Skip to Main Content
Table 6: 
Successive halves of the BF dataset used in Figure 2. Proportion of FCE and BEA-19 train is held constant during down-sampling. Learning rates are tuned based on the test set of the CoNLL-2013 shared task.
Dataset proportionexampleslearning rate
full 60011 3 × 10−5 
∼1/2 29998 3 × 10−5 
∼1/4 15121 25 × 10−6 
∼1/8 7608 1 × 10−7 
∼1/16 3749 1 × 10−7 
∼1/32 1841 1 × 10−7 
∼1/64 905 1 × 10−7 
Dataset proportionexampleslearning rate
full 60011 3 × 10−5 
∼1/2 29998 3 × 10−5 
∼1/4 15121 25 × 10−6 
∼1/8 7608 1 × 10−7 
∼1/16 3749 1 × 10−7 
∼1/32 1841 1 × 10−7 
∼1/64 905 1 × 10−7 
View Large
Close Modal
This Feature Is Available To Subscribers Only

or Create an Account

Close Modal
Close Modal
This site uses cookies. By continuing to use our website, you are agreeing to our privacy policy.
Accept