Datasets:
Difference between this dataset and the OPUS version?
Hi, thank you for curating the dataset! I see there's another bible corpus on huggingface at https://huggingface.co/datasets/bible_para, which seems to come from https://opus.nlpl.eu/bible-uedin/corpus/version/bible-uedin and https://github.com/christos-c/bible-corpus. Do you know what the main differences between these two datasets are?
There end up being a lot of small problems when you just scrape the text without any preparation and cleaning. We talk a little bit about the preprocessing and cleaning we've done in https://arxiv.org/abs/2304.09919 . We've also taken care to align verse segments to matching verse segments, which can be challenging when different Bibles sometimes use different numbering schemes (also in the paper). But, probably the biggest difference is in the number of languages covered.
I see, thanks a lot!