Transformer to CNN: Label-scarce distillation for efficient text classification

TL;DR

This paper introduces a convolutional student model trained via distillation from a large NLP model, achieving significant speed and size reductions while maintaining or improving performance on text classification tasks.

Contribution

It presents a novel CNN-based student architecture trained through distillation to efficiently perform text classification with limited labeled data.

Findings

300x inference speedup

39x reduction in parameters

Student surpasses teacher in some tasks

Abstract

Significant advances have been made in Natural Language Processing (NLP) modelling since the beginning of 2018. The new approaches allow for accurate results, even when there is little labelled data, because these NLP models can benefit from training on both task-agnostic and task-specific unlabelled data. However, these advantages come with significant size and computational costs. This workshop paper outlines how our proposed convolutional student architecture, having been trained by a distillation process from a large-scale model, can achieve 300x inference speedup and 39x reduction in parameter count. In some cases, the student model performance surpasses its teacher on the studied tasks.