Classifying with multiple binary classifiers
So far we have focused on binary classifiers, which classify with one of two possible labels. The same techniques for training a binary classifier can also be used to create a multi-class classifier, which is a classifier that can classify with one of the many possible labels. But there are also cases where you need to be able to classify with multiple labels. A classifier that can return more than one label is a multi-label classifier.
A common technique for creating a multi-label classifier is to combine many binary classifiers, one for each label. You train each binary classifier so that it either returns a known label or returns something else to signal that the label does not apply. Then, you can run all the binary classifiers on your feature set to collect all the applicable labels.
Getting ready
The reuters corpus contains multi-labeled text that we can use for training and evaluation:
>>> from nltk.corpus import reuters >>...
