One of the first steps in the utterance interpretation pipeline of many
task-oriented conversational ai systems is to identify user intents and the
corresponding slots. Since data collection for machine learning models for this
task is time-consuming, it is desirable to make use of exi