Figure 4
From: Statistical algorithms for ontology-based annotation of scientific literature

Decision Trees. In this figure we can see an abstract going through a few steps of the annotation process for both a regular naive Bayes classifier trained on the gold standard corpus and a Bayesian decision tree. The abstract classified by the naive Bayes classifier is classified without regard to decisions already made by the classifier. Therefore, it is classified with the label False Font as its stimulus modality even though its stimulus type was Auditory. By contrast, the when the Bayesian decision tree needs to identify a Stimulus Type it uses a classifier trained on a set of abstracts which are all annotated with the label Auditory and thus picks Chord Sequence as the abstract's Stimulus Type.