Google translator has improved its accuracy over time by using neural networks to analyze large amounts of text in different languages. This allows the system to learn patterns and relationships between languages from real examples rather than being programmed with rules. As a result, Google translator can now understand context and nuance better when translating between many language pairs.