Multilingual Epidemic Event Extraction : From Simple Classification Methods to Open Information Extraction (OIE) and Ontology

Proceedings of Recent Advances in Natural Language Processing, pages 1227–1233 Sep 1–3, 2021. https://doi.org/10.26615/978-954-452-072-4_138 1227 Multilingual Epidemic Event Extraction : From simple Classification methods to Open Information Extraction (OIE) and Ontology Sihem Sahnoun Sorbonne University France sahnounsihem@yahoo.com Ga¨el Lejeune Sorbonne University France gael.lejeune@sorbonne-universite.fr Abstract There is an incredible amount of information available in the form of textual documents due to the growth of information sources. In or- der to get the information into an actionable way, it is common to use information extrac- tion and more specifically the event extraction, it became crucial in various domains even in public health. In this paper, We address the problem of the epidemic event extraction in potentially any language, so that we tested dif- ferent corpora on an existed multilingual sys- tem for tele-epidemiology: the Data Analysis for Information Extraction in any Language (DANIEL) system. We focused on the influ- ence of the number of documents on the perfor- mance of the system, on average results show that it is able to achieve a precision and recall around 82% but when we resorted to the evalu- ation by event by checking whether it has been really detected or not, results are not satisfac- tory according to this paper’s evaluation. Our idea is to propose a system that uses an on- tology which includes information in different languages and covers specific epidemiological concepts, it is also based on the multilingual open information extraction for the relation ex- traction step to reduce the expert intervention and to restrict the content for each text. We de- scribe a methodology of five main stages: Pre- processing, relation extraction, named entity recognition (NER), event recognition and the matching between the information extracted and the ontology. 1 Introduction Infectious diseases are responsible for the morbid- ity and mortality and like we see today with the Covid-19 pandemic, surveillance provides us with information to improve our knowledge of their epi- demiology (space-time dynamics, evolution of clin- ical and microbiological characteristics), in order to identify an appropriate control and prevention mea- sures. The growth of digital data sources provide an avenue for data-driven surveillance, referred to as Epidemic Intelligence. The purpose of epidemic intelligence is to detect, analyze and monitor poten- tial health threats over time (Nash and Geng, 2020). It requires the development of tools dedicated to the collection and processing of unstructured textual data published on the Web. Information Extrac- tion (IE) (Martinez-Rodriguez et al., 2020) is one of the areas of active research in artificial intelli- gence and made it possible to analyze data from web sources. Different tasks of the IE systems have been suggested such as named entity recognition (NER) to identify real-world objects such as names of people, locations, names of diseases, etc. The relation extraction , which aims to find a seman- tic relation between two entities in a text (Elloumi et al., 2012). The event extraction is also an im- portant task in the field of IE to detect, from the text, the …

Content truncated to 500 words