Customers now have early access to a new type of classifiers we call Redaction Classifiers. These classifiers are helpful to identify and remove "noise" such as personal information, email signatures, logs, auto-responses, and even code.

A redaction classifier can be trained like any other classifier by feeding it samples from your data set. Once you have a training set of 10 samples you're able to adjust the settings in Caravel to hide this data from scoring and reporting.