How to learn highly compact yet effective sentence representation? Pre-trained language models have been effective in many NLP tasks. However, these models are often huge and produce large sentence embeddings. Moreover, there is a big performance gap between large and small models. In this paper, we propose Homomorphic Projective Distillation (HPD) to learn compressed sentence embeddings. Our method augments a small Transformer encoder model with learnable projection layers to produce compact representations while mimicking a large pre-trained language model to retain the sentence representation quality. We evaluate our method with different model sizes on both semantic textual similarity (STS) and semantic retrieval (SR) tasks. Experiments show that our method achieves 2.7-4.5 points performance gain on STS tasks compared with previous best representations of the same size. In SR tasks, our method improves retrieval speed (8.2$\times$) and memory usage (8.0$\times$) compared with state-of-the-art large models.
Compressing Sentence Representation for Semantic Retrieval via Homomorphic Projective Distillation
Homomorphic Projective Distillation (HPD) enhances compact sentence representations by augmenting small Transformer models, achieving performance gains on similarity and retrieval tasks while improving speed and memory usage.
- Year
- 2022
- Venue
- Findings (ACL) 2022 5
- Authors
- 4
- Hosting
- Abstract onlyARXIV-DEFAULT
Cite
Notes
Only stored in your browser.
Attribution
- Abstract & full text
- arxiv.org/abs/2203.07687ARXIV-DEFAULT
- TL;DR
- Semantic Scholar