NUBes: A Corpus of Negation and Uncertainty in Spanish Clinical Texts

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 5772–5781 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 5772 NUBES: A Corpus of Negation and Uncertainty in Spanish Clinical Texts Salvador Lima1, Naiara Perez1, Montse Cuadros1, German Rigau2 1SNLT group at Vicomtech Foundation, Basque Research and Technology Alliance (BRTA), Donostia/San-Sebasti´an, 20009, Spain 2IXA group. HiTZ centre. University of the Basque Country UPV/EHU, Donostia/San-Sebasti´an, 20018, Spain {slima, nperez, mcuadros}@vicomtech.org, german.rigau@ehu.es Abstract This paper introduces the first version of the NUBES corpus (Negation and Uncertainty annotations in Biomedical texts in Spanish). The corpus is part of an on-going research and currently consists of 29,682 sentences obtained from anonymised health records annotated with negation and uncertainty. The article includes an exhaustive comparison with similar corpora in Spanish, and presents the main annotation and design decisions. Additionally, we perform preliminary experiments using deep learning algorithms to validate the annotated dataset. As far as we know, NUBES is the largest publicly available corpus for negation in Spanish and the first that also incorporates the annotation of speculation cues, scopes, and events. Keywords: negation, uncertainty, clinical texts, Spanish 1. Introduction The aim of Natural Language Understanding is to capture the intended meaning of texts or utterances. However, un- til recently, research has predominantly focused only on propositional aspects of meaning. Truly understanding lan- guage involves taking into account many linguistic aspects which are usually overlooked. These linguistic phenom- ena are sometimes referred to as Extra-Propositional As- pects of Meaning (EPAM) (Morante and Sporleder, 2012). Some examples of EPAM include factuality, uncertainty, opinions, beliefs, intentions or subjectivity. Documents en- riched with this kind of information can be of utmost im- portance. For instance, in a domain such as the biomedical, the implicit meaning of a sentence can be crucial to differ- entiate whether a patient suffers from a disease or not, or whether they should be taking or not a given drug. One way to learn these nuances is through means of an an- notated corpus. Unfortunately, there are not many corpora that cover these phenomena. Just a few consider negation, a key aspect of factuality –and, even fewer, uncertainty. Negation is understood as an element that modifies the truth value of an event or a statement, or that makes explicit that an event is absent; uncertainty (also called speculation) oc- curs when a speaker is not sure whether an event or state- ment is true. Usually, negation and uncertainty are annotated in two parts: on the one hand, the phrase that triggers the change of meaning (called ‘cue’, ‘trigger’ or ‘marker’) and, on the other hand, the words that are affected by them (called ‘scope’). For a higher level of granularity, there are other elements that can be annotated, such as the element most clearly affected by the cue (called ‘event’1) or the element that reinforces or diminishes the meaning of the cue (called ‘polarity’). A typical annotation that includes all these ele- ments is shown in …

Content truncated to 500 words