Dataset size
It seems that this dataset has only 567 samples... compiling the dataset from scratch via the .txt files from Github yields 6827 samples. Is it only a subset of the full available dataset? Why that? What am I missing here?
Hi @mnbucher ,
According to the paper describing this dataset: https://aclanthology.org/W18-0535/
The corpus consists of 189 texts, each in three versions (567 in total).
Hi @albertvillanova thanks for the pointer! I guess the main difference is that the paper states "189 texts", which is not on a sentence level. Each text contains a different number of sentences and the script on my side created the dataset on a sentence level. The dataset provided here on HF aggregated the texts by the full text and not on a sentence level. Totally fine from my side. Maybe we can mention this somewhere in the Dataset card.
Yes @mnbucher , I agree we better mention this in the dataset card to make it totally clear. Thanks! :)
@albertvillanova fantastic, thanks!