Accelerating Natural Language Understanding in Task-Oriented Dialog

TL;DR

This paper presents a lightweight convolutional model for task-oriented dialog understanding that maintains high accuracy while significantly reducing model size and increasing inference speed, enabling on-device deployment.

Contribution

The authors introduce a simple convolutional model with structured pruning that rivals BERT's performance on benchmarks and accelerates inference on CPUs by over 60 times.

Findings

Compressed convolutional model achieves comparable results to BERT on ATIS and Snips.

Model with under 100K parameters performs nearly as well as large transformers.

Multi-task model predicts intents and slots 63x faster than DistilBERT.

Abstract

Task-oriented dialog models typically leverage complex neural architectures and large-scale, pre-trained Transformers to achieve state-of-the-art performance on popular natural language understanding benchmarks. However, these models frequently have in excess of tens of millions of parameters, making them impossible to deploy on-device where resource-efficiency is a major concern. In this work, we show that a simple convolutional model compressed with structured pruning achieves largely comparable results to BERT on ATIS and Snips, with under 100K parameters. Moreover, we perform acceleration experiments on CPUs, where we observe our multi-task model predicts intents and slots nearly 63x faster than even DistilBERT.