A Corpus for Outbreak Detection of Diseases Prevalent in Latin America

Proceedings of the 24th Conference on Computational Natural Language Learning, pages 543–551 Online, November 19-20, 2020. c⃝2020 Association for Computational Linguistics https://doi.org/10.18653/v1/P17 543 A Corpus for Outbreak Detection of Diseases Prevalent in Latin America Antonella Dellanzo Dept. of Computer Science FCEyN Universidad de Buenos Aires, Argentina antodellanzo@gmail.com Viviana Cotik Dept. of Computer Science FCEyN Universidad de Buenos Aires, Argentina vcotik@dc.uba.ar Jos´e Ochoa-Luna Dept. of Computer Science Universidad Cat´olica San Pablo, Arequipa, Peru jeochoa@ucsp.edu.pe Abstract In this paper we present an annotated cor- pus which can be used for training and test- ing algorithms to automatically extract infor- mation about diseases outbreaks from news and health reports. We also propose initial ap- proaches to extract information from it. The corpus has been constructed with two main tasks in mind. The first one, to extract entities about outbreaks such as disease, host, location among others. The second one, to retrieve re- lations among entities, for instance, in such ge- ographic location fifteen cases of a given dis- ease were reported. Overall, our goal is to offer resources and tools to perform an auto- mated analysis so as to support early detec- tion of disease outbreaks and therefore dimin- ish their spreading. 1 Introduction In recent years, several contagious diseases have aroused. Many of them, such as dengue, Guillain- Barr´e, Zika and microcephaly, occur in tropical regions of the world as in Latin America. There are also other diseases such as Chagas, also referred to as American trypanosomiasis, which have existed for many decades, but not many resources are used to treat them, probably because they do not appear in densely populated areas and their usual hosts belong to populations of scarce socioeconomic re- sources. On the other hand, there are endemic dis- eases, such as hantavirus, which appear recurrently in some regions of Argentina and Chile, among other countries. Finally, we are facing nowadays the COVID-19 pandemics. Those illnesses have many issues in common: 1) they are dangerous in terms of number of losses of human life or lifelong consequences to those who carried them, 2) their economic impact can be very important (as in loss of tourism incomes with hantavirus, and loss in almost all economic areas with the isolation imposed by different gov- ernments due to the COVID-19 pandemics), and 3) when an outbreak occurs there is few available data that can support for data-driven decision making policies. In this work we attack the third issue, we argue that information including number of cases, host and location of the host, among others, is crucial in order to understand the propagation of the illness and diminish its spreading. When available, the information usually is of poor quality (incomplete, incorrect, inconsistent, not publicly available, and it lacks the timeliness attribute, among others). Thus, our goal is to contribute with automated tools and resources that allow to alert and inform about possible outbreak diseases and some illnesses prevalent in Latin America. To do so, we have con- structed an annotated corpus based on …

Content truncated to 500 words