Copyright and text and data mining (TDM)
Copyright Act, 2003, fair dealing and compilation copyright
Copyright Act 2003 (Act No. 21 of 2003), ss. 2, 5, 50-52 (literary work, protected works, research/private study, criticism/review/reporting, determining fair dealing)Copyright Act, 2003 (Act No. 21 of 2003), full text, WIPO Lex
In force since 30 November 2004. Binds public and private bodies.
What this law does
Section 50 permits fair dealing with a literary, dramatic, musical or artistic work for the purposes of research or private study without infringing copyright, subject to section 52's four-factor fairness test. Section 51 permits fair dealing for criticism or review of a work or a performance, and a separate fair-dealing ground for reporting current events using a protected work other than a photograph, each ground (except reporting current events) requiring a sufficient acknowledgement.
Section 52 directs a court determining whether a dealing is fair to weigh the nature of the work, the extent and substantiality of the part used, the purpose and character of the use, and the effect on the potential market for or commercial value of the work.
The Act has no express text-and-data-mining exception, so reproducing scraped Vincentian text at scale to train a model can rest only on the narrow research-or-private-study ground, whose private, non-commercial character is a poor fit for large-scale commercial training. Section 2 defines a literary work to include a written table or compilation.
Section 5(1)(a) makes copyright subsist only in an original literary, dramatic, musical or artistic work, so a compilation is protected only as an ordinary literary work conditioned on originality of its selection or arrangement. The Act confers no sui generis database right, and extracting unoriginal underlying data from a compilation falls outside copyright altogether.
What it requires