3 ms·
> I'd have liked to see some analysis of abandonment rate normalized for reading complexity, but that's hard to characterize ... Maybe one could do something l
by gwern 7y ago
> I'd have liked to see some analysis of abandonment rate normalized for reading complexity, but that's hard to characterize ...
Maybe one could do something like the Flesch-Kincaid reading index and variants of those and include other metrics like # of unique words. But it's be a real pain in the arse to do - you can get the books off Libgen, surely, but you'd have to choose manually and then often it'll be a PDF where the OCR or text layer is hot garbage...
- hug 7y agoAssuming that this is all largely fiction, you'll generally get a decent ePub of essentially any book published by a large publisher, and given that ePub is essentially just XHTML it's not "hard" to parse well enough for your purposes. The margin of error for parsing the books is going to be smaller than any margin of error forced on you by GoodReads, anyway... ...but it's still probably not worth it.