question regarding dataset statistics
Hi there,
I notice that some statistics of the dataset is different from what is described in the paper. For example, according to the paper, there are in total 12293 chapter-level instances with input length of ~5K tokens and summary of ~500 tokens. However, in your dataset there are 9600/1431/1484 training, test and dev examples (adding up to 12515 instances), with on average ~3k8 words input and about 360 words in summary. Do you have any idea why this is the case by any chance?
Hey, good spot. I looked back at my version of the repo after having run the scripts recommended in the original repo and I found that there were c.200 sections that returned errors because the urls listed are no longer available:
https://github.com/KMFODA/booksum/blob/main/scripts/data_collection/pinkmonkey/section_errors.txt
That should also probably explain why the word counts are different.
That makes sense. Thank you for the answer!