Why is it so hard?

  • Variability: A single meaning can be expressed in multiple different ways.
  • Ambiguity: A single expression can have multiple meanings depending on the context (“Go to the bank” - 😉).
  • Sparsity: Basically, Zipf’s Law.
  • Robustness(?)
  • Context Dependence
  • Language Diversity
  • In other words, uncertainty. That’s why, ML is absolutely necessary for NLP.