EPO published Google LLC's patent EP3966811A1 titled "Two-Pass End to End Speech Recognition" on March 25, 2026. The patent relates to neural network-based speech recognition technology using a two-pass approach and classifies under IPC groups G10L (speech recognition) and G06N (AI/neural networks including deep learning variants). Inventors include Tara C. Sainath, Yanzhang He, Bo Li, and others from Google.

This is a patent grant notice, not a regulatory rule imposing obligations. Technology companies developing speech recognition systems may wish to review this patent to ensure their products do not infringe on the claimed methods. No compliance deadlines, reporting requirements, or penalties are associated with this publication.