Confusion matrices are the tried-and-true way to evaluate the metrics for machine learning classifiers. Is updating them really so confusing?

Long story short, humans are slow, taking at least 10 seconds to label one example. However, a machine can make the same judgement in milliseconds.

The history of using machines to understand the meaning of natural language started as early as World War II and continues through present day.

Until recently, natural language processing (NLP) applications have been limited in their ability to understand nuance and context. What's changed?