Custom Classifier is build on the latest advances in zero-shot and few-shot machine learning techniques. These techniques let us build a text classification model with very few training examples or, in our case, no training examples at all. Essentially, this means that you can now define your categories and build a text classifier without training a machine learning model. The only optimization you need is to select the category list that gives the best performance on your text. Check out our research in this area here.