Classification module provides several supervised learning approaches including various flavours of methods such as KNN, SVM, Rocchio etc. Classification always includes two steps:
- During the training phase document assignments to predefined classes are learned, using positive (and optionally negative) examples resulting in computation of a classification model.
- During the classification phase the computed classification model is used to compute assignments (incl. confidence values) of new, previously unseen documents to the learned classes.
Examples of classifier applications are filtering of spam emails or sentiment detection. Support for multi-label classification as well as for incremental updating of classification models is provided. Multiple classification models can be maintained and operated simultaneously.