This repository accompanies a short paper for the first Computational Humanities Research Workshop (18-20 november 2020; to be held online):
M. Kestemont & F. Karsdorp, 'Estimating the Loss of Medieval Literature with an Unseen Species model from Ecodiversity'. Computational Humanities Research Workshop. Amsterdam [online], 18-20 november 2020.
This repository that contains all the code and software (with some documentation interwoven) necessary to replicate our findings; this (and each future release) will be archived on Zenodo. The paper offers a more technical complement to and extension of a Dutch-language journal article that was previously published, but which was targeted at a broader audience of literary historians:
Kestemont, M. & Karsdorp, F., ‘Het Atlantis van de Middelnederlandse ridderepiek. Een schatting van het tekstverlies met methodes uit de ecodiversiteit’, Spiegel der Letteren 61:3 (2019), 271-290.
We ran all the code in an Anaconda installation of Python 3.6.10. The top-level directory of this repository has a requirements.txt file that you can use to install the more specific version of the third-party packages used here. Note that the data used here feature a number of minor improvements in comparison to that for the SDL paper (with some insignificant fluctuations in the results). The data format is documented in the notebook under /notebooks.
Erratum (22/02/2021): In follow-up work, we have discovered that the published version of the paper contains an inaccurate interpretation of the output of the minsample() function, as originally defined in this repository. In the terminology of Chao et al. (2009), this function returned m = n * x*, and not n + m, as we incorrectly claimed in the paper. Because n <<< m, however, the results don't shift much and the main arguments of the paper still stand firmly. This does mean that the document survival estimate has to be adjusted downwards. A more correct estimate for the original number of documents in the population therefore is:
Minimum sampling: 2214.77 CI(1231.19 - 4173.42)
Consequently, the correct document survival estimate is rather: 7.54% (~ 167/2215 MSS). The code and graphs in this repository have been corrected in the latest release. (None of the other results reported in the paper are affected by this issue.)