Hullo folks -- Do you happen to have WMT En-->De systems, from any year, that you would be willing to lend to me?
I'm doing a retrospective survey of open-source systems, and am trying to avoid having to re-train systems based on the papers. I don't need/want your data nor your training parameters/script. I’d just like the model (decoder), config file (if any), and any preprocessing info/scripts for decoding new data. Any open-source toolkit, any type, any year! As long as i can still use the model locally to re-decode the WMT test sets, it's all good. I'm looking at EN-->DE because it's been part of the modern WMT bakeoffs the whole time (2006-2019). Finally, an excuse to dig out your pre-neural systems! On the other end of the spectrum, even your most recent WMT submission would also be helpful. Individual systems will be anonymized, but authors will be cited appropriately. As far as licensing, this is strictly non-commercial use, and I'm happy to delete after use. Happy to answer questions, etc. Cheers, ~amittai PS Google drive, dropbox, ftp, http, whatever works for you. I'm also willing to paypal/venmo for the cost of a flash drive and shipping. What is internet? PPS If you have no En->De but want to help and only have other language pairs, I'll happily take those also. _______________________________________________ Mt-list site list Mt-list@eamt.org http://lists.eamt.org/mailman/listinfo/mt-list