Text by example of base | % correct prediction of text fragment sequence | Advantage of NFIM (%) | ||||||||
By unconditional probability | By 1st fragment | By 2nd fragment | By mean probability | Method 1 | Method 2 | Method 3 | Method 4 | NFIM | ||
“Russian” | 164.760 | 139.896 | 139.819 | 130.470 | 130.486 | 130.485 | 130.215 | 130.206 | 130.210 | −0.00769 |
“German” | 134.129 | 112.559 | 112.505 | 104.600 | 104.623 | 104.611 | 104.378 | 104.364 | 104.373 | −0.01375 |
“Ukrainian” | 146.245 | 121.950 | 121.923 | 114.341 | 114.372 | 114.365 | 114.105 | 114.098 | 114.106 | −0.01675 |
“English” | 104.389 | 88.326 | 88.238 | 81.159 | 81.169 | 81.158 | 80.918 | 80.913 | 80.923 | −0.03079 |