LIDSNet: A Lightweight on-device Intent Detection model using Deep Siamese Network

TL;DR

LIDSNet is a lightweight, accurate intent detection model designed for on-device deployment, utilizing a Deep Siamese Network with character-level features and transfer learning to achieve high accuracy with minimal parameters.

Contribution

The paper introduces LIDSNet, a novel lightweight intent detection model that combines a Deep Siamese Network with character-level features and transfer learning for efficient on-device NLP applications.

Findings

Achieves 98.00% accuracy on SNIPS dataset

Contains under 0.59 million parameters

Is 41 times lighter and 30 times faster than MobileBERT on a Galaxy S20

Abstract

Intent detection is a crucial task in any Natural Language Understanding (NLU) system and forms the foundation of a task-oriented dialogue system. To build high-quality real-world conversational solutions for edge devices, there is a need for deploying intent detection model on device. This necessitates a light-weight, fast, and accurate model that can perform efficiently in a resource-constrained environment. To this end, we propose LIDSNet, a novel lightweight on-device intent detection model, which accurately predicts the message intent by utilizing a Deep Siamese Network for learning better sentence representations. We use character-level features to enrich the sentence-level representations and empirically demonstrate the advantage of transfer learning by utilizing pre-trained embeddings. Furthermore, to investigate the efficacy of the modules in our architecture, we conduct an ablation study and arrive at our optimal model. Experimental results prove that LIDSNet achieves state-of-the-art competitive accuracy of 98.00% and 95.97% on SNIPS and ATIS public datasets respectively, with under 0.59M parameters. We further benchmark LIDSNet against fine-tuned BERTs and show that our model is at least 41x lighter and 30x faster during inference than MobileBERT on Samsung Galaxy S20 device, justifying its efficiency on resource-constrained edge devices.