaluated on the clinical texts of the BioScope corpus, a freely available resource consisting of medical and biological texts: fulllength articles, scientific abstracts, and clinical reports. The results obtained by our system were compared with those of two different systems, one based on regular expressions and the other based on machine learning. Our system's results outperformed the results obtained by these two systems. In the signal detection task, the F-score value was 97.3% in negation and 94.9% in speculation. In the scope-finding task, a token was correctly classified if it had been properly identified as being inside or outside the scope of all the negation signals present in the sentence. Our proposal showed an F score of 93.2% in negation and 80.9% in speculation. Additionally, the percentage of correct scopes (those with all their tokens correctly classified) was evaluated obtaining F scores of 90.9% in negation and 71.9% in speculation.

Choosing the right tokenizer is a non-trivial task, especially in the biomedical domain, where it poses additional challenges, which if not resolved means the propagation of errors in successive Natural Language Processing analysis pipeline. This paper aims to identify these problematic cases and analyze the output that, a representative and widely used set of tokenizers, shows on them. This work will aid the decision making process of choosing the right strategy according to the downstream application. In addition, it will help developers to create accurate tokenization tools or improve the existing ones. A total of 14 problematic cases were described, showing biomedical samples for each of them. The outputs of 12 tokenizers were provided and discussed in relation to the level of agreement among tools.

No abstract

The thesis proposed here intends to assist information retrieval and text mining tasks through the negation and speculation detection focusing on two different areas. In the biomedical domain, the existence of an annotated corpus with this kind of information has made possible the development of an effective system to automatically detect these language forms. In the review domain, we have annotated for negation, speculation and their scope a set of reviews. Information retrieval. Negation and speculation detection. Biomedical and review domains.