There’s an article over at MIT Technology Review discussing a paper in Annals of Applied Statistics on using statistical methods to date medieval charters that lack an internal date. I don’t pretend to understand the math involved in selecting the precise method—it’s all in the paper over at ArXiv—but the simple version is that the authors have automated the comparison of the language and terminology in the charters.
Instead of looking only for significant phrases that humans have identified as being era-specific, the program compares every phrase of every length to every phrase of every length in a set of 3,000 machine-readable British charters with known dates.
The results are pretty cool. Not only did the program do well dating 300 or so charters for whom the date was known, but it also managed to date a previously unknown charter to within 4-8 years of the estimate made by a historian who examined it.
Real kudos go, though, to the creators of the Documents of Early England Data Set (DEEDS), who started transcribing charters into a machine-readable format long before anything like this paper was possible. Did I mention that it’s a University of Toronto project?
Have a medieval charter you need dated? Just try the web-facing version: D+MECH.
(h/t Medieval News)