Table 8

Distribution of instances by frequency and percentage (%) for each meta-knowledge value
Value ST FullText FullText(%)/ST(%)
frequency % frequency %
Investigation 131 5.2 201 5.8 1.11
Analysis 312 12.4 477 13.7 1.10
Observation 1,154 45.8 1,465 42.0 0.92
Fact 89 3.5 94 2.7 0.76
Method 0 0.0 4 0.1 N/A
Other 832 33.0 1,244 35.7 1.08
L3 2,415 95.9 3,283 94.2 0.98
L2 87 3.5 162 4.6 1.35
L1 16 0.6 40 1.1 1.81
Positive 2,389 94.9 3,294 94.5 1.00
Negative 129 5.1 191 5.5 1.07
High 165 6.6 241 6.9 1.06
Low 18 0.7 39 1.1 1.57
Neutral 2,335 92.7 3,205 92.0 0.99
Current 2,487 98.8 3,415 98.0 0.99
Other 31 1.2 70 2.0 1.63
Event Total 2,518 100.0 3,485 1.0 100.00

Distributions of predictions on test set of the ST corpus and on the full-text subset of the BioNLP-ST’11 GENIA corpus are reported. EventMine-MK used the model trained on the ST-MK corpus with the +GENIA setting, as in Table 5. The last column shows the ratio of the percentage in the full-text subset to the percentage in test set of the ST corpus.

Miwa et al.

Miwa et al. BMC Bioinformatics 2012 13:108   doi:10.1186/1471-2105-13-108

Open Data