Francisco Rangel a,bPaolo Rosso b
a Autoritas Consulting, S.A., Spain
b Pattern Recognition and Human Language Technologies (PRHLT), Universitat Politècnica de València, Spain

Language and Law Journal, 2018 (in press)

Evaluation campaigns allow for the creation of a common framework for research, making possible comparability and reproducibility in science. Furthermore, the huge amount of publicly available data in the different social platforms (social big data) favours evaluation tasks proliferation, for example in forensic linguistics.

However, due to the implications that the release of the data may have on the privacy of people, rules for their protection must be laid down. These norms have been defined by the European Commission in the General Data Protection Regulation (GDPR) of April 27, 2016. Moreover, for the collection and distribution of data, each social media platform defines its legal base to use its data.

In this paper, we describe the GDPR articles that apply for the organisation of evaluation tasks. Moreover, we propose a methodology to follow at the time of the organisation of evaluation tasks. Finally, we show a case study about the organisation of the PAN forensic linguistic tasks on author profiling at CLEF that we have been organising since 2013, showing how both GDPR and Twitter Terms of Service have been met when creating and distributing the corpora.