Building a Comprehensive Romanian Knowledge Base for Drug Administration

Proceedings of Recent Advances in Natural Language Processing, pages 829–836, Varna, Bulgaria, Sep 2–4, 2019. https://doi.org/10.26615/978-954-452-056-4_096 829 Building a Comprehensive Romanian Knowledge Base for Drug Administration Bogdan Nicula1, Mihai Dascalu1, Maria-Dorinela Sirbu1, Stefan Tr˘aus,an-Matu1 and Alexandru Nuta2 1University Politehnica of Bucharest, 313 Splaiul Independentei, 060042, Bucharest, Romania 2All Business Management, Str. Vasile Conta 19, sc. A, et. 1, ap. 8, Bucharest, Romania Abstract Information on drug administration is ob- tained traditionally from doctors and phar- macists, as well as leaflets which provide in most cases cumbersome and hard-to- follow details. Thus, the need for med- ical knowledge bases emerges to provide access to concrete and well-structured in- formation which can play an important role in informing patients. This paper introduces a Romanian medical knowl- edge base focused on drug-drug interac- tions, on representing relevant drug in- formation, and on symptom-disease re- lations. The knowledge base was cre- ated by extracting and transforming in- formation using Natural Language Pro- cessing techniques from both structured and unstructured sources, together with manual annotations. The resulting Ro- manian ontologies are aligned with larger, well-established, English medical ontolo- gies. Our knowledge base supports queries regarding drugs (e.g., active ingredients, concentration, expiration date), drug-drug interaction, symptom-disease relations, as well as drug-symptom relations (e.g., searching for the drug that might be most useful for treating a given set of symp- toms). 1 Introduction The conventional way of accessing information regarding drug administration and storage strate- gies, recommendations and precautions, effects and side-effects is via medical leaflets. How- ever, most of the times the text is too complex, too cluttered with information, that the leaflet ends up being ignored by the consumer. At the same time, when consumers must take multiple drugs as result of overlapping treatment schemes, it becomes increasingly difficult for them to keep in mind all mentioned precautions and counter- indications. This paper presents a knowledge base built upon medical leaflets and existing med- ical ontologies aimed at aiding Romanian con- sumers when adding new drugs to their treatment schemes. The knowledge base consists of a set of two ontologies built from information extracted from websites of pharmaceutical producers and na- tional agencies, which were combined with En- glish medical ontologies. The first ontology is used to better structure the leaflet content and pro- vide easy access to information concerning ad- ministration and storage strategies, together with possible side-effects. This ontology is aligned with a larger English-based ontology - DINTO (Herrero-Zazo et al., 2015) - in order to discover incompatibilities between drugs and to warn the user whether two administered drugs might in- teract one with another. The second ontology is focused on diseases. Translations for both dis- eases and symptoms were added for Romanian language, allowing customers to lookup possible explanations for their symptoms. Moreover, the gap between the two ontologies is filled in by in- dexing the description and recommendation texts for the drug leaflets, therefore enabling users to directly search what drugs might help them deal with certain symptoms. These tools are …

Content truncated to 500 words