Skip to Main Content
Table 2.

Progress on classification results over approximately 1 year, evaluated on a fixed holdout set of 9,708 examples. In parallel with these various iterations on the classification algorithms, the training data was raised from 30,665 (initial evaluation with BidGRU) to 38,925 examples (last evaluation with SciBERT) via an active learning approach.

ApproachF-score
ContrastingSupportingMentioning
BidGRU .206 .554 .964 
BidGRU + metaclassifier .260 .590 .964 
BidGRU + ELMo .405 .590 .969 
BidGRU + ELMo + ensemble (10 classifiers) .460 .605 .972 
SciBERT .590 .648 .973 
Observed distribution 0.8% 6.5% 92.6% 
ApproachF-score
ContrastingSupportingMentioning
BidGRU .206 .554 .964 
BidGRU + metaclassifier .260 .590 .964 
BidGRU + ELMo .405 .590 .969 
BidGRU + ELMo + ensemble (10 classifiers) .460 .605 .972 
SciBERT .590 .648 .973 
Observed distribution 0.8% 6.5% 92.6% 
Close Modal

or Create an Account

Close Modal
Close Modal