A question about test and validation split

[bedapudi6788]

Hi, In your paper you mentioned that you used a 20-10-70 split. I am interested in knowing how this split is performed. For eg: if website Real-A has RA articles, Real-B has RB articles, Fake-A has FA articles, Fake-B has FB articles, is your split article wise i.e: Train(0.2RA + 0.2RB, 0.2FA + 0.2FB) or is it overall. i.e Train 0.2(RA + RB) or did you keep some sites in test and some in train. Thank you.