Assuming all the information were available, I would like to see how many words (of the published version; not all the intermediate drafts) were written during

Assuming all the information were available, I would like to see how many words (of the published version; not all the intermediate drafts) were written during a given year. If it took 2 years to write 100,000 words for a release on 2014-01-01 then 2012 and 2013 would be at 50,000 words each. That would indeed be a 'constant X amount of words for however long the line between publications is'. (This also assumes no publication delay. I have no idea how to tell how long it took to get edited and reviewed.) It's possible to fake it. You could compute the average publication rate then backtrack that amount for each publication date, assuming a constant rate. Then let overlapping publications sum, and use a smoothing function to interpolate. It's still a mess. Especially if there are any posthumous publications. Yes, getting the actual rate requires a lot of research, including perhaps talking to people who study those authors. Eg, a Twain scholar probably knows a lot more about when things were published. But this gets into serious labor of love (or Master's thesis) territory.