Block-wise Word Embedding Compression Revisited: Better Weighting and Structuring

Jong-Ryul Lee, Yong-Ju Lee, Yong-Hyuk Moon


Abstract
Word embedding is essential for neural network models for various natural language processing tasks. Since the word embedding usually has a considerable size, in order to deploy a neural network model having it on edge devices, it should be effectively compressed. There was a study for proposing a block-wise low-rank approximation method for word embedding, called GroupReduce. Even if their structure is effective, the properties behind the concept of the block-wise word embedding compression were not sufficiently explored. Motivated by this, we improve GroupReduce in terms of word weighting and structuring. For word weighting, we propose a simple yet effective method inspired by the term frequency-inverse document frequency method and a novel differentiable method. Based on them, we construct a discriminative word embedding compression algorithm. In the experiments, we demonstrate that the proposed algorithm more effectively finds word weights than competitors in most cases. In addition, we show that the proposed algorithm can act like a framework through successful cooperation with quantization.
Anthology ID:
2021.findings-emnlp.372
Volume:
Findings of the Association for Computational Linguistics: EMNLP 2021
Month:
November
Year:
2021
Address:
Punta Cana, Dominican Republic
Editors:
Marie-Francine Moens, Xuanjing Huang, Lucia Specia, Scott Wen-tau Yih
Venue:
Findings
SIG:
SIGDAT
Publisher:
Association for Computational Linguistics
Note:
Pages:
4379–4388
Language:
URL:
https://aclanthology.org/2021.findings-emnlp.372
DOI:
10.18653/v1/2021.findings-emnlp.372
Bibkey:
Cite (ACL):
Jong-Ryul Lee, Yong-Ju Lee, and Yong-Hyuk Moon. 2021. Block-wise Word Embedding Compression Revisited: Better Weighting and Structuring. In Findings of the Association for Computational Linguistics: EMNLP 2021, pages 4379–4388, Punta Cana, Dominican Republic. Association for Computational Linguistics.
Cite (Informal):
Block-wise Word Embedding Compression Revisited: Better Weighting and Structuring (Lee et al., Findings 2021)
Copy Citation:
PDF:
https://aclanthology.org/2021.findings-emnlp.372.pdf
Video:
 https://aclanthology.org/2021.findings-emnlp.372.mp4
Code
 etri-edgeai/nn-comp-discblock
Data
SQuADSSTSST-5WikiText-103WikiText-2