English

DIET: Lightweight Language Understanding for Dialogue Systems

Computation and Language 2020-05-12 v3

Abstract

Large-scale pre-trained language models have shown impressive results on language understanding benchmarks like GLUE and SuperGLUE, improving considerably over other pre-training methods like distributed representations (GloVe) and purely supervised approaches. We introduce the Dual Intent and Entity Transformer (DIET) architecture, and study the effectiveness of different pre-trained representations on intent and entity prediction, two common dialogue language understanding tasks. DIET advances the state of the art on a complex multi-domain NLU dataset and achieves similarly high performance on other simpler datasets. Surprisingly, we show that there is no clear benefit to using large pre-trained models for this task, and in fact DIET improves upon the current state of the art even in a purely supervised setup without any pre-trained embeddings. Our best performing model outperforms fine-tuning BERT and is about six times faster to train.

Keywords

Cite

@article{arxiv.2004.09936,
  title  = {DIET: Lightweight Language Understanding for Dialogue Systems},
  author = {Tanja Bunk and Daksh Varshneya and Vladimir Vlasov and Alan Nichol},
  journal= {arXiv preprint arXiv:2004.09936},
  year   = {2020}
}

Comments

v3: Updated results for the best model

R2 v1 2026-06-23T14:59:39.763Z