Abstract: An application server for improving the data diversity of a corpus of training data for training a classifier is provided. The application server comprises one or more hardware processors configured to execute a set of instructions to: obtain a first set of text data; determining a set of phrases from the first set of text data; training the classifier using the set of phrases; classify the set of phrases using the trained classifier; and determine a level of accuracy of a classification by the trained classifier with the set of parameters. The application server may retrain the classifier on a second set of text data, or update a set of rules associated with the classifier, in response to determining that the level of accuracy is below a predetermined threshold.