Extreme Model Compression for On-device Natural Language Understanding
This work provides a method for significantly reducing the size of NLU models, enabling their deployment on resource-constrained devices for commercial NLU systems.
This paper addresses the challenge of deploying large Natural Language Understanding (NLU) models on resource-constrained devices by proposing a task-aware, end-to-end compression approach. It achieves a 97.4% compression rate with less than 3.7% degradation in predictive performance on a large-scale commercial NLU system.
In this paper, we propose and experiment with techniques for extreme compression of neural natural language understanding (NLU) models, making them suitable for execution on resource-constrained devices. We propose a task-aware, end-to-end compression approach that performs word-embedding compression jointly with NLU task learning. We show our results on a large-scale, commercial NLU system trained on a varied set of intents with huge vocabulary sizes. Our approach outperforms a range of baselines and achieves a compression rate of 97.4% with less than 3.7% degradation in predictive performance. Our analysis indicates that the signal from the downstream task is important for effective compression with minimal degradation in performance.