Sentence Classifiers:

Each SVM was trained on a specialised corpus. Around 2000 features (= document terms stemmed using the Lovins stemmer) were used.

The precise figures are as follows:

Disease 2682 / Function 2033 / Structure 2694