Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Obtaining Better Static Word Embeddings Using Contextual Embedding Models

About

The advent of contextual word embeddings -- representations of words which incorporate semantic and syntactic information from their context -- has led to tremendous improvements on a wide variety of NLP tasks. However, recent contextual models have prohibitively high computational cost in many use-cases and are often hard to interpret. In this work, we demonstrate that our proposed distillation method, which is a simple extension of CBOW-based training, allows to significantly improve computational efficiency of NLP applications, while outperforming the quality of existing static embeddings trained from scratch as well as those distilled from previously proposed methods. As a side-effect, our approach also allows a fair comparison of both contextual and static embeddings via standard lexical evaluation tasks.

Prakhar Gupta, Martin Jaggi• 2021

Related benchmarks

TaskDatasetResultRank
Subjectivity ClassificationSubj
Accuracy92.4
266
Sentiment AnalysisMR
Accuracy0.808
142
Sentiment AnalysisCR
Accuracy83.6
123
Word SimilarityWS-353
Spearman Correlation (WS-353)0.7638
54
Word SimilarityRG-65
Spearman Correlation0.8085
35
Word SimilarityRG-65 (test)
Spearman Correlation0.835
33
Word SimilaritySimLex999 (test)
Spearman Correlation0.554
30
Word SimilaritySimVerb-3500 (test)
Spearman Correlation0.473
27
Word SimilarityWS-353 (test)
Spearman Correlation0.7638
18
Word SimilarityWS-353 SIM (test)
Spearman Correlation0.764
15
Showing 10 of 19 rows

Other info

Code

Follow for update