# Categorizing text into dynamic amount of categories

I'm looking for a supervized system/approach, that could learn how to categorize incoming texts/documents, where new categories can be added over time and the training set will be small. The trained model should not be static and should be able to evolve with adding new categories or evaluating new documents.

For each document it should first give it's suggestion that can be then corrected.

In order for model to flag the need of a new output class by inferring on a test document, you could include a class "Unknown" in addition to existing N classes (hence, the output layer now contains N+1 classes). If the model predicts "Unknown" with highest probability, you could add a new class to the output layer after examining the test data.