Questions to answer:

  • How to gather training data for the IDM?
  • How much data is enough to train something meaningful?
  • How to annotate our data?