Skip to main content
. Author manuscript; available in PMC: 2024 Apr 28.
Published in final edited form as: Genet Med. 2022 May 25;24(8):1593–1603. doi: 10.1016/j.gim.2022.04.025

Table 1.

Performance of the respective models to differentiate category 1 from non–category 1 articles on the held-out test set of articles

Model TPR FPR FNR TNR
Snorkel LFs only 0.143 0.378 0.571 0.469
Snorkel RFs only 0.214 0.000 0.786 1.000
Snorkel BERTs only 1.000 0.055 0.000 0.945
Snorkel LFs + RFs 0.143 0.378 0.643 0.469
Snorkel LFs + BERTs 0.143 0.378 0.857 0.622
Snorkel RFs + BERTs 1.000 0.055 0.000 0.945
Snorkel LFs + RFs + BERTs 0.143 0.378 0.857 0.622

Bold values show the results for the highest-perfoming models.

BERT, bidirectional encoder representations from transformers; FNR, false negative rate; FPR, false positive rate; LF, labeling functions; RF, random forest; TNR, true negative rate; TPR, true positive rate.