Personalized Speech recognition on mobile devices

TL;DR

This paper presents a highly efficient, accurate, and low-latency speech recognition system optimized for mobile devices, combining quantized LSTM models, SVD compression, and adaptive language modeling.

Contribution

It introduces a novel mobile speech recognition system using quantized LSTM with CTC, SVD compression, and a unified Bayesian-interpolated language model for multiple domains.

Findings

Achieves 13.5% WER on dictation task

Runs seven times faster than real-time on Nexus 5

Uses a single adaptive language model for multiple domains

Abstract

We describe a large vocabulary speech recognition system that is accurate, has low latency, and yet has a small enough memory and computational footprint to run faster than real-time on a Nexus 5 Android smartphone. We employ a quantized Long Short-Term Memory (LSTM) acoustic model trained with connectionist temporal classification (CTC) to directly predict phoneme targets, and further reduce its memory footprint using an SVD-based compression scheme. Additionally, we minimize our memory footprint by using a single language model for both dictation and voice command domains, constructed using Bayesian interpolation. Finally, in order to properly handle device-specific information, such as proper names and other context-dependent information, we inject vocabulary items into the decoder graph and bias the language model on-the-fly. Our system achieves 13.5% word error rate on an open-ended dictation task, running with a median speed that is seven times faster than real-time.