Table 1

Datasets used in our experiment


Dataset
Size (# of abstracts)

Training
True positive (TP)
3,536

True negative (TN)
1,959

Likely-positive (LP)
18,930

Unlabeled (U)
105,000
Test
Positive
338

Negative
339

Tsai et al. BMC Bioinformatics 2008 9(Suppl 1):S3   doi:10.1186/1471-2105-9-S1-S3