Generating samples for the other languages

[eisenjulian]

Hello,
Thanks a lot for putting this together. I have two questions:

1. Upon trying to run sampling_rcv2 it seems that I need a single file per dialect containing the file name and class for each of the articles. Is this file available somewhere? If I create such a file would it a valuable pull request?
2. Have you tried extending the benchmark to the rest of the languages? Is there an issue with the label distribution that prevents it? Didn't find any mention of this in the paper.

Thanks again!

[icoxfog417]

I also looking for required files. But TREC site is down now. As a matter of fact, we can't get any data except the manipulating script.