Table 1B.
Corpus No. articles with N codes Total articles
1 2 3 4
Training1544488860916401
Test 200026822312712941
Test 20011842220208

[i] Some of the articles within the training set were obtained in more than one of the queries; thus these articles have more than a single relevant GO classification. This table lists the number of abstracts in each data set and the number of abstracts with one, two, three, and four relevant codes.