Skip to Main Content
Table 5

Comparing our tuned metric to the best rivaling metric at wmt14, for each individual language pair (this best rival differs across language pairs) at the segment-level using Kendall's τ. Statistically significant improvements are marked with ** for p-value < 0.01.

Systemfr-ende-enhi-encs-enru-enOverall
DiscoTKparty 0.433** 0.380** 0.434 0.328** 0.355** 0.386** 
Best at wmt14 0.417 0.345 0.438 0.284 0.336 0.364 
 +0.016 +0.035 −0.004 +0.044 +0.019 +0.024 
Systemfr-ende-enhi-encs-enru-enOverall
DiscoTKparty 0.433** 0.380** 0.434 0.328** 0.355** 0.386** 
Best at wmt14 0.417 0.345 0.438 0.284 0.336 0.364 
 +0.016 +0.035 −0.004 +0.044 +0.019 +0.024 
Close Modal

or Create an Account

Close Modal
Close Modal