Text and data mining in health research: reflections on copyright

Abstract

The usage of computer systems in the collection, organization, and analysis of large volumes of data and texts and, in this context, text and data mining (TDM) 1 techniques is critical to contemporary data-intensive studies, such as those related to the COVID-19 pandemic 2,3,4,5,6,7,8 . In addition to the ethical issues concerning research data appropriation 8 , copyright brings specific challenges regarding the use of TDM, due to the objects to which protection applies, its temporal extension, broad scope of assigned rights, and scarcity of explicit limitations. Because of their importance for research, our study is focused on these obstacles and ongoing regulatory actions.Copyright does not protect facts, information or data, nor even the content of a work, as its object of protection is the literary or artistic form in which they are expressed and communicated. However, when these elements are combined, organized or systematized in a database that is minimally original in terms of content selection, organization or arrangement, this material will have its access and use controlled by its owner and then protected by copyright 9,10,11 . Also, legal protection by copyright of databases is just one of the layers of control over access and use of data and information 11 . A second layer involves technological restrictions to access or use of databases 12 , and a third layer refers to legal prohibitions to circumvent these mechanisms, even when there is no direct violation of any right over the material 13 . It has a direct impact on essential research activities, such as verification, reproducibility, and communication of results, by allowing or limiting who can perform what type of research, use what type of material, and under what conditions 14,15,16 . This is highlighted in TDM, as it requires uses from as many available sources as possible, which do not always have open access. Therefore, using only this type of source with this method can lead to exacerbated biases in the results 15 .There is an ongoing demand for regulatory reforms at a global level, given the normative, technological, and contractual extension of control powers of data holders, which imposes legal challenges on numerous data-intensive research activities 17,18,19 -a fact particularly observed in countries of the Global South 19,20,21 . Given the above, several legal systems around the world started in the last decade reforms in their copyright legislation to include standards with a focus on ensuring the legality of TDM 22 , such as the European Union 23,24 and countries like Japan 25,26 and Singapore 27 .In Brazil, potential challenges and opportunities have been identified regarding the use of dataintensive technologies in the texts of the Brazilian Artificial Intelligence Strategy (EBIA, acronym in Portuguese) 28 and the Brazilian National Intellectual Property Strategy (ENPI, acronym in Portu-